Skip to contentSkip to conversation

gitlab: avoid timeouts in windows cross builds

PATCH In review Daniel P. Berrangé 5 participants 13 messages Aug 10, 2021 – Aug 18, 2021

lore
Share & tools
lore
#0 lore raw

cover gitlab: avoid timeouts in windows cross builds

The win64 cross build is quite often hitting the job timeout, despite having an elevated timeout of 80 minutes. A typical time is 75-78 minutes in my tests, which leaves little headroom.

I'm not inclined to increase the timeout beyond 80 minutes as this is already unreasonably long.

Interestingly the win32 job doesn't timeout in the same way. In local tests I've found the mingw32 gcc is as much as 50% faster than mingw64 gcc in building QEMU. This explains at least why we only see win64 gitlab builds timeout normally.

None the less, this series thus cuts down on the amount of stuff built in both win32 and win64 jobs, as well as cross build jobs more generally.

With this applied:

- win32 job runs in 40 minutes instead of 55 minutes
- win64 job runs in 50 minutes instead of 75 minutes
- other cross build jobs are a few minutes faster

this should give us plenty of headroom to avoid hitting the job timeouts.

Daniel P. Berrangé (2):
gitlab: exclude sparc-softmmu and riscv32-softmmu from cross builds
gitlab: skip many more targets in windows cross builds

 .gitlab-ci.d/crossbuild-template.yml | 3 ++-
 .gitlab-ci.d/crossbuilds.yml         | 2 ++
 2 files changed, 4 insertions(+), 1 deletion(-)
Signature · 1 line
2.31.1
#1 lore raw

PATCH 1/2 gitlab: exclude sparc-softmmu and riscv32-softmmu from cross builds

We need to cut down compile time by excluding more targets. Both these targets still have their 64-bit variant enabled, so the loss of coverage is mitigated to some degree.

---
 .gitlab-ci.d/crossbuild-template.yml | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
.gitlab-ci.d/crossbuild-template.yml +2 -1
index 7d3ad00a1e..cfb576b54c 100644
@@ -9,7 +9,8 @@
       ../configure --enable-werror --disable-docs $QEMU_CONFIGURE_OPTS
         --disable-user --target-list-exclude="arm-softmmu cris-softmmu
           i386-softmmu microblaze-softmmu mips-softmmu mipsel-softmmu
-          mips64-softmmu ppc-softmmu sh4-softmmu xtensa-softmmu"
+          mips64-softmmu ppc-softmmu riscv32-softmmu sh4-softmmu
+          sparc-softmmu xtensa-softmmu"
     - make -j$(expr $(nproc) + 1) all check-build $MAKE_CHECK_ARGS
     - if grep -q "EXESUF=.exe" config-host.mak;
       then make installer;
Signature · 1 line
2.31.1
Reviewed-by on v1 1/2
#2 lore raw

Re: [PATCH 1/2] gitlab: exclude sparc-softmmu and riscv32-softmmu from cross builds

Quoted 10 lines · Daniel P. Berrangé

We need to cut down compile time by excluding more targets. Both these targets still have their 64-bit variant enabled, so the loss of coverage is mitigated to some degree.

---
 .gitlab-ci.d/crossbuild-template.yml | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
#3 lore raw

Re: [PATCH 1/2] gitlab: exclude sparc-softmmu and riscv32-softmmu from cross builds

In reply to Daniel P. Berrangé

+Mark/Artyom

Quoted 14 lines · Daniel P. Berrangé

We need to cut down compile time by excluding more targets. Both these targets still have their 64-bit variant enabled, so the loss of coverage is mitigated to some degree.

