git: fix riscv64-linux build

[liberodark]

fix git issue :

git-riscv64-linux> Test Summary Report
git-riscv64-linux> -------------------
git-riscv64-linux> t0213-trace2-ancestry.sh                         (Wstat: 256 (exited 1) Tests: 5 Failed: 4)
git-riscv64-linux>   Failed tests:  2-5
git-riscv64-linux>   Non-zero exit status: 1
git-riscv64-linux> Files=1028, Tests=29512, 1604 wallclock secs (41.52 usr  2.33 sys + 15862.29 cusr 1850.84 csys = 17756.98 CPU)
git-riscv64-linux> Result: FAIL
git-riscv64-linux> make[1]: *** [Makefile:78: prove] Error 1
git-riscv64-linux> make[1]: Leaving directory '/build/git-2.54.0/t'
git-riscv64-linux> make: *** [Makefile:3391: test] Error 2

Things done

• Built on platform:
  • x86_64-linux
  • aarch64-linux
  • aarch64-darwin
  • riscv64-linux
• Tested, as applicable:
  • NixOS tests in nixos/tests.
  • Package tests at passthru.tests.
  • Tests in lib/tests or pkgs/test for functions and "core" functionality.
• Ran nixpkgs-review on this PR. See nixpkgs-review usage.
• Tested basic functionality of all binary files, usually in ./result/bin/.
• Nixpkgs Release Notes
  • Package update: when the change is major or breaking.
• NixOS Release Notes
  • Module addition: when adding a new NixOS module.
  • Module update: when the change is significant.
• Fits CONTRIBUTING.md, pkgs/README.md, maintainers/README.md and other READMEs.
• Follows the automation/AI policy.

[philiptaron]

Can you do a little more root-causing on why this is? Is there an upstream issue?

[liberodark]

Can you do a little more root-causing on why this is? Is there an upstream issue?

Hi,

I'm actually looking into this right now; I'm not sure the issue lies with upstream.
But I hope Tristan will also have time to provide some feedback on his end.
I've observed this behavior with qemu.
Unfortunately, it takes several hours to attempt to reproduce the issue.
I have reproduced it four times using qemu.
I am currently working with riscv64 hardware.

Best Regards

[me-and]

Yeah, I'm not comfortable disabling tests just because they fail. Those tests are there for a reason, and we should understand what the failure means or implies before disabling it.

[liberodark]

Yeah, I'm not comfortable disabling tests just because they fail. Those tests are there for a reason, and we should understand what the failure means or implies before disabling it.

Hi,

Me too.
That is why I am waiting for the tests to finish and for further feedback before trying to reproduce it.
I can move it to draft if you like.
PS : To be precise, it’s not the entire test that’s failing.
But I’m still waiting for some feedback sorry, it’s taking a very, very long time.

Best Regards

[philiptaron]

But I’m still waiting for some feedback sorry, it’s taking a very, very long time.

Yeah, this is the RISC-V sadness. Hopefully the newer chips are faster and more effective.

I moved it to draft; please ping me if you get something definitive.

[JamieMagee]

I think it's a QEMU emulation artifact, not a real riscv64 failure.

On native hardware it passes:

t0213-trace2-ancestry.sh ... ok
Result: PASS

The test checks git's trace2 cmd_ancestry event, which git builds by walking the process tree and reading each process's comm from /proc/<pid>/stat. Tests 2-5 truncate the ancestry at the literal string test-tool and compare against test-tool/git test-tool.

[liberodark]

I think it's a QEMU emulation artifact, not a real riscv64 failure.

On native hardware it passes:

t0213-trace2-ancestry.sh ... ok
Result: PASS

The test checks git's trace2 cmd_ancestry event, which git builds by walking the process tree and reading each process's comm from /proc/<pid>/stat. Tests 2-5 truncate the ancestry at the literal string test-tool and compare against test-tool/git test-tool.

Hi,

I think too yes.
Thank for your test.

Best Regards

[JamieMagee]

I've made an attempt to fix this upstream: gitgitgadget/git#2168