Bug 226506 – [webkitcorepy] TaskPool shouldn't fork when 1 process is needed

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[Jonathan Bedard]

If we only need a single process, the TaskPool object should not fork. This is important for certain integration tests which use mocking structures that don't function across process boundaries.

[Radar WebKit Bug Importer]

<rdar://problem/78724554>

[Jonathan Bedard]

Created attachment 430279 [details]
Patch

[Jonathan Bedard]

Created attachment 430409 [details]
Patch

Comment on attachment 430409 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=430409&action=review

r=me with comment.

> Tools/Scripts/libraries/webkitcorepy/webkitcorepy/task_pool.py:373
>          from mock import patch

Should we ensure we always import this or it's not needed in the case of no-fork?

[Jonathan Bedard]

(In reply to dewei_zhu from comment #4)
> Comment on attachment 430409 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=430409&action=review
> 
> r=me with comment.
> 
> > Tools/Scripts/libraries/webkitcorepy/webkitcorepy/task_pool.py:373
> >          from mock import patch
> 
> Should we ensure we always import this or it's not needed in the case of
> no-fork?

Not needed in the case of no-fork.

The reason for including it in the fork case, for those who are curious, is that autoinstalling a package on a child process is racey. This is something we should consider resolving with a file lock (because independently running Python processes have a similar problem), but it hasn't been much of a problem in practice, so we haven't done anything to fix it.

[EWS]

Committed r278380 (238407@main): <https://commits.webkit.org/238407@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 430409 [details].