---
 .gitlab-ci.d/crossbuild-template.yml | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
.gitlab-ci.d/crossbuild-template.yml +0 -0
index 7d3ad00a1e..cfb576b54c 100644

.gitlab-ci.d/crossbuild-template.yml line 13

Quoted 6 lines
.gitlab-ci.d/crossbuild-template.yml +2 -1
@@ -9,7 +9,8 @@
       ../configure --enable-werror --disable-docs $QEMU_CONFIGURE_OPTS
         --disable-user --target-list-exclude="arm-softmmu cris-softmmu
           i386-softmmu microblaze-softmmu mips-softmmu mipsel-softmmu
-          mips64-softmmu ppc-softmmu sh4-softmmu xtensa-softmmu"
+          mips64-softmmu ppc-softmmu riscv32-softmmu sh4-softmmu
+          sparc-softmmu xtensa-softmmu"

I disagree for sparc, as there are some differences (sparc64 is not a superset of it). However I'm not against this patch.

Quoted 4 lines
.gitlab-ci.d/crossbuild-template.yml +0 -0
     - make -j$(expr $(nproc) + 1) all check-build $MAKE_CHECK_ARGS
     - if grep -q "EXESUF=.exe" config-host.mak;
       then make installer;
#4 lore raw

Re: [PATCH 1/2] gitlab: exclude sparc-softmmu and riscv32-softmmu from cross builds

In reply to Daniel P. Berrangé
Quoted 14 lines · Daniel P. Berrangé

We need to cut down compile time by excluding more targets. Both these targets still have their 64-bit variant enabled, so the loss of coverage is mitigated to some degree.

---
  .gitlab-ci.d/crossbuild-template.yml | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
.gitlab-ci.d/crossbuild-template.yml +0 -0
index 7d3ad00a1e..cfb576b54c 100644

.gitlab-ci.d/crossbuild-template.yml line 16

Quoted 9 lines
.gitlab-ci.d/crossbuild-template.yml +2 -1
@@ -9,7 +9,8 @@
        ../configure --enable-werror --disable-docs $QEMU_CONFIGURE_OPTS
          --disable-user --target-list-exclude="arm-softmmu cris-softmmu
            i386-softmmu microblaze-softmmu mips-softmmu mipsel-softmmu
-          mips64-softmmu ppc-softmmu sh4-softmmu xtensa-softmmu"
+          mips64-softmmu ppc-softmmu riscv32-softmmu sh4-softmmu
+          sparc-softmmu xtensa-softmmu"
      - make -j$(expr $(nproc) + 1) all check-build $MAKE_CHECK_ARGS
      - if grep -q "EXESUF=.exe" config-host.mak;
        then make installer;

I'd prefer to keep sparc-softmmu if possible, simply because my everyday platform is Linux and so having a cross-build for Windows will catch things that I may miss on a day-to-day basis. Is sparc-softmmu currently enabled as part of the native MINGW64 build?

If I go to my Gitlab QEMU fork Settings -> CI/CD -> Variables there is an option to set variables that can be used in job scripts. Perhaps this could be used so that I can configure my personal QEMU fork to always run sparc-softmmu builds when preparing PRs even if they aren't enabled for everyone by default? At least this would then allow me to spot any breakage before sending a final PR to Peter.

ATB,

Mark.

#5 lore raw

Re: [PATCH 1/2] gitlab: exclude sparc-softmmu and riscv32-softmmu from cross builds

Quoted 22 lines · Mark Cave-Ayland
Quoted 20 lines · Daniel P. Berrangé

We need to cut down compile time by excluding more targets. Both these targets still have their 64-bit variant enabled, so the loss of coverage is mitigated to some degree.

