Offer to run CI/PR builds in Visual Studio Team Services

For a long time already, we have Git's source code continuously tested via Travis CI, see e.g. https://travis-ci.org/git/git/builds/421738884. It has served us well, and more and more developers actually pay attention and benefit from the testing this gives us.

It is also an invaluable tool for contributors who can validate their code contributions via PRs on GitHub, e.g. to verify that their tests do actually run on macOS (i.e. with the BSD family of Unix tools instead of the GNU one).

The one sad part about this is the Windows support. Travis lacks it, and we work around that by using Visual Studio Team Services (VSTS) indirectly: one phase in Travis would trigger a build, wait for its log, and then paste that log.

As Git's Windows builds (and tests!) take quite a bit of time, Travis often timed out, or somehow the trigger did not work, and for security reasons (the Windows builds are performed in a private pool of containers), the Windows builds are completely disabled for Pull Requests on GitHub.

One might ask why we did not use Visual Studio Team Services directly. There were a couple of reasons for that:

* most notably, VSTS's build logs could not be viewed anonymously,
* while VSTS had Linux and Windows agents, it lacked macOS agents,
* etc

The main two reasons no longer apply: macOS agents are available now
[https://docs.microsoft.com/en-us/vsts/release-notes/2018/jul-10-vsts], and there is a limited preview of "public projects"
[https://blogs.msdn.microsoft.com/devops/2018/04/27/vsts-public-projects-limited-preview/]
, i.e. it is possible to configure a VSTS project so that anybody can view the logs.

I had secured such a public project for Git for Windows already, and I recently also got one for Git. For now, the latter is hooked up with my personal git.git fork on GitHub, but it is my hope that I convince y'all that these VSTS builds are a good idea, and then hook it up with
https://github.com/git/git.

As a special treat, this patch series adds the ability to present the outcome of Git's test suite as JUnit-style .xml files. This allows the VSTS build to present fun diagrams, trends, and makes it a lot easier to drill down to test failures than before. See for example
https://git.visualstudio.com/git/_build/results?buildId=113&view=ms.vss-test-web.test-result-details
[https://git.visualstudio.com/git/_build/results?buildId=113&view=ms.vss-test-web.test-result-details]
(you can click on the label of the failed test, and then see the detailed output in the right pane).

This patch series took way more time than I had originally planned, but I think that in particular the advanced display of the test results was worth it. Please let me know what you think about this.

Johannes Schindelin (9):
ci: rename the library of common functions
ci/lib.sh: encapsulate Travis-specific things
test-date: add a subcommand to measure times in shell scripts
tests: optionally write results as JUnit-style .xml ci/lib.sh: add support for VSTS CI
Add a build definition for VSTS
tests: include detailed trace logs with --write-junit-xml upon failure
tests: record more stderr with --write-junit-xml in case of failure
README: add a build badge (status of the VSTS build)

 .vsts-ci.yml                   | 296 +++++++++++++++++++++++++++++++++
 README.md                      |   2 +
 ci/install-dependencies.sh     |   5 +-
 ci/{lib-travisci.sh => lib.sh} |  67 ++++++--
 ci/mount-fileshare.sh          |  26 +++
 ci/print-test-failures.sh      |   4 +-
 ci/run-build-and-tests.sh      |   2 +-
 ci/run-linux32-docker.sh       |   2 +-
 ci/run-static-analysis.sh      |   2 +-
 ci/run-windows-build.sh        |   2 +-
 ci/test-documentation.sh       |   3 +-
 t/.gitignore                   |   1 +
 t/helper/test-date.c           |  12 ++
 t/test-lib.sh                  | 142 ++++++++++++++++
 14 files changed, 544 insertions(+), 22 deletions(-)
 create mode 100644 .vsts-ci.yml
 rename ci/{lib-travisci.sh => lib.sh} (61%)
 create mode 100755 ci/mount-fileshare.sh

base-commit: 2f743933341f276111103550fbf383a34dfcfd38
Published-As: https://github.com/gitgitgadget/git/releases/tags/pr-31%2Fdscho%2Fvsts-ci-v1
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-31/dscho/vsts-ci-v1
Pull-Request: https://github.com/gitgitgadget/git/pull/31

gitgitgadget

From: Johannes Schindelin <[email protected]>

The name is hard-coded to reflect that we use Travis CI for continuous testing.

In the next commits, we will extend this to be able use Visual Studio Team Services, too.

So let's adjust the name to make it more generic.

---
 ci/install-dependencies.sh     | 2 +-
 ci/{lib-travisci.sh => lib.sh} | 0
 ci/print-test-failures.sh      | 2 +-
 ci/run-build-and-tests.sh      | 2 +-
 ci/run-linux32-docker.sh       | 2 +-
 ci/run-static-analysis.sh      | 2 +-
 ci/run-windows-build.sh        | 2 +-
 ci/test-documentation.sh       | 2 +-
 8 files changed, 7 insertions(+), 7 deletions(-)
 rename ci/{lib-travisci.sh => lib.sh} (100%)

index 75a9fd2475..961064658e 100755
@@ -3,7 +3,7 @@
 # Install dependencies required to build and test Git on Linux and macOS
 #
 
-. ${0%/*}/lib-travisci.sh
+. ${0%/*}/lib.sh
 
 P4WHENCE=http://filehost.perforce.com/perforce/r$LINUX_P4_VERSION
 LFSWHENCE=https://github.com/github/git-lfs/releases/download/v$LINUX_GIT_LFS_VERSION

similarity index 100%
rename from ci/lib-travisci.sh
rename to ci/lib.sh

index d55460a212..7aef39a2fd 100755
@@ -3,7 +3,7 @@
 # Print output of failing tests
 #
 
-. ${0%/*}/lib-travisci.sh
+. ${0%/*}/lib.sh
 
 # Tracing executed commands would produce too much noise in the loop below.
 set +x

index 2a5bff4a1c..e28ac2fb9a 100755
@@ -3,7 +3,7 @@
 # Build and test Git
 #
 
-. ${0%/*}/lib-travisci.sh
+. ${0%/*}/lib.sh
 
 ln -s "$cache_dir/.prove" t/.prove
 

index 21637903ce..751acfcf8a 100755
@@ -3,7 +3,7 @@
 # Download and run Docker image to build and test 32-bit Git
 #
 
-. ${0%/*}/lib-travisci.sh
+. ${0%/*}/lib.sh
 
 docker pull daald/ubuntu32:xenial
 

index 5688f261d0..dc189c7456 100755
@@ -3,7 +3,7 @@
 # Perform various static code analysis checks
 #
 
-. ${0%/*}/lib-travisci.sh
+. ${0%/*}/lib.sh
 
 make --jobs=2 coccicheck
 

index d99a180e52..a73a4eca0a 100755
@@ -6,7 +6,7 @@
 # supported) and a commit hash.
 #
 
-. ${0%/*}/lib-travisci.sh
+. ${0%/*}/lib.sh
 
 test $# -ne 2 && echo "Unexpected number of parameters" && exit 1
 test -z "$GFW_CI_TOKEN" && echo "GFW_CI_TOKEN not defined" && exit

index a20de9ca12..d3cdbac73f 100755
@@ -3,7 +3,7 @@
 # Perform sanity checks on documentation and build it.
 #
 
-. ${0%/*}/lib-travisci.sh
+. ${0%/*}/lib.sh
 
 gem install asciidoctor
 

gitgitgadget

From: Johannes Schindelin <[email protected]>

The upcoming patches will allow building git.git via VSTS CI, where variable names and URLs look a bit different than in Travis CI.

---
 ci/install-dependencies.sh |  3 ++-
 ci/lib.sh                  | 44 +++++++++++++++++++++++++++-----------
 ci/print-test-failures.sh  |  2 +-
 ci/test-documentation.sh   |  1 +
 4 files changed, 35 insertions(+), 15 deletions(-)

index 961064658e..6d92cc1cba 100755
@@ -28,7 +28,8 @@ osx-clang|osx-gcc)
 	brew update --quiet
 	# Uncomment this if you want to run perf tests:
 	# brew install gnu-time
-	brew install git-lfs gettext
+	test -z "$BREW_INSTALL_PACKAGES" ||
+	eval brew install $BREW_INSTALL_PACKAGES
 	brew link --force gettext
 	brew install caskroom/cask/perforce
 	;;

index 7aef39a2fd..d2045b63a6 100755
@@ -69,7 +69,7 @@ do
 	fi
 done
 
-if [ $combined_trash_size -gt 0 ]
+if [ -n "$TRAVIS_JOB_ID" -a $combined_trash_size -gt 0 ]
 then
 	echo "------------------------------------------------------------------------"
 	echo "Trash directories embedded in this log can be extracted by running:"

index d3cdbac73f..7d0beb2832 100755
@@ -5,6 +5,7 @@
 
 . ${0%/*}/lib.sh
 
+test -n "$ALREADY_HAVE_ASCIIDOCTOR" ||
 gem install asciidoctor
 
 make check-builtins

gitgitgadget

Hi Eric,

Oh right! Fixed in https://github.com/gitgitgadget/git/pull/31 (and I also opened https://github.com/git/git/pull/531 to verify that Travis CI still works).

Thanks,
Dscho

Hi Sebastian,

Good call.

Will fix,
Dscho

From: Johannes Schindelin <[email protected]>

In the next commit, we want to teach Git's test suite to optionally output test results in JUnit-style .xml files. These files contain information about the time spent. So we need a way to measure time.

While we could use date +%s for that, this will give us only seconds, i.e. very coarse-grained timings.

GNU date supports date +%s.%N (i.e. nanosecond-precision output), but there is no equivalent in BSD date (read: on macOS, we would not be able to obtain precise timings).

So let's introduce test-tool date getnanos, with an optional start time, that outputs preciser values.

Granted, it is a bit pointless to try measuring times accurately in shell scripts, certainly to nanosecond precision. But it is better than second-granularity.

---
 t/helper/test-date.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

index a0837371ab..792a805374 100644
@@ -7,6 +7,7 @@ static const char *usage_msg = "\n"
 "  test-tool date parse [date]...\n"
 "  test-tool date approxidate [date]...\n"
 "  test-tool date timestamp [date]...\n"
+"  test-tool date getnanos [start-nanos]\n"
 "  test-tool date is64bit\n"
 "  test-tool date time_t-is64bit\n";
 

@@ -82,6 +83,15 @@ static void parse_approx_timestamp(const char **argv, struct timeval *now)
 	}
 }
 
+static void getnanos(const char **argv, struct timeval *now)
+{
+	double seconds = getnanotime() / 1.0e9;
+
+	if (*argv)
+		seconds -= strtod(*argv, NULL);
+	printf("%lf\n", seconds);
+}
+
 int cmd__date(int argc, const char **argv)
 {
 	struct timeval now;

@@ -108,6 +118,8 @@ int cmd__date(int argc, const char **argv)
 		parse_approxidate(argv+1, &now);
 	else if (!strcmp(*argv, "timestamp"))
 		parse_approx_timestamp(argv+1, &now);
+	else if (!strcmp(*argv, "getnanos"))
+		getnanos(argv+1, &now);
 	else if (!strcmp(*argv, "is64bit"))
 		return sizeof(timestamp_t) == 8 ? 0 : 1;
 	else if (!strcmp(*argv, "time_t-is64bit"))

gitgitgadget

From: Johannes Schindelin <[email protected]>

This will come in handy when publishing the results of Git's test suite during an automated VSTS CI run.

---
 t/.gitignore  |  1 +
 t/test-lib.sh | 98 +++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 99 insertions(+)

index 348715f0e4..91cf5772fe 100644
@@ -2,3 +2,4 @@
 /test-results
 /.prove
 /chainlinttmp
+/out/

gitgitgadget

Hi Eric,

My primary aim is to display the test results in the web interface, see e.g.
https://git.visualstudio.com/git/_build/results?buildId=128&view=ms.vss-test-web.test-result-details

The parser for JUnit XML (and in fact, the JUnit XML schema itself) do not allow for such a warning. If you add a <failure>, then the build fails.

And we do not want the build to fail. Historically, I saw quite a couple of "vanished" breakages depending on the platform where I ran the tests.

Okay. But since we are talking about displaying some chunk of text, I would rather just delete the CR here anyway.

Ciao,
Dscho

From: Johannes Schindelin <[email protected]>

This patch introduces a conditional arm that defines some environment variables and a function that displays the URL given the job id (to identify previous runs for known-good trees).

---
 ci/lib.sh | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)

index 657ff88672..f2f6d70c72 100755
@@ -19,6 +19,29 @@ then
 	BREW_INSTALL_PACKAGES="git-lfs gettext"
 	export GIT_PROVE_OPTS="--timer --jobs 3 --state=failed,slow,save"
 	export GIT_TEST_OPTS="--verbose-log -x --immediate"
+elif test -n "$SYSTEM_TASKDEFINITIONSURI"
+then
+	# We are running in VSTS CI
+	CI_BRANCH="$BUILD_SOURCEBRANCH"
+	CI_COMMIT="$BUILD_SOURCEVERSION"
+	CI_JOB_ID="$BUILD_BUILDID"
+	CI_JOB_NUMBER="$BUILD_BUILDNUMBER"
+	CI_OS_NAME="$(echo "$AGENT_OS" | tr A-Z a-z)"
+	test darwin != "$CI_OS_NAME" || CI_OS_NAME=osx
+	CI_REPO_SLUG="$(expr "$BUILD_REPOSITORY_URI" : '.*/\([^/]*/[^/]*\)$')"
+	CC="${CC:-gcc}"
+
+	# use a subdirectory of the cache dir (because the file share is shared
+	# among *all* phases)
+	cache_dir="$HOME/vsts-cache/$SYSTEM_PHASENAME"
+
+	url_for_job_id () {
+		echo "$SYSTEM_TASKDEFINITIONSURI$SYSTEM_TEAMPROJECT/_build/results?buildId=$1"
+	}
+
+	BREW_INSTALL_PACKAGES=
+	export GIT_PROVE_OPTS="--timer --jobs 10 --state=failed,slow,save"
+	export GIT_TEST_OPTS="--quiet --write-junit-xml"
 fi
 
 skip_branch_tip_with_tag () {

gitgitgadget

From: Johannes Schindelin <[email protected]>

This commit adds a .vsts-ci.yml which is Visual Studio Team Services' equivalent to Travis CI's .travis.yml.

To make things a bit easier to understand, we refrain from using the matrix feature here because (while it is powerful) it can be a bit confusing to users who are not familiar with CI setups. Therefore, we use a separate phase even for similar configurations (such as GCC vs Clang on Linux, GCC vs Clang on macOS).

Also, we make use of the shiny new feature we just introduced where the test suite can output JUnit-style .xml files. This information is made available in a nice UI that allows the viewer to filter by phase and/or test number, and to see trends such as: number of (failing) tests, time spent running the test suite, etc.

