RE: [Test Failures] Git v2.37.1 (was RE: [ANNOUNCE] Git v2.37.1 and others)

[rsbecker]

On July 13, 2022 10:24 AM, I wrote:
>On July 13, 2022 8:11 AM, I wrote:
>>On July 12, 2022 1:07 PM, Junio C Hamano wrote:
>>>Git v2.37.1, together with v2.30.5, v2.31.4, v2.32.3, v2.33.4,
>>>v2.34.4, v2.35.4, and
>>>v2.36.2 for older maintenance tracks, are now available at the usual places.
>>>
>>>These are to address CVE-2022-29187, where the fixes in v2.36.1 and
>>>below to address CVE-2022-24765 released earlier may not have been
>complete.
>>
>>Following are net new test failures with 2.37.1 compared with 2.37.0
>>are as follows on NonStop ia64 and x86 platforms:
>>
>>t5516-fetch-push subtests 53, 113
>>
>>t5545-push-options subtest 9
>
>Test passes when not run in Jenkins CI/CD
>
>>t5601-clone subtest 8
>
>Test passes when not run in Jenkins CI/CD
>
>>t7502-commit-porcelain subtests 20-26, 29-33, 42-47
>
>Test passes when not run in Jenkins CI/CD
>
>>t7528-signed-commit-ssh subtests 4, 5
>
>Test passes when not run in Jenkins CI/CD
>
>So it looks like a Jenkins/git test interaction is the issue here. I'm wondering
>whether running the test with </dev/null and 2>/dev/null might make a
>difference when running in Jenkins. The tty where Jenkins was started is in a
>permanently disconnected state.
>
>Mostly ignore the subtest failures, but still, would like to know why these subtests
>in particular.

Strangely, setting stdin to /dev/null and stderr to /dev/null makes no difference in the test results. stdout is the Jenkins pipe, which is Java of course, but roughly a popen(x, "r"). I have no explanation at this point why specific tests fail - other than OpenSSH, which I know attempts to prompt if it thinks the tty is even remotely reasonable.