---
  .gitlab-ci.d/crossbuild-template.yml | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
.gitlab-ci.d/crossbuild-template.yml +2 -1
index 7d3ad00a1e..cfb576b54c 100644
@@ -9,7 +9,8 @@
        ../configure --enable-werror --disable-docs $QEMU_CONFIGURE_OPTS
          --disable-user --target-list-exclude="arm-softmmu cris-softmmu
            i386-softmmu microblaze-softmmu mips-softmmu mipsel-softmmu
-          mips64-softmmu ppc-softmmu sh4-softmmu xtensa-softmmu"
+          mips64-softmmu ppc-softmmu riscv32-softmmu sh4-softmmu
+          sparc-softmmu xtensa-softmmu"
Quoted 8 lines

.gitlab-ci.d/crossbuild-template.yml line 16

Quoted 3 lines
.gitlab-ci.d/crossbuild-template.yml +0 -0
      - make -j$(expr $(nproc) + 1) all check-build $MAKE_CHECK_ARGS
      - if grep -q "EXESUF=.exe" config-host.mak;
        then make installer;

I'd prefer to keep sparc-softmmu if possible, simply because my everyday platform is Linux and so having a cross-build for Windows will catch things that I may miss on a day-to-day basis. Is sparc-softmmu currently enabled as part of the native MINGW64 build?

Note the builds still include 'sparc64-softmmu', so we're only loosing cover in places where it diverges fromthe 64-bit build, but this will sstil get coverage in native builds.

Quoted 6 lines

If I go to my Gitlab QEMU fork Settings -> CI/CD -> Variables there is an option to set variables that can be used in job scripts. Perhaps this could be used so that I can configure my personal QEMU fork to always run sparc-softmmu builds when preparing PRs even if they aren't enabled for everyone by default? At least this would then allow me to spot any breakage before sending a final PR to Peter.

Separately from this I'm doing some work to make things more configurable

https://lists.gnu.org/archive/html/qemu-devel/2021-08/msg02102.html

but not on this level of granularity.

Rather than globally excluding from all cross builds, I think we ought to split it up more equitably acrss the different builds in some way.

Regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#6 lore raw

Re: [PATCH 1/2] gitlab: exclude sparc-softmmu and riscv32-softmmu from cross builds

Quoted 32 lines · Daniel P. Berrangé
Quoted 22 lines · Mark Cave-Ayland
Quoted 20 lines · Daniel P. Berrangé

We need to cut down compile time by excluding more targets. Both these targets still have their 64-bit variant enabled, so the loss of coverage is mitigated to some degree.

---
   .gitlab-ci.d/crossbuild-template.yml | 3 ++-
   1 file changed, 2 insertions(+), 1 deletion(-)
.gitlab-ci.d/crossbuild-template.yml +2 -1
index 7d3ad00a1e..cfb576b54c 100644
@@ -9,7 +9,8 @@
         ../configure --enable-werror --disable-docs $QEMU_CONFIGURE_OPTS
           --disable-user --target-list-exclude="arm-softmmu cris-softmmu
             i386-softmmu microblaze-softmmu mips-softmmu mipsel-softmmu
-          mips64-softmmu ppc-softmmu sh4-softmmu xtensa-softmmu"
+          mips64-softmmu ppc-softmmu riscv32-softmmu sh4-softmmu
+          sparc-softmmu xtensa-softmmu"
Quoted 8 lines

.gitlab-ci.d/crossbuild-template.yml line 16

Quoted 3 lines
.gitlab-ci.d/crossbuild-template.yml +0 -0
       - make -j$(expr $(nproc) + 1) all check-build $MAKE_CHECK_ARGS
       - if grep -q "EXESUF=.exe" config-host.mak;
         then make installer;

I'd prefer to keep sparc-softmmu if possible, simply because my everyday platform is Linux and so having a cross-build for Windows will catch things that I may miss on a day-to-day basis. Is sparc-softmmu currently enabled as part of the native MINGW64 build?

Quoted 3 lines

Note the builds still include 'sparc64-softmmu', so we're only loosing cover in places where it diverges fromthe 64-bit build, but this will sstil get coverage in native builds.

Okay I see. From my perspective sparc-softmmu covers more corner cases simply because everything is wired up as cMMIO peripherals on a custum bus compared with sparc64-softmmu which is a PCI machine and so there is already a lot of overlap there. However...