---
 .vsts-ci.yml          | 296 ++++++++++++++++++++++++++++++++++++++++++
 ci/mount-fileshare.sh |  26 ++++
 2 files changed, 322 insertions(+)
 create mode 100644 .vsts-ci.yml
 create mode 100755 ci/mount-fileshare.sh

new file mode 100644
index 0000000000..ff1d952df9
@@ -0,0 +1,296 @@
+resources:
+- repo: self
+  fetchDepth: 1
+
+phases:
+- phase: linux_clang
+  displayName: linux-clang
+  condition: succeeded()
+  queue:
+    name: Hosted Ubuntu 1604
+  steps:
+  - bash: |
+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
+
+       sudo apt-get update
+       sudo apt-get -y install git gcc make libssl-dev libcurl4-openssl-dev libexpat-dev tcl tk gettext git-email zlib1g-dev apache2-bin
+
+       export CC=clang
+
+       ci/install-dependencies.sh
+       ci/run-build-and-tests.sh || {
+           ci/print-test-failures.sh
+           exit 1
+       }
+
+       sudo umount "$HOME/vsts-cache" || exit 1
+    displayName: 'ci/run-build-and-tests.sh'
+  - task: PublishTestResults@2
+    displayName: 'Publish Test Results **/TEST-*.xml'
+    inputs:
+      mergeTestResults: true
+      testRunTitle: 'linux-clang'
+      platform: Linux
+      publishRunAttachments: false
+    condition: succeededOrFailed()
+
+- phase: linux_gcc
+  displayName: linux-gcc
+  condition: succeeded()
+  queue:
+    name: Hosted Ubuntu 1604
+  steps:
+  - bash: |
+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
+
+       sudo apt-get update
+       sudo apt-get -y install git gcc make libssl-dev libcurl4-openssl-dev libexpat-dev tcl tk gettext git-email zlib1g-dev apache2-bin
+
+       ci/install-dependencies.sh
+       ci/run-build-and-tests.sh || {
+           ci/print-test-failures.sh
+           exit 1
+       }
+
+       sudo umount "$HOME/vsts-cache" || exit 1
+    displayName: 'ci/run-build-and-tests.sh'
+  - task: PublishTestResults@2
+    displayName: 'Publish Test Results **/TEST-*.xml'
+    inputs:
+      mergeTestResults: true
+      testRunTitle: 'linux-gcc'
+      platform: Linux
+      publishRunAttachments: false
+    condition: succeededOrFailed()
+
+- phase: osx_clang
+  displayName: osx-clang
+  condition: succeeded()
+  queue:
+    name: Hosted macOS
+  steps:
+  - bash: |
+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
+
+       export CC=clang
+
+       ci/install-dependencies.sh
+       ci/run-build-and-tests.sh || {
+           ci/print-test-failures.sh
+           exit 1
+       }
+
+       umount "$HOME/vsts-cache" || exit 1
+    displayName: 'ci/run-build-and-tests.sh'
+  - task: PublishTestResults@2
+    displayName: 'Publish Test Results **/TEST-*.xml'
+    inputs:
+      mergeTestResults: true
+      testRunTitle: 'osx-clang'
+      platform: macOS
+      publishRunAttachments: false
+    condition: succeededOrFailed()
+
+- phase: osx_gcc
+  displayName: osx-gcc
+  condition: succeeded()
+  queue:
+    name: Hosted macOS
+  steps:
+  - bash: |
+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
+
+       ci/install-dependencies.sh
+       ci/run-build-and-tests.sh || {
+           ci/print-test-failures.sh
+           exit 1
+       }
+
+       umount "$HOME/vsts-cache" || exit 1
+    displayName: 'ci/run-build-and-tests.sh'
+  - task: PublishTestResults@2
+    displayName: 'Publish Test Results **/TEST-*.xml'
+    inputs:
+      mergeTestResults: true
+      testRunTitle: 'osx-gcc'
+      platform: macOS
+      publishRunAttachments: false
+    condition: succeededOrFailed()
+
+- phase: gettext_poison
+  displayName: GETTEXT_POISON
+  condition: succeeded()
+  queue:
+    name: Hosted Ubuntu 1604
+  steps:
+  - bash: |
+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
+
+       sudo apt-get update
+       sudo apt-get -y install git gcc make libssl-dev libcurl4-openssl-dev libexpat-dev tcl tk gettext git-email zlib1g-dev
+
+       export jobname=GETTEXT_POISON
+
+       ci/run-build-and-tests.sh || {
+           ci/print-test-failures.sh
+           exit 1
+       }
+
+       sudo umount "$HOME/vsts-cache" || exit 1
+    displayName: 'ci/run-build-and-tests.sh'
+  - task: PublishTestResults@2
+    displayName: 'Publish Test Results **/TEST-*.xml'
+    inputs:
+      mergeTestResults: true
+      testRunTitle: 'gettext-poison'
+      platform: Linux
+      publishRunAttachments: false
+    condition: succeededOrFailed()
+
+- phase: windows
+  displayName: Windows
+  condition: succeeded()
+  queue:
+    name: Hosted VS2017
+    timeoutInMinutes: 240
+  steps:
+  - powershell: |
+       # Helper to check the error level of the latest command (exit with error when appropriate)
+       function c() { if (!$?) { exit(1) } }
+
+       net use s: \\gitfileshare.file.core.windows.net\vsts-cache "$(gitfileshare.pwd)" /user:AZURE\gitfileshare /persistent:no; c
+       cmd /c mklink /d "$(Build.SourcesDirectory)\vsts-cache" S:\; c
+
+       # Add build agent's MinGit to PATH
+       $env:PATH = $env:AGENT_HOMEDIRECTORY +"\externals\\git\cmd;" +$env:PATH
+
+       # Helper to initialize (or update) a Git worktree
+       function init ($path, $url, $set_origin) {
+           if (Test-Path $path) {
+               cd $path; c
+               if (Test-Path .git) {
+                   git init; c
+               } else {
+                   git status
+               }
+           } else {
+               git init $path; c
+               cd $path; c
+           }
+           git config core.autocrlf false; c
+           git config core.untrackedCache true; c
+           if (($set_origin -ne 0) -and !(git config remote.origin.url)) {
+               git remote add origin $url; c
+           }
+           git fetch --depth=1 $url master; c
+           git reset --hard FETCH_HEAD; c
+           git clean -df; c
+       }
+
+       # Initialize Git for Windows' SDK
+       $sdk_path = "$(Build.SourcesDirectory)\git-sdk-64"
+       init "$sdk_path" "https://git-for-windows.visualstudio.com/_git/git-sdk-64" 0
+       init usr\src\build-extra https://github.com/git-for-windows/build-extra 1
+
+       cd "$(Build.SourcesDirectory)"; c
+
+       $env:HOME = "$(Build.SourcesDirectory)"
+       $env:MSYSTEM = "MINGW64"
+       git-sdk-64\git-cmd --command=usr\\bin\\bash.exe -lc @"
+         . ci/lib.sh
+
+         make -j10 DEVELOPER=1 NO_PERL=1 || exit 1
+         NO_PERL=1 NO_SVN_TESTS=1 GIT_TEST_OPTS=\"--quiet --write-junit-xml\" time make -j15 -k DEVELOPER=1 test || {
+           NO_PERL=1 NO_SVN_TESTS=1 GIT_TEST_OPTS=\"-i -v -x\" make -k failed; exit 1
+         }
+
+         save_good_tree
+       "@
+
+       cmd /c rmdir "$(Build.SourcesDirectory)\vsts-cache"
+    displayName: 'build & test'
+  - task: PublishTestResults@2
+    displayName: 'Publish Test Results **/TEST-*.xml'
+    inputs:
+      mergeTestResults: true
+      testRunTitle: 'windows'
+      platform: Windows
+      publishRunAttachments: false
+    condition: succeededOrFailed()
+
+- phase: linux32
+  displayName: Linux32
+  condition: succeeded()
+  queue:
+    name: Hosted Ubuntu 1604
+  steps:
+  - bash: |
+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
+
+       sudo apt-get update
+       sudo apt-get -y install \
+           apt-transport-https \
+           ca-certificates \
+           curl \
+           software-properties-common
+       curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add -
+       sudo add-apt-repository \
+          "deb [arch=amd64] https://download.docker.com/linux/ubuntu \
+          $(lsb_release -cs) \
+          stable"
+       sudo apt-get update
+       sudo apt-get -y install docker-ce
+
+       sudo AGENT_OS="$AGENT_OS" BUILD_BUILDNUMBER="$BUILD_BUILDNUMBER" BUILD_REPOSITORY_URI="$BUILD_REPOSITORY_URI" BUILD_SOURCEBRANCH="$BUILD_SOURCEBRANCH" BUILD_SOURCEVERSION="$BUILD_SOURCEVERSION" SYSTEM_PHASENAME="$SYSTEM_PHASENAME" SYSTEM_TASKDEFINITIONSURI="$SYSTEM_TASKDEFINITIONSURI" SYSTEM_TEAMPROJECT="$SYSTEM_TEAMPROJECT" CC=$CC MAKEFLAGS=-j3 bash -lxc ci/run-linux32-docker.sh || exit 1
+
+       sudo chmod a+r t/out/TEST-*.xml
+
+       sudo umount "$HOME/vsts-cache" || exit 1
+    displayName: 'ci/run-linux32-docker.sh'
+  - task: PublishTestResults@2
+    displayName: 'Publish Test Results **/TEST-*.xml'
+    inputs:
+      mergeTestResults: true
+      testRunTitle: 'linux32'
+      platform: Linux
+      publishRunAttachments: false
+    condition: succeededOrFailed()
+
+- phase: static_analysis
+  displayName: StaticAnalysis
+  condition: succeeded()
+  queue:
+    name: Hosted Ubuntu 1604
+  steps:
+  - bash: |
+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
+
+       sudo apt-get update
+       sudo apt-get install -y coccinelle
+
+       export jobname=StaticAnalysis
+
+       ci/run-static-analysis.sh || exit 1
+
+       sudo umount "$HOME/vsts-cache" || exit 1
+    displayName: 'ci/run-static-analysis.sh'
+
+- phase: documentation
+  displayName: Documentation
+  condition: succeeded()
+  queue:
+    name: Hosted Ubuntu 1604
+  steps:
+  - bash: |
+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
+
+       sudo apt-get update
+       sudo apt-get install -y asciidoc xmlto asciidoctor
+
+       export ALREADY_HAVE_ASCIIDOCTOR=yes.
+       export jobname=Documentation
+
+       ci/test-documentation.sh || exit 1
+
+       sudo umount "$HOME/vsts-cache" || exit 1
+    displayName: 'ci/test-documentation.sh'

new file mode 100755
index 0000000000..5fb5f74b70
@@ -0,0 +1,26 @@
+#!/bin/sh
+
+die () {
+	echo "$*" >&2
+	exit 1
+}
+
+test $# = 4 ||
+die "Usage: $0 <share> <username> <password> <mountpoint"
+
+mkdir -p "$4" || die "Could not create $4"
+
+case "$(uname -s)" in
+Linux)
+	sudo mount -t cifs -o vers=3.0,username="$2",password="$3",dir_mode=0777,file_mode=0777,serverino "$1" "$4"
+	;;
+Darwin)
+	pass="$(echo "$3" | sed -e 's/\//%2F/g' -e 's/+/%2B/g')" &&
+	mount -t smbfs,soft "smb://$2:$pass@${1#//}" "$4"
+	;;
+*)
+	die "No support for $(uname -s)"
+	;;
+esac ||
+die "Could not mount $4"
+

gitgitgadget

From: Johannes Schindelin <[email protected]>

The JUnit XML format lends itself to be presented in a powerful UI, where you can drill down to the information you are interested in very quickly.

For test failures, this usually means that you want to see the detailed trace of the failing tests.

With Travis CI, we passed the --verbose-log option to get those traces. However, that seems excessive, as we do not need/use the logs in almost all of those cases: only when a test fails do we have a way to include the trace.

So let's do something different in VSTS: let's run all the tests with --quiet first, and only if a failure is encountered, try to trace the commands as they are executed.

Of course, we cannot turn on --verbose-log after the fact. So let's just re-run the test with all the same options, adding --verbose-log. And then munging the output file into the JUnit XML on the fly.

Note: there is an off chance that re-running the test in verbose mode "fixes" the failures (and this does happen from time to time!). That is a possibility we should be able to live with. Ideally, we would label this as "Passed upon rerun", and that outcome even exists on VSTS, but it is not available when using the JUnit XML format for now:
https://github.com/Microsoft/vsts-agent/blob/master/src/Agent.Worker/TestResults/JunitResultReader.cs

This patch contains a slightly inelegant workaround for the p4 and git-daemon tests: when we try to re-run the p4/git-daemon tests after the daemon has been started already, we need to kill said daemon. We do this by detecting the presence of the kill_p4d and stop_git_daemon shell functions and calling them if available.

---
 t/test-lib.sh | 40 +++++++++++++++++++++++++++++++++++++++-
 1 file changed, 39 insertions(+), 1 deletion(-)

index 50a65a600e..ea4ed250cc 100644
@@ -85,6 +85,13 @@ done,*)
 	test "$(cat "$BASE.exit")" = 0
 	exit
 	;;
+*' --write-junit-xml '*)
+	# record how to call this script *with* --verbose-log, in case
+	# we encounter a breakage
+	junit_rerun_options_sq="$(printf '%s\n' "$0" --verbose-log -x "$@" |
+		sed -e "s/'/'\\\\''/g" -e "s/^/'/" -e "s/\$/'/" |
+		tr '\012' ' ')"
+	;;
 esac
 
 # For repeatability, reset the environment to known value.

@@ -445,10 +452,37 @@ test_ok_ () {
 test_failure_ () {
 	if test -n "$write_junit_xml"
 	then
+		if test -z "$GIT_TEST_TEE_OUTPUT_FILE"
+		then
+			case "$(type kill_p4d 2>/dev/null | head -n 1)" in
+			*function*) kill_p4d;;
+			esac
+
+			case "$(type stop_git_daemon 2>/dev/null |
+				head -n 1)" in
+			*function*) stop_git_daemon;;
+			esac
+
+			# re-run with --verbose-log
+			echo "# Re-running: $junit_rerun_options_sq" >&2
+
+			cd "$TEST_DIRECTORY" &&
+			eval "${TEST_SHELL_PATH}" "$junit_rerun_options_sq" \
+				>/dev/null 2>&1
+			status=$?
+
+			say_color "" "$(test 0 = $status ||
+				echo "not ")ok $test_count - (re-ran with trace)"
+			say "1..$test_count"
+			GIT_EXIT_OK=t
+			exit $status
+		fi
+
 		junit_insert="<failure message=\"not ok $test_count -"
 		junit_insert="$junit_insert $(xml_attr_encode "$1")\">"
 		junit_insert="$junit_insert $(xml_attr_encode \
-			"$(printf '%s\n' "$@" | sed 1d)")"
+			"$(cat "$GIT_TEST_TEE_OUTPUT_FILE")")"
+		>"$GIT_TEST_TEE_OUTPUT_FILE"
 		junit_insert="$junit_insert</failure>"
 		write_junit_xml_testcase "$1" "      $junit_insert"
 	fi

@@ -733,6 +767,10 @@ test_start_ () {
 	if test -n "$write_junit_xml"
 	then
 		junit_start=$(test-tool date getnanos)
+
+		# truncate output
+		test -z "$GIT_TEST_TEE_OUTPUT_FILE" ||
+		>"$GIT_TEST_TEE_OUTPUT_FILE"
 	fi
 }
 

gitgitgadget

Hi Eric,

Or we could implement test_atexit (similar to test_when_finished, but to be executed at test_done time). I guess that's what the p4 and daemon tests really needed to begin with (and probably also the apache2-using tests).

Now that you mention it... I agree. This is getting long.

In the short run, I have two things to consider, though: I want to make this work first, then think about introducing a layer of abstraction, and I want to go on vacation tomorrow.

So I agree that this is something to be considered in the long run, i.e. not right now ;-)

Thanks,
Dscho

Is this re-running just an individual test on its own or all the tests within a single file?

The latter shouldn't need this at all. And the former, I'm not sure will actually work - most of the tests assume some particular p4 state. But perhaps I'm missing something?

I also think it does look kind of ugly. And if there's one thing I've learned, it's that the ugly hack you write today with the words "we'll tidy this up later" goes on to live with you forever!

Hi Luke,

Upon encountering a failed test, it is re-running the entire test script afresh.

Please do not let me die dumb. In other words, I would love for you to explain what exactly you mean by that sentence.

No, the former would not work at all. Not only for the p4 tests: Git's tests frequently commit the deadly sin of relying on output of one another (wreaking havoc e.g. when test cases are skipped due to missing prerequisites, and latter test cases relying on their output). It is not the only thing that is wrong with the test suite, of course.

Okay.

(And having read lib-git-p4.sh, I kind of see where you learned that.)

But maybe you also have some splendid idea what to do instead? Because doing something about it, that we need. We can't just say "oh, the only solution we found is ugly, so let's not do it at all".

I am even going so far as to say: unless you have a better idea, it is pretty detrimental to criticize the current approach. It is the opposite of constructive.

So let's hear some ideas how to improve the situation, m'kay?

