Workflow enhancement to handle Git failures

[KapilPowar]

runtimes/backlog/issues/1184

Update test workflow to handle Git failures.

• fixes issue of corrupted git directory resulting in test failure

Signed-off-by: Kapil Powar [email protected]

CC: @llxia

[KapilPowar]

Here are test runs for the changes
Grinder #1635
Grinder #1636

[llxia]

LGTM

[llxia]

FYI @AdamBrousseau @pshipton

[AdamBrousseau]

I would recommend a better commit comment. You have linked to an internal IBM issue which nobody outside of IBM will be able to see. Your commit comment should explain what you're changing and why. This repo has a contributing.md (here). I try to follow, and by default I direct people to OpenJ9's guidelines on commit comments. You will thank yourself later when you're scrolling through the history trying to track down a change.

[AdamBrousseau]

Will this only handle the issue where the .git dir is missing? I was originally thinking it would cover all cases where the git update fails.

[llxia]

This PR will check the first git cmd, that is git rev-parse HEAD. If this cmd does not run successfully (regardless of the error), the program will delete the test dir and re-clone. This should cover most of the cases that we have seen so far.
However, it will not catch the case where it fails at later git cmd.

I think we should get this change in to ease the problems. Meanwhile, working on a solution to handle all git cases. To be honest, I think we should re-write the logic in this file.

btw, If the material is in a bad state (i.e., jenkins user cannot delete the material), there is not much we can do.

[LongyuZhang]

LGTM