Quoted 7 lines
Quoted 6 lines

If I go to my Gitlab QEMU fork Settings -> CI/CD -> Variables there is an option to set variables that can be used in job scripts. Perhaps this could be used so that I can configure my personal QEMU fork to always run sparc-softmmu builds when preparing PRs even if they aren't enabled for everyone by default? At least this would then allow me to spot any breakage before sending a final PR to Peter.

Quoted 8 lines

Separately from this I'm doing some work to make things more configurable

   https://lists.gnu.org/archive/html/qemu-devel/2021-08/msg02102.html

but not on this level of granularity.

Rather than globally excluding from all cross builds, I think we ought to split it up more equitably acrss the different builds in some way.

From my perspective as long as I have something in Gitlab that I can push and then get a green CI build that covers both sparc platforms on Linux/Windows then I'm not too concerned about the implementation details. As long as the combination is covered in a job somewhere in your latest set of build changes then that's okay with me.

ATB,

Mark.

#7 lore raw

Re: [PATCH 1/2] gitlab: exclude sparc-softmmu and riscv32-softmmu from cross builds

Quoted 65 lines · Mark Cave-Ayland
Quoted 39 lines · Daniel P. Berrangé
Quoted 29 lines · Mark Cave-Ayland
Quoted 27 lines · Daniel P. Berrangé

We need to cut down compile time by excluding more targets. Both these targets still have their 64-bit variant enabled, so the loss of coverage
is mitigated to some degree.

---
   .gitlab-ci.d/crossbuild-template.yml | 3 ++-

   1 file changed, 2 insertions(+), 1 deletion(-)
.gitlab-ci.d/crossbuild-template.yml +0 -0
index 7d3ad00a1e..cfb576b54c 100644
@@ -9,7 +9,8 @@

         ../configure --enable-werror --disable-docs $QEMU_CONFIGURE_OPTS
           --disable-user --target-list-exclude="arm-softmmu cris-softmmu
             i386-softmmu microblaze-softmmu mips-softmmu mipsel-softmmu
-          mips64-softmmu ppc-softmmu sh4-softmmu xtensa-softmmu" +          mips64-softmmu ppc-softmmu riscv32-softmmu sh4-softmmu +          sparc-softmmu xtensa-softmmu"
       - make -j$(expr $(nproc) + 1) all check-build $MAKE_CHECK_ARGS
       - if grep -q "EXESUF=.exe" config-host.mak;

Quoted 8 lines
Quoted 1 line

         then make installer;

I'd prefer to keep sparc-softmmu if possible, simply because my everyday platform is Linux and so having a cross-build for Windows will catch things
that I may miss on a day-to-day basis. Is sparc-softmmu currently enabled as
part of the native MINGW64 build?

Quoted 3 lines

Note the builds still include  'sparc64-softmmu', so we're only loosing cover in places where it diverges fromthe 64-bit build, but this will sstil get coverage in native builds.

Okay I see. From my perspective sparc-softmmu covers more corner cases simply because everything is wired up as cMMIO peripherals on a custum bus compared with sparc64-softmmu which is a PCI machine and so there is already a lot of overlap there. However...

Quoted 16 lines
Quoted 9 lines

If I go to my Gitlab QEMU fork Settings -> CI/CD -> Variables there is an
option to set variables that can be used in job scripts. Perhaps this could
be used so that I can configure my personal QEMU fork to always run sparc-softmmu builds when preparing PRs even if they aren't enabled for everyone by default? At least this would then allow me to spot any breakage
before sending a final PR to Peter.

Separately from this I'm doing some work to make things more configurable

   https://lists.gnu.org/archive/html/qemu-devel/2021-08/msg02102.html

but not on this level of granularity.

Quoted 8 lines
Quoted 2 lines

Rather than globally excluding from all cross builds, I think we ought to split it up more equitably acrss the different builds in some way.