Just as a reminder, this is the problem I want to solve: I want to run the tests in a light-weight manner, with minimal output, and only in case of an error do I want to crank up the verbosity. Instead of wasting most of the effort to log everything and then throwing it away in most of the common cases, I suggest to re-run the entire test.

What do you suggest to solve this?

Ciao,
Johannes

Just re-run the script. You shouldn't need to kill p4d, as each script starts up its own instance of p4d, and shuts it down when it exits.

$ cd t
$ ./t9800-git-p4-basic.sh
Ctrl^C
$ ./t9800-git-p4-basic.sh -v

There's a cleanup() function in lib-git-p4.sh which kills the p4d server, and that's invoked via:

trap cleanup EXIT

That's the only cleanup that each of the scripts require AFAIK.

I don't know about any other tests, but the git-p4 tests don't take any longer in verbose mode. So one simple solution is to just run it in verbose mode - unless there are other tests which have more overhead.

The trap/exit/cleanup method that the git-p4 tests already use would seem to be ideally suited to cleaning up everything on exit.

There might be some specific tests where this doesn't quite work out, if you let me know what they are I can have a look at fixing them for you.

Thanks,
Luke

Hi Luke,

That Ctrl^C is not possible in the automated builds (which are the subject of this patch series).

What *is* possible is to detect in the test script that there was a breakage, and then re-run the test script.

Unfortunately, at this stage the first test script has not exited yet, so your p4d has not exited yet, and we have to clean up in some way before re-running (because the port in use is determined by the test script's number, so we really need to kill the old daemon).

And this is exactly what my patches try to achieve here.

Good.

However, we really will want to consider introducing something consistent here, something that works for *all* test scripts. IOW if any test script wants to start a process for the life-time of that script, and needs to kill that process at the end, we will want to have a consistent, documented way to register a function (or commands) to be called at the end.

And that function (or commands) need to be run also when stopping everything in preparation for a re-run.

That statement is too general to be correct. It may be the case in your setup, but I found that even a redirected, chatty stderr can slow down things on Windows substantially, probably due to the POSIX emulation going on in the background that needs to prepare for all kinds of eventualities (that we do not even need, but the MSYS2 runtime has no way of knowing that).

It is a bit of an overhead. As the entire test run on Windows takes *quite* a while (due to said POSIX emulation issues), I was not able to quantify it exactly.

Yes. On exit. But I cannot exit here, not before re-running the script.

Thanks. I will let you know when I encounter other tests than the p4 and git-daemon tests (which I fixed in this patch series).

Ciao,
Dscho

What about the very different approach of capturing the full "verbose" output the executed tests in addition to whatever is actually output to the terminal? If a test fails, then (and only then) you can insert the captured verbose output into the JUnit XML file. This way (if we always have the full verbose output at hand), you don't need to re-run the test at all.

I've cc:'d Peff and Jonathan since I believe they are more familiar with how all the capturing / output-redirection works in the test suite.

I don't think there's much to know beyond what you wrote. The "--verbose" case does not really cost any more than the non-verbose one, because the only difference is whether output goes to a file versus /dev/null. But the commands all call write() regardless.

For --verbose-log, it does of course cost a little extra to run one tee per script, and to write a few kb of logs. I doubt those are measurable versus the rest of a script run, but I'm happy to be disproven by numbers. There are some gymnastics done to re-exec the test script with the same shell, but AFAIK those are robust and don't cost a lot (again, one extra process per script run).

I'm not overly concerned about the cost of re-running a test, since the idea is that failed tests should be rare. I would be a little worried about flaky tests mysteriously righting themselves on the second run (so you know a failure happened, but you have no good output to describe it).

I do agree that a test_atexit() cleanup would make a lot more sense than what's in the original patch. And that's nicer than the exit trap we're using already, because you may note that each caller has to manually restore the original 'die' handler that test-lib.sh installs.

That would also help with bitrot. If this funky cleanup case only causes problems with junit output, then other people are likely to forget to do it, and the mess falls onto the junit folks (i.e., Dscho). But if the tests start _relying_ on test_atexit() being called (i.e., don't call stop_git_daemon manually, but assume that the atexit handler does so), then the responsible authors are a lot more likely to notice and fix it early.

-Peff

Hi Peff,

This really only holds true on systems where redirected output is essentially a no-op.

For the MSYS2-emulated case, this is incorrect. And as that case is already suffering, time-wise, I would be loathe to pile more onto it.

I tried, and failed, to quantify this properly. Essentially because a single test run takes over 1h15m, so I tried to do it in the cloud, but those timings are too variable for anything approaching an accurate measurement.

It *is* actually a bit worse here (but only in the case of failures, which should be the minority of the cases): imagine an expensive test like t0027 that takes something like 14 minutes to run, and then a test failure near the end: re-running with verbose output will cost another 14 minutes.

However, the benefit of allowing to pass flakey tests outweighs the costs here, if you ask me.

This is actually a benefit.

Pretty much all the times I can remember a test being flakey (with the one notable exception of the O_APPEND issue with GIT_TRACE), the *tests* were buggy.

Plus, it is less concerning if a test fails occasionally vs a test that fails all the time.

My plan is to indicate the outcome of "Passed upon rerun" in the test output eventually, as soon as there is server-side support for it.

(Sidenote: there is technically already server-side support for it, but we would have to generate Visual Studio-style test output, and I had the hunch that some die-hard Linuxers among the core Git developers would take objection to that.)

Okay, then. I will work on this.

Ciao,
Dscho

Hi Eric,

I fear it is not really possible to do a "verbose but not really" mode. I want the console output to be quiet, there is no use in chatting up the build log with these messages, as we have to run the tests in parallel, so the output would be utterly hard to interpret anyway. At the same time, I want verbose output for use in the test results. It is not really possible to tee all output, then "quiet down" the part that makes it into the log.

Yep. That's what my patch series does.

If a test fails, then (and only then) I re-run the script with verbose output that is immediately moved into that JUnit XML file.

But that way, if we always have the full verbose output, the build log will be unnecessarily verbose and confusing.

Okay.

Ciao,
Dscho

P.S.: An unintended side effect of re-running the tests is to identify flakey tests. I do not yet have a way to represent this outcome in the test result, but I deem this an additional benefit in favor of keeping my current strategy.

From: Johannes Schindelin <[email protected]>

Sometimes, failures in a test case are actually caused by issues in earlier test cases.

To make it easier to see those issues, let's attach the output from before the failing test case (i.e. stdout/stderr since the previous failing test case, or the start of the test script). This will be visible in the "Attachments" of the details of the failed test.

---
 t/test-lib.sh | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)

index ea4ed250cc..6fc03d5a3b 100644
@@ -484,6 +484,9 @@ test_failure_ () {
 			"$(cat "$GIT_TEST_TEE_OUTPUT_FILE")")"
 		>"$GIT_TEST_TEE_OUTPUT_FILE"
 		junit_insert="$junit_insert</failure>"
+		junit_insert="$junit_insert<system-err>$(xml_attr_encode \
+			"$(cat "$GIT_TEST_TEE_OUTPUT_FILE.err")")</system-err>"
+		>"$GIT_TEST_TEE_OUTPUT_FILE.err"
 		write_junit_xml_testcase "$1" "      $junit_insert"
 	fi
 	test_failure=$(($test_failure + 1))

@@ -768,9 +771,12 @@ test_start_ () {
 	then
 		junit_start=$(test-tool date getnanos)
 
-		# truncate output
-		test -z "$GIT_TEST_TEE_OUTPUT_FILE" ||
-		>"$GIT_TEST_TEE_OUTPUT_FILE"
+		# append to future <system-err>; truncate output
+		test -z "$GIT_TEST_TEE_OUTPUT_FILE" || {
+			cat "$GIT_TEST_TEE_OUTPUT_FILE" \
+				>>"$GIT_TEST_TEE_OUTPUT_FILE.err"
+			>"$GIT_TEST_TEE_OUTPUT_FILE"
+		}
 	fi
 }
 

gitgitgadget

From: Johannes Schindelin <[email protected]>

Just like so many other OSS projects, we now also have a build badge.

---
 README.md | 2 ++
 1 file changed, 2 insertions(+)

index f920a42fad..f260e78042 100644
@@ -1,3 +1,5 @@
+[![Build Status](https://git.visualstudio.com/git/_apis/build/status/test-git.git)](https://git.visualstudio.com/git/_build/latest?definitionId=2)
+
 Git - fast, scalable, distributed revision control system
 =========================================================
 

gitgitgadget

I'm sorry if this has been discussed before, but as this recap doesn't mention it: Has AppVeyor been considered as an option? It seems to be the defacto standard for Windows CI for projects on GitHub.

Sebastian Schuberth

There was
https://public-inbox.org/git/[email protected]/
and
https://public-inbox.org/git/[email protected]/

For a long time already, we have Git's source code continuously tested via Travis CI, see e.g. https://travis-ci.org/git/git/builds/421738884. It has served us well, and more and more developers actually pay attention and benefit from the testing this gives us.

It is also an invaluable tool for contributors who can validate their code contributions via PRs on GitHub, e.g. to verify that their tests do actually run on macOS (i.e. with the BSD family of Unix tools instead of the GNU one).

The one sad part about this is the Windows support. Travis lacks it, and we work around that by using Azure Pipelines (the CI part of Azure DevOps, formerly known as Visual Studio Team Services) indirectly: one phase in Travis would trigger a build, wait for its log, and then paste that log.

As Git's Windows builds (and tests!) take quite a bit of time, Travis often timed out, or somehow the trigger did not work, and for security reasons (the Windows builds are performed in a private pool of containers), the Windows builds are completely disabled for Pull Requests on GitHub.

One might ask why we did not use Azure Pipelines directly. There were a couple of reasons for that:

* most notably, Azure Pipelines' build logs could not be viewed anonymously,
* while Azure Pipelines had Linux and Windows agents, it lacked macOS agents,
* etc

The main two reasons no longer apply: macOS agents are available now
[https://docs.microsoft.com/en-us/azure/devops/release-notes/2018/jul-10-vsts]
, and there is a limited preview of "public projects"
[https://blogs.msdn.microsoft.com/devops/2018/04/27/vsts-public-projects-limited-preview/]
, i.e. it is possible to configure a Azure Pipelines project so that anybody can view the logs.

I had secured such a public project for Git for Windows already, and I recently also got one for Git. For now, the latter is hooked up with my personal git.git fork on GitHub, but it is my hope that I convince y'all that these Azure Pipelines builds are a good idea, and then hook it up with
https://github.com/git/git.

As a special treat, this patch series adds the ability to present the outcome of Git's test suite as JUnit-style .xml files. This allows the Azure Pipelines build to present fun diagrams, trends, and makes it a lot easier to drill down to test failures than before. See for example
https://dev.azure.com/git/git/_build/results?buildId=113&view=ms.vss-test-web.test-result-details
[https://dev.azure.com/git/git/_build/results?buildId=113&view=ms.vss-test-web.test-result-details]
(you can click on the label of the failed test, and then see the detailed output in the right pane).

This patch series took way more time than I had originally planned, but I think that in particular the advanced display of the test results was worth it. Please let me know what you think about this.

Changes since v1:

* Removed a superfluous eval.
* Added the commit that fixes the Travis PR builds targeting master that just happens to be tagged (see
https://travis-ci.org/git/git/jobs/424276413 for an incorrectly-skipped build).
* The commit messages and the cover letter now reflect the name change from Visual Studio Team Services to Azure DevOps (and in particular, Azure Pipelines for the automated builds).
* Now we're using test_atexit (which we introduced for that purpose) instead of hard-coding kill_p4d and stop_git_daemon.
* The build should now also succeed for Pull Requests (where secret variables are not available, for security reasons, and as a consequence the file share cannot be mounted).
* The shell scripted parts now use proper && chains.

Johannes Schindelin (13):
ci: rename the library of common functions
ci/lib.sh: encapsulate Travis-specific things
test-date: add a subcommand to measure times in shell scripts
tests: optionally write results as JUnit-style .xml ci/lib.sh: add support for Azure Pipelines
Add a build definition for Azure DevOps
tests: introduce test_atexit
git-daemon: use test_atexit in the tests
git-p4: use test_atexit to kill the daemon
tests: include detailed trace logs with --write-junit-xml upon failure
tests: record more stderr with --write-junit-xml in case of failure
README: add a build badge (status of the Azure Pipelines build)
travis: fix skipping tagged releases

 README.md                                  |   2 +
 azure-pipelines.yml                        | 319 +++++++++++++++++++++
 ci/install-dependencies.sh                 |   5 +-
 ci/{lib-travisci.sh => lib.sh}             |  67 ++++-
 ci/mount-fileshare.sh                      |  26 ++
 ci/print-test-failures.sh                  |   4 +-
 ci/run-build-and-tests.sh                  |   2 +-
 ci/run-linux32-docker.sh                   |   2 +-
 ci/run-static-analysis.sh                  |   2 +-
 ci/run-windows-build.sh                    |   2 +-
 ci/test-documentation.sh                   |   3 +-
 t/.gitignore                               |   1 +
 t/helper/test-date.c                       |  12 +
 t/interop/i5500-git-daemon.sh              |   1 -
 t/lib-git-daemon.sh                        |   3 +-
 t/lib-git-p4.sh                            |  10 +-
 t/t0000-basic.sh                           |  20 ++
 t/t5570-git-daemon.sh                      |   1 -
 t/t9800-git-p4-basic.sh                    |   4 -
 t/t9801-git-p4-branch.sh                   |   4 -
 t/t9802-git-p4-filetype.sh                 |   4 -
 t/t9803-git-p4-shell-metachars.sh          |   4 -
 t/t9804-git-p4-label.sh                    |   4 -
 t/t9805-git-p4-skip-submit-edit.sh         |   4 -
 t/t9806-git-p4-options.sh                  |   5 -
 t/t9807-git-p4-submit.sh                   |   4 -
 t/t9808-git-p4-chdir.sh                    |   4 -
 t/t9809-git-p4-client-view.sh              |   4 -
 t/t9810-git-p4-rcs.sh                      |   4 -
 t/t9811-git-p4-label-import.sh             |   5 -
 t/t9812-git-p4-wildcards.sh                |   4 -
 t/t9813-git-p4-preserve-users.sh           |   4 -
 t/t9814-git-p4-rename.sh                   |   4 -
 t/t9815-git-p4-submit-fail.sh              |   4 -
 t/t9816-git-p4-locked.sh                   |   4 -
 t/t9817-git-p4-exclude.sh                  |   4 -
 t/t9818-git-p4-block.sh                    |   4 -
 t/t9819-git-p4-case-folding.sh             |   4 -
 t/t9820-git-p4-editor-handling.sh          |   4 -
 t/t9821-git-p4-path-variations.sh          |   4 -
 t/t9822-git-p4-path-encoding.sh            |   4 -
 t/t9823-git-p4-mock-lfs.sh                 |   4 -
 t/t9824-git-p4-git-lfs.sh                  |   4 -
 t/t9825-git-p4-handle-utf16-without-bom.sh |   4 -
 t/t9826-git-p4-keep-empty-commits.sh       |   4 -
 t/t9827-git-p4-change-filetype.sh          |   4 -
 t/t9828-git-p4-map-user.sh                 |   4 -
 t/t9829-git-p4-jobs.sh                     |   4 -
 t/t9830-git-p4-symlink-dir.sh              |   4 -
 t/t9831-git-p4-triggers.sh                 |   4 -
 t/t9832-unshelve.sh                        |   4 -
 t/t9833-errors.sh                          |   5 -
 t/test-lib-functions.sh                    |  29 ++
 t/test-lib.sh                              | 140 +++++++++
 54 files changed, 616 insertions(+), 174 deletions(-)
 create mode 100644 azure-pipelines.yml
 rename ci/{lib-travisci.sh => lib.sh} (61%)
 create mode 100755 ci/mount-fileshare.sh

base-commit: 5a0cc8aca797dbd7d2be3b67458ff880ed45cddf
Published-As: https://github.com/gitgitgadget/git/releases/tags/pr-31%2Fdscho%2Fvsts-ci-v2
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-31/dscho/vsts-ci-v2
Pull-Request: https://github.com/gitgitgadget/git/pull/31

Range-diff vs v1:

1: 858b80bfcf ! 1: c963184510 ci: rename the library of common functions
@@ -5,8 +5,8 @@
The name is hard-coded to reflect that we use Travis CI for continuous testing.

