Bug 128406 – Fix bug in stack limit adjustments in JSLock.

• Status: RESOLVED
• Resolution: FIXED
• Component: JavaScriptCore

[Mark Lam]

This is the real fix for https://bugs.webkit.org/show_bug.cgi?id=128347 and will undo the workaround committed there.

[Mark Lam]

Created attachment 223505 [details]
the patch.

[Geoffrey Garen]

Comment on attachment 223505 [details]
the patch.

Let's turn this into three patches:

(1) A three-line patch that removes the problematic assignment of nullptr even if not dropping the lock, and removes the problematic testing of "m_vm->stackPointerAtVMEntry == entryStackPointer".

(2) A mechanical patch that re-enables stack checking.

(3) The re-architectural you've proposed here, which removes the iOS-specific code path.

For (3), I think we need a little research into why that path currently exists, and is iOS-specific. I believe it was needed to fix some nasty bugs -- so we shouldn't just remove it without knowledge of why it's there.

[Mark Lam]

Created attachment 223509 [details]
reduced patch with only the JSLock fix.

[Geoffrey Garen]

Comment on attachment 223509 [details]
reduced patch with only the JSLock fix.

r=me

[Mark Lam]

Thanks.  Landed in r163661: <http://trac.webkit.org/r163661>.

[Michael Saboff]

Comment on attachment 223509 [details]
reduced patch with only the JSLock fix.

This patch should have deleted entryStackPointer since it is no longer needed.

[Mark Lam]

(In reply to comment #6)
> (From update of attachment 223509 [details])
> This patch should have deleted entryStackPointer since it is no longer needed.

Took care of that in https://webkit.org/b/128413, fixed in <http://trac.webkit.org/r163665>.

[Mark Lam]

ref: <rdar://problem/16008492>