Bug 225660 – [BMalloc] Temporarily disable bmalloc in debug builds while examining 76167814

• Component: bmalloc
• Status: RESOLVED WONTFIX

[Dean Jackson]

[BMalloc] Temporarily disable bmalloc in debug builds while examining 76167814

[Radar WebKit Bug Importer]

<rdar://problem/77852400>

[Geoffrey Garen]

Kinda sketchy to lose debug testing coverage for bmalloc and gigacage. Maybe release coverage is sufficient? Maybe.

[Dean Jackson]

Created attachment 428292 [details]
Patch

[Dean Jackson]

(In reply to Geoffrey Garen from comment #2)
> Kinda sketchy to lose debug testing coverage for bmalloc and gigacage. Maybe
> release coverage is sufficient? Maybe.

The original bug hasn't been touched in weeks, which is more sketchy :)

[Dean Jackson]

(In reply to Dean Jackson from comment #4)
> (In reply to Geoffrey Garen from comment #2)
> > Kinda sketchy to lose debug testing coverage for bmalloc and gigacage. Maybe
> > release coverage is sufficient? Maybe.
> 
> The original bug hasn't been touched in weeks, which is more sketchy :)

Sorry, I meant months.

[Chris Dumez]

Comment on attachment 428292 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=428292&action=review

> Source/bmalloc/ChangeLog:3
> +        [BMalloc] Temporarily disable bmalloc in debug builds while examining 76167814

Why can't people use the environment variables to disable bmalloc when debugging? Disabling BMalloc for everybody in debug seems aggressive.

[Dean Jackson]

(In reply to Chris Dumez from comment #6)
> Comment on attachment 428292 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=428292&action=review
> 
> > Source/bmalloc/ChangeLog:3
> > +        [BMalloc] Temporarily disable bmalloc in debug builds while examining 76167814
> 
> Why can't people use the environment variables to disable bmalloc when
> debugging? Disabling BMalloc for everybody in debug seems aggressive.

Because it is difficult to do this for the simulator, and people will have to remember to undo the workaround. Seems better to track this officially.

Well, the best would be to address the real bug, which is almost certainly in the toolchain :)

[Chris Dumez]

We're loosing useful test coverage on bots and local machines just for the purpose of making the debugger work. It doesn't seem like a good trade-offs. We want to be running something as close as possible to the configuration we ship.

[Dean Jackson]

(In reply to Chris Dumez from comment #8)
> We're loosing useful test coverage on bots and local machines just for the
> purpose of making the debugger work. It doesn't seem like a good trade-offs.
> We want to be running something as close as possible to the configuration we
> ship.

We don't ship debug builds.

But to be clear, I don't want to do this. I want the actual bug to be examined - it's been more than a month and nothing has happened.

I should remove the r? flag because I don't expect anyone will want to land this, but it serves as a patch that people can apply to get things working locally.