- In the next commits, we will extend this to be able use Visual Studio
- Team Services, too.
+ In the next commits, we will extend this to be able use Azure DevOps,
+ too.

So let's adjust the name to make it more generic.

2: 18e6beec5f ! 2: 815152e0f5 ci/lib.sh: encapsulate Travis-specific things
@@ -2,8 +2,9 @@

ci/lib.sh: encapsulate Travis-specific things

- The upcoming patches will allow building git.git via VSTS CI, where
- variable names and URLs look a bit different than in Travis CI.
+ The upcoming patches will allow building git.git via Azure Pipelines
+ (i.e. Azure DevOps' Continuous Integration), where variable names and
+ URLs look a bit different than in Travis CI.

@@ -16,7 +17,7 @@
# brew install gnu-time
- brew install git-lfs gettext
+ test -z "$BREW_INSTALL_PACKAGES" ||
-+ eval brew install $BREW_INSTALL_PACKAGES
++ brew install $BREW_INSTALL_PACKAGES
brew link --force gettext
brew install caskroom/cask/perforce ;;
@@ -27,7 +28,7 @@
@@
# Library of functions shared by all CI scripts

-+if test -n "$TRAVIS_COMMIT"
++if test true = "$TRAVIS"
+then
+ # We are running within Travis CI
+ CI_BRANCH="$TRAVIS_BRANCH"

  3:  064db77669 =  3:  52337f1875 test-date: add a subcommand to measure times in shell scripts
  4:  2b5d678594 !  4:  cf4c5ae470 tests: optionally write results as JUnit-style .xml

@@ -3,7 +3,7 @@
tests: optionally write results as JUnit-style .xml

This will come in handy when publishing the results of Git's test suite
- during an automated VSTS CI run.
+ during an automated Azure DevOps run.

5: 127dfcfb09 ! 5: 486d1d2518 ci/lib.sh: add support for VSTS CI

     @@ -1,6 +1,6 @@
      Author: Johannes Schindelin <[email protected]>
      
     -    ci/lib.sh: add support for VSTS CI
     +    ci/lib.sh: add support for Azure Pipelines
      
          This patch introduces a conditional arm that defines some environment
          variables and a function that displays the URL given the job id (to
     @@ -17,7 +17,7 @@
       	export GIT_TEST_OPTS="--verbose-log -x --immediate"
      +elif test -n "$SYSTEM_TASKDEFINITIONSURI"
      +then
     -+	# We are running in VSTS CI
     ++	# We are running in Azure Pipelines
      +	CI_BRANCH="$BUILD_SOURCEBRANCH"
      +	CI_COMMIT="$BUILD_SOURCEVERSION"
      +	CI_JOB_ID="$BUILD_BUILDID"
     @@ -29,7 +29,7 @@
      +
      +	# use a subdirectory of the cache dir (because the file share is shared
      +	# among *all* phases)
     -+	cache_dir="$HOME/vsts-cache/$SYSTEM_PHASENAME"
     ++	cache_dir="$HOME/test-cache/$SYSTEM_PHASENAME"
      +
      +	url_for_job_id () {
      +		echo "$SYSTEM_TASKDEFINITIONSURI$SYSTEM_TEAMPROJECT/_build/results?buildId=$1"
  6:  ab66060c57 !  6:  1a22efe849 Add a build definition for VSTS
     @@ -1,8 +1,8 @@
      Author: Johannes Schindelin <[email protected]>
      
     -    Add a build definition for VSTS
     +    Add a build definition for Azure DevOps
      
     -    This commit adds a .vsts-ci.yml which is Visual Studio Team Services'
     +    This commit adds an azure-pipelines.yml file which is Azure DevOps'
          equivalent to Travis CI's .travis.yml.
      
          To make things a bit easier to understand, we refrain from using the
     @@ -19,10 +19,10 @@
      
          Signed-off-by: Johannes Schindelin <[email protected]>
      
     -diff --git a/.vsts-ci.yml b/.vsts-ci.yml
     +diff --git a/azure-pipelines.yml b/azure-pipelines.yml
      new file mode 100644
      --- /dev/null
     -+++ b/.vsts-ci.yml
     ++++ b/azure-pipelines.yml
      @@
      +resources:
      +- repo: self
     @@ -36,12 +36,12 @@
      +    name: Hosted Ubuntu 1604
      +  steps:
      +  - bash: |
     -+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
      +
     -+       sudo apt-get update
     -+       sudo apt-get -y install git gcc make libssl-dev libcurl4-openssl-dev libexpat-dev tcl tk gettext git-email zlib1g-dev apache2-bin
     ++       sudo apt-get update &&
     ++       sudo apt-get -y install git gcc make libssl-dev libcurl4-openssl-dev libexpat-dev tcl tk gettext git-email zlib1g-dev apache2-bin &&
      +
     -+       export CC=clang
     ++       export CC=clang || exit 1
      +
      +       ci/install-dependencies.sh
      +       ci/run-build-and-tests.sh || {
     @@ -49,8 +49,10 @@
      +           exit 1
      +       }
      +
     -+       sudo umount "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || sudo umount "$HOME/test-cache" || exit 1
      +    displayName: 'ci/run-build-and-tests.sh'
     ++    env:
     ++      GITFILESHAREPWD: $(gitfileshare.pwd)
      +  - task: PublishTestResults@2
      +    displayName: 'Publish Test Results **/TEST-*.xml'
      +    inputs:
     @@ -67,10 +69,10 @@
      +    name: Hosted Ubuntu 1604
      +  steps:
      +  - bash: |
     -+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
      +
     -+       sudo apt-get update
     -+       sudo apt-get -y install git gcc make libssl-dev libcurl4-openssl-dev libexpat-dev tcl tk gettext git-email zlib1g-dev apache2-bin
     ++       sudo apt-get update &&
     ++       sudo apt-get -y install git gcc make libssl-dev libcurl4-openssl-dev libexpat-dev tcl tk gettext git-email zlib1g-dev apache2-bin || exit 1
      +
      +       ci/install-dependencies.sh
      +       ci/run-build-and-tests.sh || {
     @@ -78,8 +80,10 @@
      +           exit 1
      +       }
      +
     -+       sudo umount "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || sudo umount "$HOME/test-cache" || exit 1
      +    displayName: 'ci/run-build-and-tests.sh'
     ++    env:
     ++      GITFILESHAREPWD: $(gitfileshare.pwd)
      +  - task: PublishTestResults@2
      +    displayName: 'Publish Test Results **/TEST-*.xml'
      +    inputs:
     @@ -96,7 +100,7 @@
      +    name: Hosted macOS
      +  steps:
      +  - bash: |
     -+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
      +
      +       export CC=clang
      +
     @@ -106,8 +110,10 @@
      +           exit 1
      +       }
      +
     -+       umount "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || umount "$HOME/test-cache" || exit 1
      +    displayName: 'ci/run-build-and-tests.sh'
     ++    env:
     ++      GITFILESHAREPWD: $(gitfileshare.pwd)
      +  - task: PublishTestResults@2
      +    displayName: 'Publish Test Results **/TEST-*.xml'
      +    inputs:
     @@ -124,7 +130,7 @@
      +    name: Hosted macOS
      +  steps:
      +  - bash: |
     -+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
      +
      +       ci/install-dependencies.sh
      +       ci/run-build-and-tests.sh || {
     @@ -132,8 +138,10 @@
      +           exit 1
      +       }
      +
     -+       umount "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || umount "$HOME/test-cache" || exit 1
      +    displayName: 'ci/run-build-and-tests.sh'
     ++    env:
     ++      GITFILESHAREPWD: $(gitfileshare.pwd)
      +  - task: PublishTestResults@2
      +    displayName: 'Publish Test Results **/TEST-*.xml'
      +    inputs:
     @@ -150,20 +158,22 @@
      +    name: Hosted Ubuntu 1604
      +  steps:
      +  - bash: |
     -+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
      +
     -+       sudo apt-get update
     -+       sudo apt-get -y install git gcc make libssl-dev libcurl4-openssl-dev libexpat-dev tcl tk gettext git-email zlib1g-dev
     ++       sudo apt-get update &&
     ++       sudo apt-get -y install git gcc make libssl-dev libcurl4-openssl-dev libexpat-dev tcl tk gettext git-email zlib1g-dev &&
      +
     -+       export jobname=GETTEXT_POISON
     ++       export jobname=GETTEXT_POISON || exit 1
      +
      +       ci/run-build-and-tests.sh || {
      +           ci/print-test-failures.sh
      +           exit 1
      +       }
      +
     -+       sudo umount "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || sudo umount "$HOME/test-cache" || exit 1
      +    displayName: 'ci/run-build-and-tests.sh'
     ++    env:
     ++      GITFILESHAREPWD: $(gitfileshare.pwd)
      +  - task: PublishTestResults@2
      +    displayName: 'Publish Test Results **/TEST-*.xml'
      +    inputs:
     @@ -184,8 +194,10 @@
      +       # Helper to check the error level of the latest command (exit with error when appropriate)
      +       function c() { if (!$?) { exit(1) } }
      +
     -+       net use s: \\gitfileshare.file.core.windows.net\vsts-cache "$(gitfileshare.pwd)" /user:AZURE\gitfileshare /persistent:no; c
     -+       cmd /c mklink /d "$(Build.SourcesDirectory)\vsts-cache" S:\; c
     ++       if ("$GITFILESHAREPWD" -ne "" -and "$GITFILESHAREPWD" -ne "`$`(gitfileshare.pwd)") {
     ++         net use s: \\gitfileshare.file.core.windows.net\test-cache "$GITFILESHAREPWD" /user:AZURE\gitfileshare /persistent:no; c
     ++         cmd /c mklink /d "$(Build.SourcesDirectory)\test-cache" S:\; c
     ++       }
      +
      +       # Add build agent's MinGit to PATH
      +       $env:PATH = $env:AGENT_HOMEDIRECTORY +"\externals\\git\cmd;" +$env:PATH
     @@ -215,7 +227,7 @@
      +
      +       # Initialize Git for Windows' SDK
      +       $sdk_path = "$(Build.SourcesDirectory)\git-sdk-64"
     -+       init "$sdk_path" "https://git-for-windows.visualstudio.com/_git/git-sdk-64" 0
     ++       init "$sdk_path" "https://dev.azure.com/git-for-windows/git-sdk-64/_git/git-sdk-64" 0
      +       init usr\src\build-extra https://github.com/git-for-windows/build-extra 1
      +
      +       cd "$(Build.SourcesDirectory)"; c
     @@ -227,14 +239,19 @@
      +
      +         make -j10 DEVELOPER=1 NO_PERL=1 || exit 1
      +         NO_PERL=1 NO_SVN_TESTS=1 GIT_TEST_OPTS=\"--quiet --write-junit-xml\" time make -j15 -k DEVELOPER=1 test || {
     -+           NO_PERL=1 NO_SVN_TESTS=1 GIT_TEST_OPTS=\"-i -v -x\" make -k failed; exit 1
     ++           NO_PERL=1 NO_SVN_TESTS=1 GIT_TEST_OPTS=\"-i -v -x\" make -k -C t failed; exit 1
      +         }
      +
      +         save_good_tree
      +       "@
     ++       c
      +
     -+       cmd /c rmdir "$(Build.SourcesDirectory)\vsts-cache"
     ++       if ("$GITFILESHAREPWD" -ne "" -and "$GITFILESHAREPWD" -ne "`$`(gitfileshare.pwd)") {
     ++         cmd /c rmdir "$(Build.SourcesDirectory)\test-cache"
     ++       }
      +    displayName: 'build & test'
     ++    env:
     ++      GITFILESHAREPWD: $(gitfileshare.pwd)
      +  - task: PublishTestResults@2
      +    displayName: 'Publish Test Results **/TEST-*.xml'
      +    inputs:
     @@ -251,28 +268,30 @@
      +    name: Hosted Ubuntu 1604
      +  steps:
      +  - bash: |
     -+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
      +
     -+       sudo apt-get update
     ++       sudo apt-get update &&
      +       sudo apt-get -y install \
      +           apt-transport-https \
      +           ca-certificates \
      +           curl \
     -+           software-properties-common
     -+       curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add -
     ++           software-properties-common &&
     ++       curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add - &&
      +       sudo add-apt-repository \
      +          "deb [arch=amd64] https://download.docker.com/linux/ubuntu \
      +          $(lsb_release -cs) \
     -+          stable"
     -+       sudo apt-get update
     -+       sudo apt-get -y install docker-ce
     ++          stable" &&
     ++       sudo apt-get update &&
     ++       sudo apt-get -y install docker-ce &&
      +
      +       sudo AGENT_OS="$AGENT_OS" BUILD_BUILDNUMBER="$BUILD_BUILDNUMBER" BUILD_REPOSITORY_URI="$BUILD_REPOSITORY_URI" BUILD_SOURCEBRANCH="$BUILD_SOURCEBRANCH" BUILD_SOURCEVERSION="$BUILD_SOURCEVERSION" SYSTEM_PHASENAME="$SYSTEM_PHASENAME" SYSTEM_TASKDEFINITIONSURI="$SYSTEM_TASKDEFINITIONSURI" SYSTEM_TEAMPROJECT="$SYSTEM_TEAMPROJECT" CC=$CC MAKEFLAGS=-j3 bash -lxc ci/run-linux32-docker.sh || exit 1
      +
      +       sudo chmod a+r t/out/TEST-*.xml
      +
     -+       sudo umount "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || sudo umount "$HOME/test-cache" || exit 1
      +    displayName: 'ci/run-linux32-docker.sh'
     ++    env:
     ++      GITFILESHAREPWD: $(gitfileshare.pwd)
      +  - task: PublishTestResults@2
      +    displayName: 'Publish Test Results **/TEST-*.xml'
      +    inputs:
     @@ -289,17 +308,19 @@
      +    name: Hosted Ubuntu 1604
      +  steps:
      +  - bash: |
     -+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
      +
     -+       sudo apt-get update
     -+       sudo apt-get install -y coccinelle
     ++       sudo apt-get update &&
     ++       sudo apt-get install -y coccinelle &&
      +
     -+       export jobname=StaticAnalysis
     ++       export jobname=StaticAnalysis &&
      +
      +       ci/run-static-analysis.sh || exit 1
      +
     -+       sudo umount "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || sudo umount "$HOME/test-cache" || exit 1
      +    displayName: 'ci/run-static-analysis.sh'
     ++    env:
     ++      GITFILESHAREPWD: $(gitfileshare.pwd)
      +
      +- phase: documentation
      +  displayName: Documentation
     @@ -308,18 +329,20 @@
      +    name: Hosted Ubuntu 1604
      +  steps:
      +  - bash: |
     -+       ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/vsts-cache gitfileshare "$(gitfileshare.pwd)" "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
      +
     -+       sudo apt-get update
     -+       sudo apt-get install -y asciidoc xmlto asciidoctor
     ++       sudo apt-get update &&
     ++       sudo apt-get install -y asciidoc xmlto asciidoctor &&
      +
     -+       export ALREADY_HAVE_ASCIIDOCTOR=yes.
     -+       export jobname=Documentation
     ++       export ALREADY_HAVE_ASCIIDOCTOR=yes. &&
     ++       export jobname=Documentation &&
      +
      +       ci/test-documentation.sh || exit 1
      +
     -+       sudo umount "$HOME/vsts-cache" || exit 1
     ++       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || sudo umount "$HOME/test-cache" || exit 1
      +    displayName: 'ci/test-documentation.sh'
     ++    env:
     ++      GITFILESHAREPWD: $(gitfileshare.pwd)
      
      diff --git a/ci/mount-fileshare.sh b/ci/mount-fileshare.sh
      new file mode 100755
  -:  ---------- >  7:  12d6137f8d tests: introduce `test_atexit`
  -:  ---------- >  8:  3bb226b79b git-daemon: use `test_atexit` in the tests
  -:  ---------- >  9:  3e2193a73d git-p4: use `test_atexit` to kill the daemon
  7:  942bf423a4 ! 10:  ae3c42519a tests: include detailed trace logs with --write-junit-xml upon failure

@@ -14,9 +14,9 @@
almost all of those cases: only when a test fails do we have a way to include the trace.