From my perspective as long as I have something in Gitlab that I can push and then get a green CI build that covers both sparc platforms on Linux/Windows then I'm not too concerned about the implementation details. As long as the combination is covered in a job somewhere in your latest set of build changes then that's okay with me.

Your use case seems similar to mine: I overload the current mainstream pipeline with extra jobs. See:
https://docs.gitlab.com/ee/ci/pipelines/settings.html#specify-a-custom-cicd-configuration-file

My namespace CI points at a custom config in some branch:
https://gitlab.com/philmd/qemu/raw/my_gci_config/.gitlab-ci.d/philmd.yml

This files simply contains:

include:
# Keep mainstream config ...
- local: '/.gitlab-ci.d/qemu-project.yml'

  # ... but include my extra jobs.
  -
'https://gitlab.com/philmd/qemu/raw/my_gci_config/.gitlab-ci.d/nvme_wip.yml'
  -
'https://gitlab.com/philmd/qemu/raw/my_gci_config/.gitlab-ci.d/mips_slow.yml'
  -
'https://gitlab.com/philmd/qemu/raw/my_gci_config/.gitlab-ci.d/philmd_runners.yml'
#8 lore raw

PATCH 2/2 gitlab: skip many more targets in windows cross builds

In reply to Daniel P. Berrangé

The windows cross builds still take way too long in gitlab CI, so need more targets to be skipped. We don't want to hurt coverage of other cross builds more though, so we let jobs fine tune with a new env variale $CROSS_SKIP_TARGETS.

We take the set of targets that are considered relatively niche or very old architectures, and skip approx half of them in win32 builds and the other half of them in win64.

---
 .gitlab-ci.d/crossbuild-template.yml | 2 +-
 .gitlab-ci.d/crossbuilds.yml         | 2 ++
 2 files changed, 3 insertions(+), 1 deletion(-)
.gitlab-ci.d/crossbuild-template.yml +1 -1
index cfb576b54c..10d22dcf6c 100644
@@ -10,7 +10,7 @@
         --disable-user --target-list-exclude="arm-softmmu cris-softmmu
           i386-softmmu microblaze-softmmu mips-softmmu mipsel-softmmu
           mips64-softmmu ppc-softmmu riscv32-softmmu sh4-softmmu
-          sparc-softmmu xtensa-softmmu"
+          sparc-softmmu xtensa-softmmu $CROSS_SKIP_TARGETS"
     - make -j$(expr $(nproc) + 1) all check-build $MAKE_CHECK_ARGS
     - if grep -q "EXESUF=.exe" config-host.mak;
       then make installer;
.gitlab-ci.d/crossbuilds.yml +2 -0
index 4ff3aa3cfc..a5f9dbcbeb 100644
@@ -160,6 +160,7 @@ cross-win32-system:
     job: win32-fedora-cross-container
   variables:
     IMAGE: fedora-win32-cross
+    CROSS_SKIP_TARGETS: or1k-softmmu rx-softmmu sh4eb-softmmu sparc64-softmmu tricore-softmmu xtensaeb-softmmu
   artifacts:
     paths:
       - build/qemu-setup*.exe
@@ -170,6 +171,7 @@ cross-win64-system:
     job: win64-fedora-cross-container
   variables:
     IMAGE: fedora-win64-cross
+    CROSS_SKIP_TARGETS: alpha-softmmu avr-softmmu hppa-softmmu m68k-softmmu microblazeel-softmmu mips64el-softmmu nios2-softmmu
   artifacts:
     paths:
       - build/qemu-setup*.exe
Signature · 1 line
2.31.1
Reviewed-by on v1 2/2
#9 lore raw

Re: [PATCH 2/2] gitlab: skip many more targets in windows cross builds

Quoted 5 lines · Daniel P. Berrangé

The windows cross builds still take way too long in gitlab CI, so need more targets to be skipped. We don't want to hurt coverage of other cross builds more though, so we let jobs fine tune with a new env variale $CROSS_SKIP_TARGETS.

s/variale/variable

Quoted 11 lines

We take the set of targets that are considered relatively niche or very old architectures, and skip approx half of them in win32 builds and the other half of them in win64.

