Only capture logcat on failure

[aurimasc]

There have been multiple instances lately where capturing logcat gets stuck.
As a mitigation, capture logcat only if execution fails. If all tests are successful, skip it.

[u-pr]

💡 Harness Review

The change gates Android log collection on Yamato’s command-block result while preserving the existing log destination and artifact collection behavior; I examined the batch control flow, invocation context, and failure paths.

_{Reviewed commit fbc5476}

_{🤖 Helpful? 👍/👎}

[todi1856]

Honestly I don't think it's good idea. In case of failure, you sometimes want to compare failed logcat against succesful logcat, thus is important to have logcat even in succesful builds too.

[todi1856]

p.s did you consider clearing logcat before test run, that would make the log smaller, and might be the reason why it sometimes gets stuck

[aurimasc]

Honestly I don't think it's good idea. In case of failure, you sometimes want to compare failed logcat against succesful logcat, thus is important to have logcat even in succesful builds too.

The problem is that the fail rate is unacceptably high. Something like every other PR has at least one failure due to this.
During last 10 days I found 4 failures for this reason, none of the PRs had any Android changes:
https://yamato.ds.unity3d.com/job/75460576/logs/execution
https://yamato.ds.unity3d.com/job/75511999/logs/execution
https://yamato.ds.unity3d.com/job/75029231/logs/execution
https://yamato.ds.unity3d.com/job/74990540/logs/execution

[todi1856]

I cannot stop you from pushing such change, but in my opinion this is wrong, at the very least try clearing the logcat before test run, maybe it's simple as that for fixing this