- So let's do something different in VSTS: let's run all the tests with
- --quiet first, and only if a failure is encountered, try to trace the
- commands as they are executed.
+ So let's do something different when using Azure DevOps: let's run all
+ the tests with --quiet first, and only if a failure is encountered,
+ try to trace the commands as they are executed.

Of course, we cannot turn on --verbose-log after the fact. So let's just re-run the test with all the same options, adding --verbose-log.
@@ -25,15 +25,10 @@
Note: there is an off chance that re-running the test in verbose mode "fixes" the failures (and this does happen from time to time!). That is a possibility we should be able to live with. Ideally, we would label
- this as "Passed upon rerun", and that outcome even exists on VSTS, but
- it is not available when using the JUnit XML format for now:
- https://github.com/Microsoft/vsts-agent/blob/master/src/Agent.Worker/TestResults/JunitResultReader.cs
-
- This patch contains a slightly inelegant workaround for the p4 and
- git-daemon tests: when we try to re-run the p4/git-daemon tests after
- the daemon has been started already, we need to kill said daemon. We do
- this by detecting the presence of the kill_p4d and stop_git_daemon
- shell functions and calling them if available.
+ this as "Passed upon rerun", and Azure Pipelines even know about that
+ outcome, but it is not available when using the JUnit XML format for
+ now:
+ https://github.com/Microsoft/azure-pipelines-agent/blob/master/src/Agent.Worker/TestResults/JunitResultReader.cs

@@ -60,14 +55,8 @@
then
+ if test -z "$GIT_TEST_TEE_OUTPUT_FILE"
+ then
-+ case "$(type kill_p4d 2>/dev/null | head -n 1)" in
-+ *function*) kill_p4d;;
-+ esac
-+
-+ case "$(type stop_git_daemon 2>/dev/null |
-+ head -n 1)" in
-+ *function*) stop_git_daemon;;
-+ esac
++ # clean up
++ test_atexit_handler
+
+ # re-run with --verbose-log
+ echo "# Re-running: $junit_rerun_options_sq" >&2

  8:  3e83c64090 = 11:  2466a48aa3 tests: record more stderr with --write-junit-xml in case of failure
  9:  dc1d890d71 ! 12:  d112b3fe86 README: add a build badge (status of the VSTS build)
     @@ -1,6 +1,6 @@
      Author: Johannes Schindelin <[email protected]>
      
     -    README: add a build badge (status of the VSTS build)
     +    README: add a build badge (status of the Azure Pipelines build)
      
          Just like so many other OSS projects, we now also have a build badge.
      
     @@ -10,7 +10,7 @@
      --- a/README.md
      +++ b/README.md
      @@
     -+[![Build Status](https://git.visualstudio.com/git/_apis/build/status/test-git.git)](https://git.visualstudio.com/git/_build/latest?definitionId=2)
     ++[![Build Status](https:/dev.azure.com/git/git/_apis/build/status/test-git.git)](https://dev.azure.com/git/git/_build/latest?definitionId=2)
      +
       Git - fast, scalable, distributed revision control system
       =========================================================

-: ---------- > 13: 0a53f37135 travis: fix skipping tagged releases

gitgitgadget

From: Johannes Schindelin <[email protected]>

The name is hard-coded to reflect that we use Travis CI for continuous testing.

In the next commits, we will extend this to be able use Azure DevOps, too.

So let's adjust the name to make it more generic.

---
 ci/install-dependencies.sh     | 2 +-
 ci/{lib-travisci.sh => lib.sh} | 0
 ci/print-test-failures.sh      | 2 +-
 ci/run-build-and-tests.sh      | 2 +-
 ci/run-linux32-docker.sh       | 2 +-
 ci/run-static-analysis.sh      | 2 +-
 ci/run-windows-build.sh        | 2 +-
 ci/test-documentation.sh       | 2 +-
 8 files changed, 7 insertions(+), 7 deletions(-)
 rename ci/{lib-travisci.sh => lib.sh} (100%)

index 75a9fd2475..961064658e 100755
@@ -3,7 +3,7 @@
 # Install dependencies required to build and test Git on Linux and macOS
 #
 
-. ${0%/*}/lib-travisci.sh
+. ${0%/*}/lib.sh
 
 P4WHENCE=http://filehost.perforce.com/perforce/r$LINUX_P4_VERSION
 LFSWHENCE=https://github.com/github/git-lfs/releases/download/v$LINUX_GIT_LFS_VERSION

similarity index 100%
rename from ci/lib-travisci.sh
rename to ci/lib.sh

index d55460a212..7aef39a2fd 100755
@@ -3,7 +3,7 @@
 # Print output of failing tests
 #
 
-. ${0%/*}/lib-travisci.sh
+. ${0%/*}/lib.sh
 
 # Tracing executed commands would produce too much noise in the loop below.
 set +x

index 2a5bff4a1c..e28ac2fb9a 100755
@@ -3,7 +3,7 @@
 # Build and test Git
 #
 
-. ${0%/*}/lib-travisci.sh
+. ${0%/*}/lib.sh
 
 ln -s "$cache_dir/.prove" t/.prove
 

index 21637903ce..751acfcf8a 100755
@@ -3,7 +3,7 @@
 # Download and run Docker image to build and test 32-bit Git
 #
 
-. ${0%/*}/lib-travisci.sh
+. ${0%/*}/lib.sh
 
 docker pull daald/ubuntu32:xenial
 

index 5688f261d0..dc189c7456 100755
@@ -3,7 +3,7 @@
 # Perform various static code analysis checks
 #
 
-. ${0%/*}/lib-travisci.sh
+. ${0%/*}/lib.sh
 
 make --jobs=2 coccicheck
 

index d99a180e52..a73a4eca0a 100755
@@ -6,7 +6,7 @@
 # supported) and a commit hash.
 #
 
-. ${0%/*}/lib-travisci.sh
+. ${0%/*}/lib.sh
 
 test $# -ne 2 && echo "Unexpected number of parameters" && exit 1
 test -z "$GFW_CI_TOKEN" && echo "GFW_CI_TOKEN not defined" && exit

index a20de9ca12..d3cdbac73f 100755
@@ -3,7 +3,7 @@
 # Perform sanity checks on documentation and build it.
 #
 
-. ${0%/*}/lib-travisci.sh
+. ${0%/*}/lib.sh
 
 gem install asciidoctor
 

gitgitgadget

"Johannes Schindelin via GitGitGadget" <[email protected]>

With many pieces still shared across these two CI systems, this looks like an excellent move.

I was wondering if the other extreme of moving ci/* to ci/travis/*, introducing ci/azure/* and sharing nothing was needed, but being able to share meaning amount of set-up sounds quite good.

From: Johannes Schindelin <[email protected]>

The upcoming patches will allow building git.git via Azure Pipelines (i.e. Azure DevOps' Continuous Integration), where variable names and URLs look a bit different than in Travis CI.

---
 ci/install-dependencies.sh |  3 ++-
 ci/lib.sh                  | 44 +++++++++++++++++++++++++++-----------
 ci/print-test-failures.sh  |  2 +-
 ci/test-documentation.sh   |  1 +
 4 files changed, 35 insertions(+), 15 deletions(-)

index 961064658e..63fa37f68f 100755
@@ -28,7 +28,8 @@ osx-clang|osx-gcc)
 	brew update --quiet
 	# Uncomment this if you want to run perf tests:
 	# brew install gnu-time
-	brew install git-lfs gettext
+	test -z "$BREW_INSTALL_PACKAGES" ||
+	brew install $BREW_INSTALL_PACKAGES
 	brew link --force gettext
 	brew install caskroom/cask/perforce
 	;;

index 7aef39a2fd..d2045b63a6 100755
@@ -69,7 +69,7 @@ do
 	fi
 done
 
-if [ $combined_trash_size -gt 0 ]
+if [ -n "$TRAVIS_JOB_ID" -a $combined_trash_size -gt 0 ]
 then
 	echo "------------------------------------------------------------------------"
 	echo "Trash directories embedded in this log can be extracted by running:"

index d3cdbac73f..7d0beb2832 100755
@@ -5,6 +5,7 @@
 
 . ${0%/*}/lib.sh
 
+test -n "$ALREADY_HAVE_ASCIIDOCTOR" ||
 gem install asciidoctor
 
 make check-builtins

gitgitgadget

Do you mean "VAR=VAL; export VAR" is kosher, "export VAR=VAL" is not?

Ah, no, you didn't. Although I think both are valid points, I think ci/lib.sh is expected to be used only inside a more predictable environment (e.g. we know the shell used is not a random POSIX shell but one that is happy with "export VAR=VAL"), so it should be OK. Showing the values of these variables in the log may still be good idea.

Sure (provided that it is an issue to begin with---if we are interested in the value of TRAVIS_BRANCH, for example, being able to see it only because "CI_BRANCH=$TRAVIS_BRANCH" assignment is made feels a bit sideways---we'd be better off explicitly logging anything we are interested in in the longer term, no?).

Yes. Travis CI runs an Ubuntu LTS, where /bin/sh is dash, which understands 'export VAR=val' just fine. I don't know what Linux distro runs on Azure Pipelines, but the build definition in patch 6 explicitly asks for Bash, so that should be fine as well.

Well, all the $TRAVIS_* and $CI_* variables are not that interessant, because they are only used in the build scripts, and then we can see their substituted values in the build logs. The same applies to the variables $cache_dir and $BREW_INSTALL_PACKAGES as well.

$GIT_PROVE_OPTS and $GIT_TEST_OPTS, however, are only used in 't/Makefile' but not in the build scripts, thus their values don't show up in the build logs.

I run some Travis CI builds with custom $GIT_TEST_OPTS, which, of course, conflicted with this patch, and I messed up the conflict resolution. I think I would have noticed sooner what went wrong, if the value of $GIT_TEST_OPTS were visible in the build logs.

I've found the output with 'set -x' sufficient so far, and don't think that an explicit logging facility is worth it.

That's an interesting perspective. I would have thoguht that the values of these two variables we can see above, that are constants without any substitution or customization, are the least interesting things to see in the log.

From: Johannes Schindelin <[email protected]>

In the next commit, we want to teach Git's test suite to optionally output test results in JUnit-style .xml files. These files contain information about the time spent. So we need a way to measure time.

While we could use date +%s for that, this will give us only seconds, i.e. very coarse-grained timings.

GNU date supports date +%s.%N (i.e. nanosecond-precision output), but there is no equivalent in BSD date (read: on macOS, we would not be able to obtain precise timings).

So let's introduce test-tool date getnanos, with an optional start time, that outputs preciser values.

Granted, it is a bit pointless to try measuring times accurately in shell scripts, certainly to nanosecond precision. But it is better than second-granularity.

---
 t/helper/test-date.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

index a0837371ab..792a805374 100644
@@ -7,6 +7,7 @@ static const char *usage_msg = "\n"
 "  test-tool date parse [date]...\n"
 "  test-tool date approxidate [date]...\n"
 "  test-tool date timestamp [date]...\n"
+"  test-tool date getnanos [start-nanos]\n"
 "  test-tool date is64bit\n"
 "  test-tool date time_t-is64bit\n";
 

@@ -82,6 +83,15 @@ static void parse_approx_timestamp(const char **argv, struct timeval *now)
 	}
 }
 
+static void getnanos(const char **argv, struct timeval *now)
+{
+	double seconds = getnanotime() / 1.0e9;
+
+	if (*argv)
+		seconds -= strtod(*argv, NULL);
+	printf("%lf\n", seconds);
+}
+
 int cmd__date(int argc, const char **argv)
 {
 	struct timeval now;

@@ -108,6 +118,8 @@ int cmd__date(int argc, const char **argv)
 		parse_approxidate(argv+1, &now);
 	else if (!strcmp(*argv, "timestamp"))
 		parse_approx_timestamp(argv+1, &now);
+	else if (!strcmp(*argv, "getnanos"))
+		getnanos(argv+1, &now);
 	else if (!strcmp(*argv, "is64bit"))
 		return sizeof(timestamp_t) == 8 ? 0 : 1;
 	else if (!strcmp(*argv, "time_t-is64bit"))

gitgitgadget

From: Johannes Schindelin <[email protected]>

This will come in handy when publishing the results of Git's test suite during an automated Azure DevOps run.

---
 t/.gitignore  |  1 +
 t/test-lib.sh | 98 +++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 99 insertions(+)

index 348715f0e4..91cf5772fe 100644
@@ -2,3 +2,4 @@
 /test-results
 /.prove
 /chainlinttmp
+/out/

gitgitgadget

From: Johannes Schindelin <[email protected]>

This patch introduces a conditional arm that defines some environment variables and a function that displays the URL given the job id (to identify previous runs for known-good trees).

---
 ci/lib.sh | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)

index 8532555b4e..584abcd529 100755
@@ -19,6 +19,29 @@ then
 	BREW_INSTALL_PACKAGES="git-lfs gettext"
 	export GIT_PROVE_OPTS="--timer --jobs 3 --state=failed,slow,save"
 	export GIT_TEST_OPTS="--verbose-log -x --immediate"
+elif test -n "$SYSTEM_TASKDEFINITIONSURI"
+then
+	# We are running in Azure Pipelines
+	CI_BRANCH="$BUILD_SOURCEBRANCH"
+	CI_COMMIT="$BUILD_SOURCEVERSION"
+	CI_JOB_ID="$BUILD_BUILDID"
+	CI_JOB_NUMBER="$BUILD_BUILDNUMBER"
+	CI_OS_NAME="$(echo "$AGENT_OS" | tr A-Z a-z)"
+	test darwin != "$CI_OS_NAME" || CI_OS_NAME=osx
+	CI_REPO_SLUG="$(expr "$BUILD_REPOSITORY_URI" : '.*/\([^/]*/[^/]*\)$')"
+	CC="${CC:-gcc}"
+
+	# use a subdirectory of the cache dir (because the file share is shared
+	# among *all* phases)
+	cache_dir="$HOME/test-cache/$SYSTEM_PHASENAME"
+
+	url_for_job_id () {
+		echo "$SYSTEM_TASKDEFINITIONSURI$SYSTEM_TEAMPROJECT/_build/results?buildId=$1"
+	}
+
+	BREW_INSTALL_PACKAGES=
+	export GIT_PROVE_OPTS="--timer --jobs 10 --state=failed,slow,save"
+	export GIT_TEST_OPTS="--quiet --write-junit-xml"
 fi
 
 skip_branch_tip_with_tag () {

gitgitgadget

Hi Gábor,

Both packages are already available on Azure Pipelines' VMs. I will mention this in the commit message.

Thanks,
Dscho

From: Johannes Schindelin <[email protected]>

This commit adds an azure-pipelines.yml file which is Azure DevOps' equivalent to Travis CI's .travis.yml.

To make things a bit easier to understand, we refrain from using the matrix feature here because (while it is powerful) it can be a bit confusing to users who are not familiar with CI setups. Therefore, we use a separate phase even for similar configurations (such as GCC vs Clang on Linux, GCC vs Clang on macOS).

Also, we make use of the shiny new feature we just introduced where the test suite can output JUnit-style .xml files. This information is made available in a nice UI that allows the viewer to filter by phase and/or test number, and to see trends such as: number of (failing) tests, time spent running the test suite, etc.