---
 .gitlab-ci.d/crossbuild-template.yml | 2 +-
 .gitlab-ci.d/crossbuilds.yml         | 2 ++
 2 files changed, 3 insertions(+), 1 deletion(-)
#10 lore raw

Re: [PATCH 2/2] gitlab: skip many more targets in windows cross builds

In reply to Daniel P. Berrangé
Quoted 41 lines · Daniel P. Berrangé

The windows cross builds still take way too long in gitlab CI, so need more targets to be skipped. We don't want to hurt coverage of other cross builds more though, so we let jobs fine tune with a new env variale $CROSS_SKIP_TARGETS.

We take the set of targets that are considered relatively niche or very old architectures, and skip approx half of them in win32 builds and the other half of them in win64.

---
 .gitlab-ci.d/crossbuild-template.yml | 2 +-
 .gitlab-ci.d/crossbuilds.yml         | 2 ++
 2 files changed, 3 insertions(+), 1 deletion(-)
.gitlab-ci.d/crossbuild-template.yml +1 -1
index cfb576b54c..10d22dcf6c 100644
@@ -10,7 +10,7 @@
         --disable-user --target-list-exclude="arm-softmmu cris-softmmu
           i386-softmmu microblaze-softmmu mips-softmmu mipsel-softmmu
           mips64-softmmu ppc-softmmu riscv32-softmmu sh4-softmmu
-          sparc-softmmu xtensa-softmmu"
+          sparc-softmmu xtensa-softmmu $CROSS_SKIP_TARGETS"
     - make -j$(expr $(nproc) + 1) all check-build $MAKE_CHECK_ARGS
     - if grep -q "EXESUF=.exe" config-host.mak;
       then make installer;
.gitlab-ci.d/crossbuilds.yml +1 -0
index 4ff3aa3cfc..a5f9dbcbeb 100644
@@ -160,6 +160,7 @@ cross-win32-system:
     job: win32-fedora-cross-container
   variables:
     IMAGE: fedora-win32-cross
+    CROSS_SKIP_TARGETS: or1k-softmmu rx-softmmu sh4eb-softmmu sparc64-softmmu tricore-softmmu xtensaeb-softmmu
   artifacts:
     paths:
       - build/qemu-setup*.exe

.gitlab-ci.d/crossbuilds.yml cross-win64-system: · line 174

Quoted 4 lines
.gitlab-ci.d/crossbuilds.yml +1 -0
@@ -170,6 +171,7 @@ cross-win64-system:
     job: win64-fedora-cross-container
   variables:
     IMAGE: fedora-win64-cross
+    CROSS_SKIP_TARGETS: alpha-softmmu avr-softmmu hppa-softmmu m68k-softmmu microblazeel-softmmu mips64el-softmmu nios2-softmmu

It looks you inverted the lists. We expect more Win64 users, and these are the targets of interest. I'd keep here (skipping in Win32):

alpha-softmmu avr-softmmu hppa-softmmu m68k-softmmu microblazeel-softmmu mips64el-softmmu sparc64-softmmu

And skip (keep them in win32):

or1k-softmmu rx-softmmu sh4eb-softmmu nios2-softmmu tricore-softmmu xtensaeb-softmmu

Quoted 4 lines
.gitlab-ci.d/crossbuilds.yml +0 -0
   artifacts:
     paths:
       - build/qemu-setup*.exe
#11 lore raw

Re: [PATCH 2/2] gitlab: skip many more targets in windows cross builds

Quoted 41 lines · Philippe Mathieu-Daudé
Quoted 40 lines · Daniel P. Berrangé

The windows cross builds still take way too long in gitlab CI, so need more targets to be skipped. We don't want to hurt coverage of other cross builds more though, so we let jobs fine tune with a new env variale $CROSS_SKIP_TARGETS.

