Skip to contentSkip to conversation

do not set GIT_TEST_MAINT_SCHEDULER where it does not matter

PATCH In review Junio C Hamano 3 participants 5 messages Mar 29, 2024 – Mar 31, 2024

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 do not set GIT_TEST_MAINT_SCHEDULER where it does not matter

31345d55 (maintenance: extract platform-specific scheduling, 2020-11-24) added code to t/test-lib.sh for everybody to set GIT_TEST_MAINT_SCHEDULER to a "safe" value and instructed the test writers to set the variable locally when their test wants to check the scheduler integration.

But it did so without "export GIT_TEST_MAINT_SCHEDULER", so the setting does not seem to have any effect anyway. Instead of setting it to a "safe" value, just unset it.

---

* t7900 (maintenance) uses many tests that does one-shot export of the variable, and one test that sets the value to its safe "failure" value and exports it at the end.

t9210 (scaler) sets up a safe fake scheduler and exports it before doing any of its tests.

Nobody else that includes t/test-lib.sh mentions this variable.

 t/test-lib.sh | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
t/test-lib.sh +3 -3
index c8af8dab79..48345864f4 100644
@@ -1959,9 +1959,9 @@ test_lazy_prereq DEFAULT_REPO_FORMAT '
 # Ensure that no test accidentally triggers a Git command
 # that runs the actual maintenance scheduler, affecting a user's
 # system permanently.
-# Tests that verify the scheduler integration must set this locally
-# to avoid errors.
-GIT_TEST_MAINT_SCHEDULER="none:exit 1"
+# Tests that verify the scheduler integration must set and
+# export this variable locally.
+sane_unset GIT_TEST_MAINT_SCHEDULER
 
 # Does this platform support `git fsmonitor--daemon`
 #
#1 lore raw
Quoted 9 lines · Junio C Hamano

31345d55 (maintenance: extract platform-specific scheduling, 2020-11-24) added code to t/test-lib.sh for everybody to set GIT_TEST_MAINT_SCHEDULER to a "safe" value and instructed the test writers to set the variable locally when their test wants to check the scheduler integration.

But it did so without "export GIT_TEST_MAINT_SCHEDULER", so the setting does not seem to have any effect anyway. Instead of setting it to a "safe" value, just unset it.

I agree that the missing export makes this a do-nothing assignment. In fact, that problem traces back to the original 2fec604f8d (maintenance: add start/stop subcommands, 2020-09-11).

w/t/test-lib.sh test_lazy_prereq DEFAULT_REPO_FORMAT ' · line 1964

Quoted 13 lines

---

w/t/test-lib.sh renamed from c/t/test-lib.sh +3 -3
@@ -1959,9 +1959,9 @@ test_lazy_prereq DEFAULT_REPO_FORMAT '
 # Ensure that no test accidentally triggers a Git command
 # that runs the actual maintenance scheduler, affecting a user's
 # system permanently.
-# Tests that verify the scheduler integration must set this locally
-# to avoid errors.
-GIT_TEST_MAINT_SCHEDULER="none:exit 1"
+# Tests that verify the scheduler integration must set and
+# export this variable locally.
+sane_unset GIT_TEST_MAINT_SCHEDULER

Clearly the idea was to protect the scheduler-configuration of the person running the test in the event that the test author forgot to set GIT_TEST_MAINT_SCHEDULER to one of the legitimate "testing values" before invoking a "destructive" command, such as git maintenance start. By defaulting to none:exit 1, the problem would be caught and reported before any damage could be done to the configuration of the person running the tests.

So, I'm somewhat skeptical of the new direction of simply unsetting GIT_TEST_MAINT_SCHEDULER since that outright removes the intended protection. I'd have expected this problem to be addressed by exporting GIT_TEST_MAINT_SCHEDULER, not by making it easier for an absent-minded test author to break his or her own configuration.

Having said that, it you do want to go the route of eliminating the (intended) protection altogether, I have a couple additional observations:

First, this change requires a corresponding update to the lead-in comment ("Ensure that no test accidentally triggers a Git command that runs the actual maintenance scheduler, affecting a user's system permanently.") since it renders that comment incorrect.