---
 azure-pipelines.yml   | 319 ++++++++++++++++++++++++++++++++++++++++++
 ci/mount-fileshare.sh |  26 ++++
 2 files changed, 345 insertions(+)
 create mode 100644 azure-pipelines.yml
 create mode 100755 ci/mount-fileshare.sh

new file mode 100644
index 0000000000..b5749121d2
@@ -0,0 +1,319 @@
+resources:
+- repo: self
+  fetchDepth: 1
+
+phases:
+- phase: linux_clang
+  displayName: linux-clang
+  condition: succeeded()
+  queue:
+    name: Hosted Ubuntu 1604
+  steps:
+  - bash: |
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
+
+       sudo apt-get update &&
+       sudo apt-get -y install git gcc make libssl-dev libcurl4-openssl-dev libexpat-dev tcl tk gettext git-email zlib1g-dev apache2-bin &&
+
+       export CC=clang || exit 1
+
+       ci/install-dependencies.sh
+       ci/run-build-and-tests.sh || {
+           ci/print-test-failures.sh
+           exit 1
+       }
+
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || sudo umount "$HOME/test-cache" || exit 1
+    displayName: 'ci/run-build-and-tests.sh'
+    env:
+      GITFILESHAREPWD: $(gitfileshare.pwd)
+  - task: PublishTestResults@2
+    displayName: 'Publish Test Results **/TEST-*.xml'
+    inputs:
+      mergeTestResults: true
+      testRunTitle: 'linux-clang'
+      platform: Linux
+      publishRunAttachments: false
+    condition: succeededOrFailed()
+
+- phase: linux_gcc
+  displayName: linux-gcc
+  condition: succeeded()
+  queue:
+    name: Hosted Ubuntu 1604
+  steps:
+  - bash: |
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
+
+       sudo apt-get update &&
+       sudo apt-get -y install git gcc make libssl-dev libcurl4-openssl-dev libexpat-dev tcl tk gettext git-email zlib1g-dev apache2-bin || exit 1
+
+       ci/install-dependencies.sh
+       ci/run-build-and-tests.sh || {
+           ci/print-test-failures.sh
+           exit 1
+       }
+
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || sudo umount "$HOME/test-cache" || exit 1
+    displayName: 'ci/run-build-and-tests.sh'
+    env:
+      GITFILESHAREPWD: $(gitfileshare.pwd)
+  - task: PublishTestResults@2
+    displayName: 'Publish Test Results **/TEST-*.xml'
+    inputs:
+      mergeTestResults: true
+      testRunTitle: 'linux-gcc'
+      platform: Linux
+      publishRunAttachments: false
+    condition: succeededOrFailed()
+
+- phase: osx_clang
+  displayName: osx-clang
+  condition: succeeded()
+  queue:
+    name: Hosted macOS
+  steps:
+  - bash: |
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
+
+       export CC=clang
+
+       ci/install-dependencies.sh
+       ci/run-build-and-tests.sh || {
+           ci/print-test-failures.sh
+           exit 1
+       }
+
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || umount "$HOME/test-cache" || exit 1
+    displayName: 'ci/run-build-and-tests.sh'
+    env:
+      GITFILESHAREPWD: $(gitfileshare.pwd)
+  - task: PublishTestResults@2
+    displayName: 'Publish Test Results **/TEST-*.xml'
+    inputs:
+      mergeTestResults: true
+      testRunTitle: 'osx-clang'
+      platform: macOS
+      publishRunAttachments: false
+    condition: succeededOrFailed()
+
+- phase: osx_gcc
+  displayName: osx-gcc
+  condition: succeeded()
+  queue:
+    name: Hosted macOS
+  steps:
+  - bash: |
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
+
+       ci/install-dependencies.sh
+       ci/run-build-and-tests.sh || {
+           ci/print-test-failures.sh
+           exit 1
+       }
+
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || umount "$HOME/test-cache" || exit 1
+    displayName: 'ci/run-build-and-tests.sh'
+    env:
+      GITFILESHAREPWD: $(gitfileshare.pwd)
+  - task: PublishTestResults@2
+    displayName: 'Publish Test Results **/TEST-*.xml'
+    inputs:
+      mergeTestResults: true
+      testRunTitle: 'osx-gcc'
+      platform: macOS
+      publishRunAttachments: false
+    condition: succeededOrFailed()
+
+- phase: gettext_poison
+  displayName: GETTEXT_POISON
+  condition: succeeded()
+  queue:
+    name: Hosted Ubuntu 1604
+  steps:
+  - bash: |
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
+
+       sudo apt-get update &&
+       sudo apt-get -y install git gcc make libssl-dev libcurl4-openssl-dev libexpat-dev tcl tk gettext git-email zlib1g-dev &&
+
+       export jobname=GETTEXT_POISON || exit 1
+
+       ci/run-build-and-tests.sh || {
+           ci/print-test-failures.sh
+           exit 1
+       }
+
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || sudo umount "$HOME/test-cache" || exit 1
+    displayName: 'ci/run-build-and-tests.sh'
+    env:
+      GITFILESHAREPWD: $(gitfileshare.pwd)
+  - task: PublishTestResults@2
+    displayName: 'Publish Test Results **/TEST-*.xml'
+    inputs:
+      mergeTestResults: true
+      testRunTitle: 'gettext-poison'
+      platform: Linux
+      publishRunAttachments: false
+    condition: succeededOrFailed()
+
+- phase: windows
+  displayName: Windows
+  condition: succeeded()
+  queue:
+    name: Hosted VS2017
+    timeoutInMinutes: 240
+  steps:
+  - powershell: |
+       # Helper to check the error level of the latest command (exit with error when appropriate)
+       function c() { if (!$?) { exit(1) } }
+
+       if ("$GITFILESHAREPWD" -ne "" -and "$GITFILESHAREPWD" -ne "`$`(gitfileshare.pwd)") {
+         net use s: \\gitfileshare.file.core.windows.net\test-cache "$GITFILESHAREPWD" /user:AZURE\gitfileshare /persistent:no; c
+         cmd /c mklink /d "$(Build.SourcesDirectory)\test-cache" S:\; c
+       }
+
+       # Add build agent's MinGit to PATH
+       $env:PATH = $env:AGENT_HOMEDIRECTORY +"\externals\\git\cmd;" +$env:PATH
+
+       # Helper to initialize (or update) a Git worktree
+       function init ($path, $url, $set_origin) {
+           if (Test-Path $path) {
+               cd $path; c
+               if (Test-Path .git) {
+                   git init; c
+               } else {
+                   git status
+               }
+           } else {
+               git init $path; c
+               cd $path; c
+           }
+           git config core.autocrlf false; c
+           git config core.untrackedCache true; c
+           if (($set_origin -ne 0) -and !(git config remote.origin.url)) {
+               git remote add origin $url; c
+           }
+           git fetch --depth=1 $url master; c
+           git reset --hard FETCH_HEAD; c
+           git clean -df; c
+       }
+
+       # Initialize Git for Windows' SDK
+       $sdk_path = "$(Build.SourcesDirectory)\git-sdk-64"
+       init "$sdk_path" "https://dev.azure.com/git-for-windows/git-sdk-64/_git/git-sdk-64" 0
+       init usr\src\build-extra https://github.com/git-for-windows/build-extra 1
+
+       cd "$(Build.SourcesDirectory)"; c
+
+       $env:HOME = "$(Build.SourcesDirectory)"
+       $env:MSYSTEM = "MINGW64"
+       git-sdk-64\git-cmd --command=usr\\bin\\bash.exe -lc @"
+         . ci/lib.sh
+
+         make -j10 DEVELOPER=1 NO_PERL=1 || exit 1
+         NO_PERL=1 NO_SVN_TESTS=1 GIT_TEST_OPTS=\"--quiet --write-junit-xml\" time make -j15 -k DEVELOPER=1 test || {
+           NO_PERL=1 NO_SVN_TESTS=1 GIT_TEST_OPTS=\"-i -v -x\" make -k -C t failed; exit 1
+         }
+
+         save_good_tree
+       "@
+       c
+
+       if ("$GITFILESHAREPWD" -ne "" -and "$GITFILESHAREPWD" -ne "`$`(gitfileshare.pwd)") {
+         cmd /c rmdir "$(Build.SourcesDirectory)\test-cache"
+       }
+    displayName: 'build & test'
+    env:
+      GITFILESHAREPWD: $(gitfileshare.pwd)
+  - task: PublishTestResults@2
+    displayName: 'Publish Test Results **/TEST-*.xml'
+    inputs:
+      mergeTestResults: true
+      testRunTitle: 'windows'
+      platform: Windows
+      publishRunAttachments: false
+    condition: succeededOrFailed()
+
+- phase: linux32
+  displayName: Linux32
+  condition: succeeded()
+  queue:
+    name: Hosted Ubuntu 1604
+  steps:
+  - bash: |
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
+
+       sudo apt-get update &&
+       sudo apt-get -y install \
+           apt-transport-https \
+           ca-certificates \
+           curl \
+           software-properties-common &&
+       curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add - &&
+       sudo add-apt-repository \
+          "deb [arch=amd64] https://download.docker.com/linux/ubuntu \
+          $(lsb_release -cs) \
+          stable" &&
+       sudo apt-get update &&
+       sudo apt-get -y install docker-ce &&
+
+       sudo AGENT_OS="$AGENT_OS" BUILD_BUILDNUMBER="$BUILD_BUILDNUMBER" BUILD_REPOSITORY_URI="$BUILD_REPOSITORY_URI" BUILD_SOURCEBRANCH="$BUILD_SOURCEBRANCH" BUILD_SOURCEVERSION="$BUILD_SOURCEVERSION" SYSTEM_PHASENAME="$SYSTEM_PHASENAME" SYSTEM_TASKDEFINITIONSURI="$SYSTEM_TASKDEFINITIONSURI" SYSTEM_TEAMPROJECT="$SYSTEM_TEAMPROJECT" CC=$CC MAKEFLAGS=-j3 bash -lxc ci/run-linux32-docker.sh || exit 1
+
+       sudo chmod a+r t/out/TEST-*.xml
+
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || sudo umount "$HOME/test-cache" || exit 1
+    displayName: 'ci/run-linux32-docker.sh'
+    env:
+      GITFILESHAREPWD: $(gitfileshare.pwd)
+  - task: PublishTestResults@2
+    displayName: 'Publish Test Results **/TEST-*.xml'
+    inputs:
+      mergeTestResults: true
+      testRunTitle: 'linux32'
+      platform: Linux
+      publishRunAttachments: false
+    condition: succeededOrFailed()
+
+- phase: static_analysis
+  displayName: StaticAnalysis
+  condition: succeeded()
+  queue:
+    name: Hosted Ubuntu 1604
+  steps:
+  - bash: |
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
+
+       sudo apt-get update &&
+       sudo apt-get install -y coccinelle &&
+
+       export jobname=StaticAnalysis &&
+
+       ci/run-static-analysis.sh || exit 1
+
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || sudo umount "$HOME/test-cache" || exit 1
+    displayName: 'ci/run-static-analysis.sh'
+    env:
+      GITFILESHAREPWD: $(gitfileshare.pwd)
+
+- phase: documentation
+  displayName: Documentation
+  condition: succeeded()
+  queue:
+    name: Hosted Ubuntu 1604
+  steps:
+  - bash: |
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || ci/mount-fileshare.sh //gitfileshare.file.core.windows.net/test-cache gitfileshare "$GITFILESHAREPWD" "$HOME/test-cache" || exit 1
+
+       sudo apt-get update &&
+       sudo apt-get install -y asciidoc xmlto asciidoctor &&
+
+       export ALREADY_HAVE_ASCIIDOCTOR=yes. &&
+       export jobname=Documentation &&
+
+       ci/test-documentation.sh || exit 1
+
+       test "$GITFILESHAREPWD" = '$(gitfileshare.pwd)' || sudo umount "$HOME/test-cache" || exit 1
+    displayName: 'ci/test-documentation.sh'
+    env:
+      GITFILESHAREPWD: $(gitfileshare.pwd)

new file mode 100755
index 0000000000..5fb5f74b70
@@ -0,0 +1,26 @@
+#!/bin/sh
+
+die () {
+	echo "$*" >&2
+	exit 1
+}
+
+test $# = 4 ||
+die "Usage: $0 <share> <username> <password> <mountpoint"
+
+mkdir -p "$4" || die "Could not create $4"
+
+case "$(uname -s)" in
+Linux)
+	sudo mount -t cifs -o vers=3.0,username="$2",password="$3",dir_mode=0777,file_mode=0777,serverino "$1" "$4"
+	;;
+Darwin)
+	pass="$(echo "$3" | sed -e 's/\//%2F/g' -e 's/+/%2B/g')" &&
+	mount -t smbfs,soft "smb://$2:$pass@${1#//}" "$4"
+	;;
+*)
+	die "No support for $(uname -s)"
+	;;
+esac ||
+die "Could not mount $4"
+

gitgitgadget

"Johannes Schindelin via GitGitGadget" <[email protected]>

I wonder if there is a need to keep what is tested by this and Travis in sync in any way, but most of the logic is not defined in these "steps" but implemented in ci/*.sh scripts to be shared, so it would be OK, I guess.

Hi Junio,

Indeed, that was my intention.

These ci scripts are not only useful for Travis and Azure Pipelines, after all, but also a good documentation how to test locally. For example, to repeat Luke's Perforce testing, I could simply use the URLs listed in those ci scripts to get almost the same setup locally (close enough for testing).

So it is not only about sharing, and ease of maintenance, but it is also about documenting.

And yes, sharing means that we do not have to waste brain cycles on keeping things in sync.

Thanks!

Thank you, both issues will be fixed in the next iteration, Dscho

Hi Gábor,

True.

FWIW the nomenclature is "build" or "job" or "build job" for the entire run, from what I understand. The "phase" is the individual chunk that is run in an individual agent, i.e. you can have a single job running test on different OSes in separate phases.

I'll add those dependencies explicitly. It does seem, however, from a cursory look at the log, that gcc-8 should not even be picked up, as it is set via the environment variable CC (which, as you point out in below-referenced thread, is not respected):

[...]

2018-10-16T10:00:36.0177072Z ++ '[' linux-gcc = linux-gcc ']'
2018-10-16T10:00:36.0177380Z ++ export CC=gcc-8
2018-10-16T10:00:36.0177630Z ++ CC=gcc-8
2018-10-16T10:00:36.0177917Z ++ case "$jobname"

[...]

(see https://dev.azure.com/git/git/_build/results?buildId=192&view=logs)

Will do.

Thanks,
Dscho

From: Johannes Schindelin <[email protected]>

When running the p4 daemon or git daemon, we want to kill it at the end of the test script.

So far, we do this "manually".

However, in the next few commits we want to teach the test suite to optionally re-run scripts with different options, therefore we will have to have a consistent way to stop daemons.

Let's introduce test_atexit, which is loosely modeled after test_when_finished (but has a broader scope: rather than running the commands after the current test case, run them when the test script finishes, and also run them when the --immediate option is in effect).

