[webkitscmpy] Handle failed `git log` process

• Component: Tools / Tests
• Status: RESOLVED FIXED

[Aakash Jain]

Incorrect commit identifier on igalia-jsc32-armv7-ews (https://ews-build.webkit.org/#/workers/81). Earlier this bot was having issues with git (e.g.: https://ews-build.webkit.org/#/builders/38/builds/31141). After the issue was fixed, the commit identifiers incorrectly started from 2@main. e.g.: https://ews-build.webkit.org/#/builders/38/builds/31142, https://ews-build.webkit.org/#/builders/38/builds/31234

[Jonathan Bedard]

This is fallout from https://bugs.webkit.org/show_bug.cgi?id=225616, we have a bad cache. This cache is trivial to clear (it's in .git/identifiers.json), but I don't know how to access this bot.

I'll need to explore a bit to understand how we got a bad cache, because that's clearly a bug. But now that we have a bad one, this behavior is not surprising.

[Angelos Oikonomopoulos]

(In reply to Jonathan Bedard from comment #1)
> This is fallout from https://bugs.webkit.org/show_bug.cgi?id=225616, we have
> a bad cache. This cache is trivial to clear (it's in .git/identifiers.json),
> but I don't know how to access this bot.

Thanks for the tip, I just deleted .git/identifiers.json in igalia-jsc32-armv7-ews-02.

> I'll need to explore a bit to understand how we got a bad cache, because
> that's clearly a bug. But now that we have a bad one, this behavior is not
> surprising.

FWIW the issue that Aakash linked to above caused git to exit early for lots of operations, so perhaps this broke some assumption about a git operation that "cannot possibly fail" in the code that manages the identifier cache?

[Jonathan Bedard]

(In reply to Angelos Oikonomopoulos from comment #2)
> (In reply to Jonathan Bedard from comment #1)
> > This is fallout from https://bugs.webkit.org/show_bug.cgi?id=225616, we have
> > a bad cache. This cache is trivial to clear (it's in .git/identifiers.json),
> > but I don't know how to access this bot.
> 
> Thanks for the tip, I just deleted .git/identifiers.json in
> igalia-jsc32-armv7-ews-02.
> 
> > I'll need to explore a bit to understand how we got a bad cache, because
> > that's clearly a bug. But now that we have a bad one, this behavior is not
> > surprising.
> 
> FWIW the issue that Aakash linked to above caused git to exit early for lots
> of operations, so perhaps this broke some assumption about a git operation
> that "cannot possibly fail" in the code that manages the identifier cache?

Almost certainly `git log`, I guess I need to start checking exit codes for that process

[Radar WebKit Bug Importer]

<rdar://problem/80520002>

[Jonathan Bedard]

Created attachment 433406 [details]
Patch

[Jonathan Bedard]

Created attachment 433422 [details]
Patch

[Aakash Jain]

Comment on attachment 433422 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=433422&action=review

rs=me

> Tools/Scripts/libraries/webkitscmpy/webkitscmpy/local/git.py:166
> +                    return

what is the consequence of returning here?

[Jonathan Bedard]

(In reply to Aakash Jain from comment #7)
> Comment on attachment 433422 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=433422&action=review
> 
> rs=me
> 
> > Tools/Scripts/libraries/webkitscmpy/webkitscmpy/local/git.py:166
> > +                    return
> 
> what is the consequence of returning here?

By returning here, we basically throw out the whole result of `git log`, since we don't cache it until after we finish the commands (because we essentially need to cache it backwards). If we don't have a cache, we compute things the way we did before the cache.

[Aakash Jain]

(In reply to Jonathan Bedard from comment #8)
> By returning here, we basically throw out the whole result of `git log`,
> since we don't cache it until after we finish the commands (because we
> essentially need to cache it backwards). If we don't have a cache, we
> compute things the way we did before the cache.
ok, sounds good.

[Jonathan Bedard]

Created attachment 433452 [details]
Patch for landing

[EWS]

Committed r279898 (239648@main): <https://commits.webkit.org/239648@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 433452 [details].