Second, it seems very unlikely that GIT_TEST_MAINT_SCHEDULER will be set in the user's environment anyhow before running the tests, so unsetting it here seems unnecessary and pointless. Instead, a cleaner approach would be to simply remove the entire hunk in t/test-lib.sh dealing with GIT_TEST_MAINT_SCHEDULER, including the comment.

#2 lore raw
Quoted 7 lines · Eric Sunshine

Clearly the idea was to protect the scheduler-configuration of the person running the test in the event that the test author forgot to set GIT_TEST_MAINT_SCHEDULER to one of the legitimate "testing values" before invoking a "destructive" command, such as git maintenance start. By defaulting to none:exit 1, the problem would be caught and reported before any damage could be done to the configuration of the person running the tests.

Yeah, if the variable were exported. So I am OK with a fix in the opposite direction.

#3 lore raw

PATCH 1/1 test-lib: fix non-functioning GIT_TEST_MAINT_SCHEDULER fallback

In reply to Junio C Hamano

From: Eric Sunshine <[email protected]>

When environment variable GIT_TEST_MAINT_SCHEDULER is set, git maintenance invokes the command specified as the variable's value rather than invoking the actual underlying platform-specific scheduler management command. By setting GIT_TEST_MAINT_SCHEDULER to some suitable value, test authors can therefore validate behavior of "destructive" git maintenance commands without having to worry about clobbering the user's own local scheduler configuration.

In order to protect an absent-minded test author from forgetting to set GIT_TEST_MAINT_SCHEDULER in the local test script (and thus clobbering his or her own scheduler configuration), t/test-lib.sh assigns an "immediately error-out" value to GIT_TEST_MAINT_SCHEDULER by default which should ensure that the problem will be caught and reported before any damage can be done to the configuration of the person running the tests.

Unfortunately, however, t/test-lib.sh neglects to export GIT_TEST_MAINT_SCHEDULER, which renders the default "error-out" assignment worthles. Fix this by exporting the variable as originally intended.

---

This is a replacement for Junio's [1]. That attempt made it easier for a test author to shoot him or herself in the foot. This replacement patch instead fixes the foot-shooting guard.

[1]: https://lore.kernel.org/git/[email protected]/

 t/test-lib.sh | 1 +
 1 file changed, 1 insertion(+)
t/test-lib.sh +1 -0
index c8af8dab79..79d3e0e7d9 100644
@@ -1962,6 +1962,7 @@ test_lazy_prereq DEFAULT_REPO_FORMAT '
 # Tests that verify the scheduler integration must set this locally
 # to avoid errors.
 GIT_TEST_MAINT_SCHEDULER="none:exit 1"
+export GIT_TEST_MAINT_SCHEDULER
 
 # Does this platform support `git fsmonitor--daemon`
 #
Signature · 1 line
2.44.0
#4 lore raw

Re: [PATCH] test-lib: fix non-functioning GIT_TEST_MAINT_SCHEDULER fallback

Quoted 16 lines · Eric Sunshine

When environment variable GIT_TEST_MAINT_SCHEDULER is set, git maintenance invokes the command specified as the variable's value rather than invoking the actual underlying platform-specific scheduler management command. By setting GIT_TEST_MAINT_SCHEDULER to some suitable value, test authors can therefore validate behavior of "destructive" git maintenance commands without having to worry about clobbering the user's own local scheduler configuration.

In order to protect an absent-minded test author from forgetting to set GIT_TEST_MAINT_SCHEDULER in the local test script (and thus clobbering his or her own scheduler configuration), t/test-lib.sh assigns an "immediately error-out" value to GIT_TEST_MAINT_SCHEDULER by default which should ensure that the problem will be caught and reported before any damage can be done to the configuration of the person running the tests.

Quoted 4 lines

Unfortunately, however, t/test-lib.sh neglects to export GIT_TEST_MAINT_SCHEDULER, which renders the default "error-out" assignment worthles. Fix this by exporting the variable as originally intended.

s/worthles/worthless/

(I won't reroll just for this minor typo.)