---
 t/t0000-basic.sh        | 18 ++++++++++++++++++
 t/test-lib-functions.sh | 29 +++++++++++++++++++++++++++++
 t/test-lib.sh           |  4 ++++
 3 files changed, 51 insertions(+)

index 391f910c6a..8c5faa6ce1 100755
@@ -821,6 +821,24 @@ test_expect_success 'tests clean up even on failures' "
 	EOF
 "
 
+test_expect_success 'test_atexit is run' "
+	test_must_fail run_sub_test_lib_test \
+		atexit-cleanup 'Run atexit commands' -i <<-\\EOF &&
+	test_expect_success 'tests clean up even after a failure' '
+		> ../../clean-atexit &&
+		test_atexit rm ../../clean-atexit &&
+		> ../../also-clean-atexit &&
+		test_atexit rm ../../also-clean-atexit &&
+		> ../../dont-clean-atexit &&
+		(exit 1)
+	'
+	test_done
+	EOF
+	test_path_exists dont-clean-atexit &&
+	test_path_is_missing clean-atexit &&
+	test_path_is_missing also-clean-atexit
+"
+
 test_expect_success 'test_oid setup' '
 	test_oid_init
 '

index 78d8c3783b..d7dd0c1be9 100644
@@ -891,6 +891,35 @@ test_when_finished () {
 		} && (exit \"\$eval_ret\"); eval_ret=\$?; $test_cleanup"
 }
 
+# This function can be used to schedule some commands to be run
+# unconditionally at the end of the test script, e.g. to stop a daemon:
+#
+#	test_expect_success 'test git daemon' '
+#		git daemon &
+#		daemon_pid=$! &&
+#		test_atexit "kill $daemon_pid" &&
+#		hello world
+#	'
+
+test_atexit () {
+	# We cannot detect when we are in a subshell in general, but by
+	# doing so on Bash is better than nothing (the test will
+	# silently pass on other shells).
+	test "${BASH_SUBSHELL-0}" = 0 ||
+	error "bug in test script: test_atexit does nothing in a subshell"
+	test_atexit_cleanup="{ $*
+		} && (exit \"\$eval_ret\"); eval_ret=\$?; $test_atexit_cleanup"
+}
+
+test_atexit_handler () {
+	test : != "$test_atexit_cleanup" || return 0
+
+	setup_malloc_check
+	test_eval_ "$test_atexit_cleanup"
+	test_atexit_cleanup=:
+	teardown_malloc_check
+}
+
 # Most tests can use the created repository, but some may need to create more.
 # Usage: test_create_repo <directory>
 test_create_repo () {

index 7ed0013f6d..6f9c1f5300 100644
@@ -413,6 +413,7 @@ test_external_has_tap=0
 
 die () {
 	code=$?
+	test_atexit_handler || code=$?
 	if test -n "$GIT_EXIT_OK"
 	then
 		exit $code

@@ -826,9 +827,12 @@ write_junit_xml_testcase () {
 	junit_have_testcase=t
 }
 
+test_atexit_cleanup=:
 test_done () {
 	GIT_EXIT_OK=t
 
+	test -n "$immediate" || test_atexit_handler
+
 	if test -n "$write_junit_xml" && test -n "$junit_xml_path"
 	then
 		test -n "$junit_have_testcase" || {

gitgitgadget

I think killing daemons on failure even with '--immediate' will make life a bit easier. I remember several occasions when I tried to debug a failing test with '--immediate', some daemon stayed alive, and then the next time I run the test it failed during setup.

Besides 'git daemon' and P4, the httpd tests could benefit from this as well.

# with '--immediate' fails.  Be careful with your commands to minimize
# any changes to the failed state.

From: Johannes Schindelin <[email protected]>

This makes use of the just-introduced consistent way to specify that a long-running process needs to be terminated at the end of a test script run.

---
 t/interop/i5500-git-daemon.sh | 1 -
 t/lib-git-daemon.sh           | 3 +--
 t/t5570-git-daemon.sh         | 1 -
 3 files changed, 1 insertion(+), 4 deletions(-)

index 1daf69420b..4d22e42f84 100755
@@ -37,5 +37,4 @@ test_expect_success "fetch with $VERSION_B" '
 	test_cmp expect actual
 '
 
-stop_git_daemon
 test_done

index edbea2d986..a896af2284 100644
@@ -13,7 +13,6 @@
 #
 #	test_expect_success ...
 #
-#	stop_git_daemon
 #	test_done
 
 test_tristate GIT_TEST_GIT_DAEMON

@@ -43,7 +42,7 @@ start_git_daemon() {
 
 	mkdir -p "$GIT_DAEMON_DOCUMENT_ROOT_PATH"
 
-	trap 'code=$?; stop_git_daemon; (exit $code); die' EXIT
+	test_atexit 'stop_git_daemon'
 
 	say >&3 "Starting git daemon ..."
 	mkfifo git_daemon_output

index 7466aad111..08f95c80a2 100755
@@ -211,5 +211,4 @@ test_expect_success FAKENC 'hostname interpolation works after LF-stripping' '
 	test_cmp expect actual
 '
 
-stop_git_daemon
 test_done

gitgitgadget

From: Johannes Schindelin <[email protected]>

This should be more reliable than the current method, and prepares the test suite for a consistent way to clean up before re-running the tests with different options.

---
 t/lib-git-p4.sh                            | 10 +---------
 t/t0000-basic.sh                           |  2 ++
 t/t9800-git-p4-basic.sh                    |  4 ----
 t/t9801-git-p4-branch.sh                   |  4 ----
 t/t9802-git-p4-filetype.sh                 |  4 ----
 t/t9803-git-p4-shell-metachars.sh          |  4 ----
 t/t9804-git-p4-label.sh                    |  4 ----
 t/t9805-git-p4-skip-submit-edit.sh         |  4 ----
 t/t9806-git-p4-options.sh                  |  5 -----
 t/t9807-git-p4-submit.sh                   |  4 ----
 t/t9808-git-p4-chdir.sh                    |  4 ----
 t/t9809-git-p4-client-view.sh              |  4 ----
 t/t9810-git-p4-rcs.sh                      |  4 ----
 t/t9811-git-p4-label-import.sh             |  5 -----
 t/t9812-git-p4-wildcards.sh                |  4 ----
 t/t9813-git-p4-preserve-users.sh           |  4 ----
 t/t9814-git-p4-rename.sh                   |  4 ----
 t/t9815-git-p4-submit-fail.sh              |  4 ----
 t/t9816-git-p4-locked.sh                   |  4 ----
 t/t9817-git-p4-exclude.sh                  |  4 ----
 t/t9818-git-p4-block.sh                    |  4 ----
 t/t9819-git-p4-case-folding.sh             |  4 ----
 t/t9820-git-p4-editor-handling.sh          |  4 ----
 t/t9821-git-p4-path-variations.sh          |  4 ----
 t/t9822-git-p4-path-encoding.sh            |  4 ----
 t/t9823-git-p4-mock-lfs.sh                 |  4 ----
 t/t9824-git-p4-git-lfs.sh                  |  4 ----
 t/t9825-git-p4-handle-utf16-without-bom.sh |  4 ----
 t/t9826-git-p4-keep-empty-commits.sh       |  4 ----
 t/t9827-git-p4-change-filetype.sh          |  4 ----
 t/t9828-git-p4-map-user.sh                 |  4 ----
 t/t9829-git-p4-jobs.sh                     |  4 ----
 t/t9830-git-p4-symlink-dir.sh              |  4 ----
 t/t9831-git-p4-triggers.sh                 |  4 ----
 t/t9832-unshelve.sh                        |  4 ----
 t/t9833-errors.sh                          |  5 -----
 36 files changed, 3 insertions(+), 148 deletions(-)

index c27599474c..f4f5d7d296 100644
@@ -74,15 +74,6 @@ cli="$TRASH_DIRECTORY/cli"
 git="$TRASH_DIRECTORY/git"
 pidfile="$TRASH_DIRECTORY/p4d.pid"
 
-# Sometimes "prove" seems to hang on exit because p4d is still running
-cleanup () {
-	if test -f "$pidfile"
-	then
-		kill -9 $(cat "$pidfile") 2>/dev/null && exit 255
-	fi
-}
-trap cleanup EXIT
-
 # git p4 submit generates a temp file, which will
 # not get cleaned up if the submission fails.  Don't
 # clutter up /tmp on the test machine.

@@ -141,6 +132,7 @@ start_p4d () {
 		# p4d failed to start
 		return 1
 	fi
+	test_atexit kill_p4d
 
 	# build a p4 user so [email protected] has an entry
 	p4_add_user author

index 8c5faa6ce1..041bd7e3ce 100755
@@ -134,6 +134,7 @@ check_sub_test_lib_test_err () {
 	)
 }
 
+cat >/dev/null <<\DDD
 test_expect_success 'pretend we have a fully passing test suite' "
 	run_sub_test_lib_test full-pass '3 passing tests' <<-\\EOF &&
 	for i in 1 2 3

@@ -820,6 +821,7 @@ test_expect_success 'tests clean up even on failures' "
 	> 1..2
 	EOF
 "
+DDD
 
 test_expect_success 'test_atexit is run' "
 	test_must_fail run_sub_test_lib_test \

index 729cd25770..5856563068 100755
@@ -326,8 +326,4 @@ test_expect_success 'submit from worktree' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 6a86d6996b..50013132c8 100755
@@ -610,8 +610,4 @@ test_expect_success 'Update a file in git side and submit to P4 using client vie
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 9978352d78..94edebe272 100755
@@ -333,8 +333,4 @@ test_expect_success SYMLINKS 'empty symlink target' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index d5c3675100..2913277013 100755
@@ -105,8 +105,4 @@ test_expect_success 'branch with shell char' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index e30f80e617..3236457106 100755
@@ -108,8 +108,4 @@ test_expect_failure 'two labels on the same changelist' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 5fbf904dc8..90ef647db7 100755
@@ -98,8 +98,4 @@ test_expect_success 'no config, edited' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 3f5291b857..4e794a01bf 100755
@@ -300,9 +300,4 @@ test_expect_success 'use --git-dir option and GIT_DIR' '
 	test_path_is_file "$git"/cli_file2.t
 '
 
-
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 2325599ee6..488d916c10 100755
@@ -542,8 +542,4 @@ test_expect_success 'submit --update-shelve' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 11d2b5102c..58a9b3b71e 100755
@@ -83,8 +83,4 @@ test_expect_success SYMLINKS 'p4 client root symlink should stay symbolic' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 897b3c3034..3cff1fce1b 100755
@@ -836,8 +836,4 @@ test_expect_success 'quotes on both sides' '
 	git_verify "cdir 1/file11" "cdir 1/file12"
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index cc53debe19..57b533dc6f 100755
@@ -360,8 +360,4 @@ test_expect_failure 'Add keywords in git which do not match the default p4 value
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 602b0a5d5c..b70e81c3cd 100755
@@ -259,9 +259,4 @@ test_expect_success 'importing labels with missing revisions' '
 	)
 '
 
-
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 0206771fbb..254a7c2446 100755
@@ -211,8 +211,4 @@ test_expect_success 'wildcard files requiring keyword scrub' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 783c6ad165..fd018c87a8 100755
@@ -138,8 +138,4 @@ test_expect_success 'not preserving user with mixed authorship' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 60baa06e27..468767cbf4 100755
@@ -242,8 +242,4 @@ test_expect_success P4D_HAVE_CONFIGURABLE_RUN_MOVE_ALLOW \
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index eaf03a6563..9779dc0d11 100755
@@ -422,8 +422,4 @@ test_expect_success 'cleanup chmod after submit cancel' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index d048bd33fa..932841003c 100755
@@ -138,8 +138,4 @@ test_expect_failure 'move with lock taken' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index aac568eadf..96d25f0c02 100755
@@ -64,8 +64,4 @@ test_expect_success 'clone, then sync with exclude' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index ce7cb22ad3..0db7ab9918 100755
@@ -146,8 +146,4 @@ test_expect_success 'Clone repo with self-sizing block size' '
 	test_line_count \> 10 log
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index d808c008c1..600ce1e0b0 100755
@@ -53,8 +53,4 @@ test_expect_failure 'Clone UC repo with lc name' '
 	test_must_fail git p4 clone //depot/uc/...
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 3c22f74bd4..fa1bba1dd9 100755
@@ -31,8 +31,4 @@ test_expect_success 'EDITOR with options' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 81e46acfa8..ef80f1690b 100755
@@ -193,8 +193,4 @@ test_expect_success 'Add a new file and clone path with new file (ignorecase)' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index c78477c19b..1bf7635016 100755
@@ -67,8 +67,4 @@ test_expect_success 'Delete iso8859-1 encoded paths and clone' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 1f2dc369bf..88b76dc4d6 100755
@@ -185,8 +185,4 @@ test_expect_success 'Run git p4 submit in repo configured with large file system
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index ed80ca858c..a28dbbdd56 100755
@@ -287,8 +287,4 @@ test_expect_success 'Add big files to repo and store files in LFS based on compr
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 1551845dc1..f049ff8229 100755
@@ -43,8 +43,4 @@ test_expect_failure 'clone depot with invalid UTF-16 file in non-verbose mode' '
 	git p4 clone --dest="$git" //depot
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index fa8b9daf1f..fd64afe064 100755
@@ -127,8 +127,4 @@ test_expect_success 'Clone repo subdir with all history' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 7433998f47..d3670bd7a2 100755
@@ -59,8 +59,4 @@ test_expect_success SYMLINKS 'change symbolic link to file' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index e20395c89f..ca6c2942bd 100755
@@ -54,8 +54,4 @@ test_expect_success 'Clone repo root path with all history' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 971aeeea1f..88cfb1fcd3 100755
@@ -92,8 +92,4 @@ test_expect_success 'check log message of changelist with more jobs' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 2ad1b0810d..3fb6960c18 100755
@@ -36,8 +36,4 @@ test_expect_success 'symlinked directory' '
 
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index be44c9751a..d743ca33ee 100755
@@ -96,8 +96,4 @@ test_expect_success 'submit description with extra info lines from verbose p4 ch
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 48ec7679b8..b649d1b7c3 100755
@@ -131,8 +131,4 @@ test_expect_success 'try to unshelve the change' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
 test_done

index 277d347012..1f3d879122 100755
@@ -72,9 +72,4 @@ test_expect_success 'git operation with expired ticket' '
 	)
 '
 
-test_expect_success 'kill p4d' '
-	kill_p4d
-'
-
-
 test_done

gitgitgadget

I'm finding that it's leaving p4d processes lying around.

e.g.

$ ./t9820-git-p4-editor-handling.sh
<passes>
$ ./t9820-git-p4-editor-handling.sh
<fails>

And also

$ ./t9800-git-p4-basic.sh
<starts running tests, but I get bored easily>
Ctrl-C

$ ./t9800-git-p4-basic.sh
<fails>

$ ps | grep p4d
21392 pts/1    00:00:00 p4d  <<<<<

Luke

Hi Luke,

That's a bummer!

Since I do not currently have a setup with p4d installed, can you run that with sh -x ... and see whether this code path is hit?

test_done () {
GIT_EXIT_OK=t

+ test -n "$immediate" || test_atexit_handler
+

        if test -n "$write_junit_xml" && test -n "$junit_xml_path"
        then

Oh, you're right. I think I need to do something in this line:

trap 'exit $?' INT

in t/test-lib.sh, something like

trap 'exit_code=$?; test_atexit_handler; exit $exit_code' INT

would you agree? (And: could you test?)

Thanks,
Dscho

All you need to do is to put p4 and p4d in your PATH.

https://www.perforce.com/downloads/helix-core-p4d
https://www.perforce.com/downloads/helix-command-line-client-p4

The server is free to use for a small number of users, you don't need to do anything to make it go.

Not sure.
Send me a patch and I can try it out.

Thanks,
Luke

Hi Luke,

I did download p4d.exe and p4.exe to $HOME/custom/p4/ (similar to the way ci/install-dependencies.sh does it), from

	http://filehost.perforce.com/perforce/r18.1/bin.ntx64/p4d.exe
	http://filehost.perforce.com/perforce/r18.1/bin.ntx64/p4.exe

and then prefixed PATH with $HOME/custom/p4, just to run t9820. However, it does not work here at all:

-- snip --
[...]
++ git p4 clone '--dest=/usr/src/git/azure-pipelines/t/trash directory.t9820-git-p4-editor-handling/git' //depot
Usage: git-p4 clone [options] //depot/path[@revRange]

[...]
-- snap --

I *think* the reason for that is that the MSYS path mangling kicks in when git.exe is called with the //depot argument (the leading // can be used in MSYS and MSYS2 to indicate that the non-MSYS .exe wants an argument that starts with a single slash, but is not a path that needs to be translated to a Windows path).

So I did the next best thing I could do: try things in the Windows Subsystem for Linux (AKA Bash on Ubuntu on Windows).

On a slight tangent: I made up my mind on the test -n "$immediate" part:
I will skip that. The daemon should be stopped also when -i was passed to the test script (to stop at the first failing test case). There is very, very little use in keeping the daemon alive in that case.

The commit message of "tests: introduce test_atexit" already made the case for that, but somehow I had not made up my mind yet.