We take the set of targets that are considered relatively niche or very old architectures, and skip approx half of them in win32 builds and the other half of them in win64.

---
 .gitlab-ci.d/crossbuild-template.yml | 2 +-
 .gitlab-ci.d/crossbuilds.yml         | 2 ++
 2 files changed, 3 insertions(+), 1 deletion(-)
.gitlab-ci.d/crossbuild-template.yml +1 -1
index cfb576b54c..10d22dcf6c 100644
@@ -10,7 +10,7 @@
         --disable-user --target-list-exclude="arm-softmmu cris-softmmu
           i386-softmmu microblaze-softmmu mips-softmmu mipsel-softmmu
           mips64-softmmu ppc-softmmu riscv32-softmmu sh4-softmmu
-          sparc-softmmu xtensa-softmmu"
+          sparc-softmmu xtensa-softmmu $CROSS_SKIP_TARGETS"
     - make -j$(expr $(nproc) + 1) all check-build $MAKE_CHECK_ARGS
     - if grep -q "EXESUF=.exe" config-host.mak;
       then make installer;
.gitlab-ci.d/crossbuilds.yml +1 -0
index 4ff3aa3cfc..a5f9dbcbeb 100644
@@ -160,6 +160,7 @@ cross-win32-system:
     job: win32-fedora-cross-container
   variables:
     IMAGE: fedora-win32-cross
+    CROSS_SKIP_TARGETS: or1k-softmmu rx-softmmu sh4eb-softmmu sparc64-softmmu tricore-softmmu xtensaeb-softmmu
   artifacts:
     paths:
       - build/qemu-setup*.exe
Quoted 8 lines

.gitlab-ci.d/crossbuilds.yml cross-win64-system: · line 174

Quoted 5 lines
.gitlab-ci.d/crossbuilds.yml +1 -0
@@ -170,6 +171,7 @@ cross-win64-system:
     job: win64-fedora-cross-container
   variables:
     IMAGE: fedora-win64-cross
+    CROSS_SKIP_TARGETS: alpha-softmmu avr-softmmu hppa-softmmu m68k-softmmu microblazeel-softmmu mips64el-softmmu nios2-softmmu

It looks you inverted the lists. We expect more Win64 users, and these are the targets of interest. I'd keep here (skipping in Win32):

I wouldn't say inverted, because the split was entirely arbitrary. While 32-bit vs 64-bit could conceivably show a difference, in practice those differences are likely to already be caught by non-windows builds. These jobs are most important for catching UNIX vs Windows differences, so in that view whether a target is built for win32 vs win64 doesn't matter.

Quoted 8 lines

alpha-softmmu avr-softmmu hppa-softmmu m68k-softmmu microblazeel-softmmu mips64el-softmmu sparc64-softmmu

And skip (keep them in win32):

or1k-softmmu rx-softmmu sh4eb-softmmu nios2-softmmu tricore-softmmu xtensaeb-softmmu

Regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#12 lore raw
In reply to Daniel P. Berrangé
Quoted 11 lines · Daniel P. Berrangé

The win64 cross build is quite often hitting the job timeout, despite having an elevated timeout of 80 minutes. A typical time is 75-78 minutes in my tests, which leaves little headroom.

I'm not inclined to increase the timeout beyond 80 minutes as this is already unreasonably long.

Interestingly the win32 job doesn't timeout in the same way. In local tests I've found the mingw32 gcc is as much as 50% faster than mingw64 gcc in building QEMU. This explains at least why we only see win64 gitlab builds timeout normally.

I've had similar experiences in the past, and once even already sent a patch:

https://patchwork.ozlabs.org/project/qemu-devel/patch/[email protected]/

But your apprach is looking way nicer, indeed!

Queued to my testing-next branch now:

https://gitlab.com/thuth/qemu/-/commits/testing-next/

Thanks,
Thomas