No, that is not expected. For me, it looks quite a bit different, though:

-- snip --
[...]
+ test_done
+ GIT_EXIT_OK=t
+ test -n
+ test_atexit_handler
+ test : != { kill_p4d

                } && (exit "$eval_ret"); eval_ret=$?; :

+ setup_malloc_check
+ MALLOC_CHECK_=3 MALLOC_PERTURB_=165
+ export MALLOC_CHECK_ MALLOC_PERTURB_
+ test_eval_ { kill_p4d

                } && (exit "$eval_ret"); eval_ret=$?; :
+ test_eval_inner_ { kill_p4d
                } && (exit "$eval_ret"); eval_ret=$?; :
+ eval
                want_trace && set -x
                { kill_p4d
                } && (exit "$eval_ret"); eval_ret=$?; :

+ want_trace
+ test = t
+ kill_p4d
+ cat /home/wsl/git/wip/t/trash directory.t9820-git-p4-editor-handling/p4d.pid
+ pid=20093
+ retry_until_fail kill 20093
+ time_in_seconds
+ cd /
+ /usr/bin/python -c import time; print(int(time.time()))
+ timeout=1539681637
+ kill 20093
+ time_in_seconds
+ cd /
+ /usr/bin/python -c import time; print(int(time.time()))
+ test 1539681577 -gt 1539681637
+ sleep 1
+ true
+ time_in_seconds
+ cd /
+ /usr/bin/python -c import time; print(int(time.time()))
+ test 1539681578 -gt 1539681871
+ sleep 1
+ kill 20093
+ retry_until_fail kill -9 20093
+ time_in_seconds
+ cd /
+ /usr/bin/python -c import time; print(int(time.time()))
+ timeout=1539681638
+ kill -9 20093
+ test_must_fail kill 20093
+ _test_ok=
+ kill 20093
+ exit_code=1
+ test 1 -eq 0
+ test_match_signal 13 1
+ test 1 = 141
+ test 1 = 269
+ return 1
+ test 1 -gt 129
+ test 1 -eq 127
+ test 1 -eq 126
+ return 0
[...]
-- snap --

As you can see, it works quite well over here.

Maybe I could trouble you to fetch the vsts-ci branch from
https://github.com/dscho/git and test things on that one, just in case that anything important got "lost in the mail"?

So I tested this in WSL, and the relevant part of the output is this:

-- snip --
[...]
+ eval test_prereq_lazily_UTF8_NFD_TO_NFC=$2
+ test_prereq_lazily_UTF8_NFD_TO_NFC=

        # check whether FS converts nfd unicode to nfc
        au^CTraceback (most recent call last):
  File "/home/virtualbox/git/wip/git-p4", line 4143, in <module>
    main()
  File "/home/virtualbox/git/wip/git-p4", line 4137, in main
    if not cmd.run(args):
  File "/home/virtualbox/git/wip/git-p4", line 3925, in run
    if not P4Sync.run(self, depotPaths):
  File "/home/virtualbox/git/wip/git-p4", line 3824, in run
    system(["git", "symbolic-ref", head_ref, self.branch])
  File "/home/virtualbox/git/wip/git-p4", line 278, in system
    retcode = subprocess.call(cmd, shell=expand)
  File "/usr/lib/python2.7/subprocess.py", line 523, in call
    return Popen(*popenargs, **kwargs).wait()
  File "/usr/lib/python2.7/subprocess.py", line 1392, in wait
    pid, sts = _eintr_retry_call(os.waitpid, self.pid, 0)
  File "/usr/lib/python2.7/subprocess.py", line 476, in _eintr_retry_call
    return func(*args)
KeyboardInterrupt
+ exit 130
+ die
+ code=130
+ test_atexit_handler
+ test : != { kill_p4d
                } && (exit "$eval_ret"); eval_ret=$?; :
[...]
-- snap --

As you can see, the Ctrl+C managed to call die very well, without my proposed change:

I *think* what happens is that the INT trap already calls exit 130 alright, and then the EXIT trap comes into effect, as planned.

Just to make sure that it was not Python that caused the exit code 130, I tried it again, and luckily hit a spot where shell code was interpreted:

-- snip --
[...]
+ test_prereq_lazily_LONG_IS_64BIT=

        test 8 -le "$(build_option sizeof-long)"

+ test_lazy_^C+ exit 130
+ die
+ code=130
+ test_atexit_handler
+ test : != { kill_p4d

                } && (exit "$eval_ret"); eval_ret=$?; :

[...]
-- snap --

Same here. The Ctrl+C triggers an exit 130, this time I am certain that it comes from the INT trap, and that die in turn is most definitely from the EXIT trap, as intended.

My tests were performed with Dash, BTW, not with Bash:

-- snip --

$ dpkg -S $(which sh)
diversion by dash from: /bin/sh
diversion by dash to: /bin/sh.distrib
dash: /bin/sh
-- snap --

But I also repeated the tests with Bash, with almost the same result: the only difference was that I did not manage to get the traces of test_atexit_handler being called, but the p4d was no longer alive, so it must have been killed as intended.

I would *love* to figure out what is happening in your setup, and to fix it. In my hands, the patches work, though...

Ciao,
Dscho

Hi Eric,

Oy, oy, oy. This is definitely a left-over from debugging (as you can imagine, it is pretty slow to run t0000-init.sh on Windows, and if I add a test case, the development cycle is much faster with the trick you see aboive). This left-over even made it into Git for Windows' master branch! (Which is the reason I missed it before contributing v2).

Will fix,
Dscho

This patch makes the test suite hang in all four Travis CI build jobs with P4 installed without any of the P4 tests finishing.

Reverting this patch from the whole patch series makes it work again.

I've also tried to revert only this first hunk of the patch below, because based on the comment I thought it's worth a try, but it didn't really help. It did make a difference: the 300s watchdog timer eventually kicked in, and then the test scripts could finish successfully... but there are a lot of P4 test scripts, and with each taking 300s the build job still timeouted.

All this may (or may not) be related to and be a different symptom of the leftover p4d processes Luke mentioned. I couldn't reproduce any of this on my machine so far.

From: Johannes Schindelin <[email protected]>

The JUnit XML format lends itself to be presented in a powerful UI, where you can drill down to the information you are interested in very quickly.

For test failures, this usually means that you want to see the detailed trace of the failing tests.

With Travis CI, we passed the --verbose-log option to get those traces. However, that seems excessive, as we do not need/use the logs in almost all of those cases: only when a test fails do we have a way to include the trace.

So let's do something different when using Azure DevOps: let's run all the tests with --quiet first, and only if a failure is encountered, try to trace the commands as they are executed.

Of course, we cannot turn on --verbose-log after the fact. So let's just re-run the test with all the same options, adding --verbose-log. And then munging the output file into the JUnit XML on the fly.

Note: there is an off chance that re-running the test in verbose mode "fixes" the failures (and this does happen from time to time!). That is a possibility we should be able to live with. Ideally, we would label this as "Passed upon rerun", and Azure Pipelines even know about that outcome, but it is not available when using the JUnit XML format for now:
https://github.com/Microsoft/azure-pipelines-agent/blob/master/src/Agent.Worker/TestResults/JunitResultReader.cs

---
 t/test-lib.sh | 34 +++++++++++++++++++++++++++++++++-
 1 file changed, 33 insertions(+), 1 deletion(-)

index 6f9c1f5300..8a60e39364 100644
@@ -85,6 +85,13 @@ done,*)
 	test "$(cat "$BASE.exit")" = 0
 	exit
 	;;
+*' --write-junit-xml '*)
+	# record how to call this script *with* --verbose-log, in case
+	# we encounter a breakage
+	junit_rerun_options_sq="$(printf '%s\n' "$0" --verbose-log -x "$@" |
+		sed -e "s/'/'\\\\''/g" -e "s/^/'/" -e "s/\$/'/" |
+		tr '\012' ' ')"
+	;;
 esac
 
 # For repeatability, reset the environment to known value.

@@ -446,10 +453,31 @@ test_ok_ () {
 test_failure_ () {
 	if test -n "$write_junit_xml"
 	then
+		if test -z "$GIT_TEST_TEE_OUTPUT_FILE"
+		then
+			# clean up
+			test_atexit_handler
+
+			# re-run with --verbose-log
+			echo "# Re-running: $junit_rerun_options_sq" >&2
+
+			cd "$TEST_DIRECTORY" &&
+			eval "${TEST_SHELL_PATH}" "$junit_rerun_options_sq" \
+				>/dev/null 2>&1
+			status=$?
+
+			say_color "" "$(test 0 = $status ||
+				echo "not ")ok $test_count - (re-ran with trace)"
+			say "1..$test_count"
+			GIT_EXIT_OK=t
+			exit $status
+		fi
+
 		junit_insert="<failure message=\"not ok $test_count -"
 		junit_insert="$junit_insert $(xml_attr_encode "$1")\">"
 		junit_insert="$junit_insert $(xml_attr_encode \
-			"$(printf '%s\n' "$@" | sed 1d)")"
+			"$(cat "$GIT_TEST_TEE_OUTPUT_FILE")")"
+		>"$GIT_TEST_TEE_OUTPUT_FILE"
 		junit_insert="$junit_insert</failure>"
 		write_junit_xml_testcase "$1" "      $junit_insert"
 	fi

@@ -734,6 +762,10 @@ test_start_ () {
 	if test -n "$write_junit_xml"
 	then
 		junit_start=$(test-tool date getnanos)
+
+		# truncate output
+		test -z "$GIT_TEST_TEE_OUTPUT_FILE" ||
+		>"$GIT_TEST_TEE_OUTPUT_FILE"
 	fi
 }
 

gitgitgadget

As someone who has dug into a few occasional failures found by Travis CI, I'd say that the output of '--verbose-log -x' is not excessive, but downright essential.

Any CI system worth its salt should provide as much information about any failures as possible, especially when it was lucky enough to stumble upon a rare and hard to reproduce non-deterministic failure.

Hi Gábor,

I agree that the output is essential for drilling down into failures. This paragraph, however, talks about the general case: where there are *no* failures. See here:

I would agree with you if more people started to pay attention to our CI failures. And if we had some sort of a development model where a CI failure would halt development on that particular topic until the failure is fixed, with the responsibility assigned to somebody to fix it.

This is not the case here, though. pu is broken for ages, at least on Windows, and even a *single* topic is enough to do that. And this is even worse with flakey tests. I cannot remember *how often* I saw CI failures in t5570-git-daemon.sh, for example. It is rare enough that it is obvious that this is a problem of the *regression test*, rather than a problem of the code that is to be tested.

So I would suggest to go forward with my proposed strategy for the moment, right up until the time when we have had the resources to fix t5570, for starters.

Ciao,
Dscho

But you don't know in advance whether there will be any failures or not, so it only makes sense to run all tests with '--verbose-log -x' by default, just in case a Heisenbug decides to make an appearance.

Some occasional failures in t5570 are actually caused by issues in Git on certain platforms:

https://public-inbox.org/git/CAM0VKj=MCS+cmOgzf_XyPeb+qZrFmuMH52-PV_NDMZA9X+rRoA@mail.gmail.com/T/#u

I don't really understand what the occasional failures in t5570 have to do with the amount of information a CI system should gather about failures in general. Or how many people pay attention to it, or what kind of development model we have, for that matter. The way I see it these are unrelated issues, and a CI system should always provide as much information about failures as possible. If only a few people pay attention to it, then for the sake of those few.

Hi Gábor,

I see it plenty of times that too many CI failures essentially render every developer numb.

If every 3rd CI run causes a failure, and seemingly every of these failures indicates a mistake in the regression test, rather than a regression, developers stop paying attention.

Which is the exact opposite of what I want to achieve here.

Ciao,
Dscho

From: Johannes Schindelin <[email protected]>

Sometimes, failures in a test case are actually caused by issues in earlier test cases.

To make it easier to see those issues, let's attach the output from before the failing test case (i.e. stdout/stderr since the previous failing test case, or the start of the test script). This will be visible in the "Attachments" of the details of the failed test.

---
 t/test-lib.sh | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)

index 8a60e39364..5f62418f9c 100644
@@ -479,6 +479,9 @@ test_failure_ () {
 			"$(cat "$GIT_TEST_TEE_OUTPUT_FILE")")"
 		>"$GIT_TEST_TEE_OUTPUT_FILE"
 		junit_insert="$junit_insert</failure>"
+		junit_insert="$junit_insert<system-err>$(xml_attr_encode \
+			"$(cat "$GIT_TEST_TEE_OUTPUT_FILE.err")")</system-err>"
+		>"$GIT_TEST_TEE_OUTPUT_FILE.err"
 		write_junit_xml_testcase "$1" "      $junit_insert"
 	fi
 	test_failure=$(($test_failure + 1))

@@ -763,9 +766,12 @@ test_start_ () {
 	then
 		junit_start=$(test-tool date getnanos)
 
-		# truncate output
-		test -z "$GIT_TEST_TEE_OUTPUT_FILE" ||
-		>"$GIT_TEST_TEE_OUTPUT_FILE"
+		# append to future <system-err>; truncate output
+		test -z "$GIT_TEST_TEE_OUTPUT_FILE" || {
+			cat "$GIT_TEST_TEE_OUTPUT_FILE" \
+				>>"$GIT_TEST_TEE_OUTPUT_FILE.err"
+			>"$GIT_TEST_TEE_OUTPUT_FILE"
+		}
 	fi
 }
 

gitgitgadget

From: Johannes Schindelin <[email protected]>

Just like so many other OSS projects, we now also have a build badge.

---
 README.md | 2 ++
 1 file changed, 2 insertions(+)

index f920a42fad..bf4780c22d 100644
@@ -1,3 +1,5 @@
+[![Build Status](https:/dev.azure.com/git/git/_apis/build/status/test-git.git)](https://dev.azure.com/git/git/_build/latest?definitionId=2)
+
 Git - fast, scalable, distributed revision control system
 =========================================================
 

gitgitgadget

Hi Eric,

I *swear* I fixed this already. Darn.

Thank you,
Dscho

From: Johannes Schindelin <[email protected]>

When building a PR, TRAVIS_BRANCH refers to the *target branch*. Therefore, if a PR targets master, and master happened to be tagged, we skipped the build by mistake.

Fix this by using TRAVIS_PULL_REQUEST_BRANCH (i.e. the *source branch*) when available, falling back to TRAVIS_BRANCH (i.e. for CI builds, also known as "push builds").

---
 ci/lib.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

index 584abcd529..e1858ae609 100755
@@ -3,7 +3,7 @@
 if test true = "$TRAVIS"
 then
 	# We are running within Travis CI
-	CI_BRANCH="$TRAVIS_BRANCH"
+	CI_BRANCH="${TRAVIS_PULL_REQUEST_BRANCH:-$TRAVIS_BRANCH}"
 	CI_COMMIT="$TRAVIS_COMMIT"
 	CI_JOB_ID="$TRAVIS_JOB_ID"
 	CI_JOB_NUMBER="$TRAVIS_JOB_NUMBER"

gitgitgadget

This all makes sense, but this patch is fixing a long-standing issue in our Travis CI build scripts (present since 09f5e9746c (travis-ci:
skip a branch build if equal tag is present, 2017-09-10)), so it should be the first in the series. So it could be picked up and perhaps even graduated faster than the rest of this patch series.

Hi Junio,

Makes sense.

Thanks,
Dscho

Hi Johannes,

Thanks for putting this together, and offering to build Git on Azure Pipelines. I haven't followed v1 of this series very closely, so please excuse me if my comments have already been addressed, and I missed them in a skim of the last revision.

Agree.

I wonder if Travis' recent announcement [1] affects this at all. To summarize [1], Travis is offering an early version of adding Windows to their list of supported builder operations systems. This brings the list to macOS, Linux, and Windows, which I think satisfies what we would like to regularly build git.git on.

Would we like to abandon Travis as our main CI service for upstream git.git, and build on Azure Pipelines only? If so, I think that this is an OK way to go, but I think that I would be opposed to having more than one build system. (FWIW, we tend to _three_ for Git LFS, and it can be a hassle at times).

I see some benefit to sticking with Travis, since we already have a build configuration that works there. But, you've done the work to "port" that build configuration over to Azure, so perhaps the point is moot.

This would be a concession of [1], in my mind: is it possible to run the tests on Windows in a time such that Travis will not time out?

That's pretty cool. Travis doesn't support this (to the best of my knowledge).

Thanks,
Taylor

[1]: https://blog.travis-ci.com/2018-10-11-windows-early-release