Bug 14707 – Crashes in StringImpl::length() seen on build bot after r24492

• Status: RESOLVED
• Resolution: FIXED
• Component: Page Loading

The leaks build slaves have been crashing repeatedly following <http://trac.webkit.org/projects/webkit/changeset/24492>. The crash happens under DocLoader::removeCachedResource().

[Geoffrey Garen]

Mitz,

I see crashes like the following:

0   <<00000000>> 	0xffff8264 __spin_lock + 4 (cpu_capabilities.h:179)
1   libSystem.B.dylib        	0x90114010 stack_logging_log_stack + 608
2   libSystem.B.dylib        	0x90003558 malloc + 688
3   com.apple.JavaScriptCore 	0x004879ec WTF::fastMalloc(unsigned long) + 104 (FastMalloc.cpp:142)
4   com.apple.WebCore        	0x011ff0e0 WebCore::newUCharVector(unsigned) + 40 (StringImpl.cpp:59)
5   com.apple.WebCore        	0x011ff4c4 WebCore::StringImpl::init(unsigned short const*, unsigned) + 132 (StringImpl.cpp:125)

Is that what you're talking about? If so, this is a known bug in Tiger malloc / leaks. We've seen it before on the leaks bot, and it's not related to r24492.

[Geoffrey Garen]

Wait a sec. I was looking at the wrong part of the crash log. Here's the crash Mitz was talking about:

0   com.apple.WebCore              	0x015eef44 WebCore::StringImpl::length() const + 20 (StringImpl.h:72)
1   com.apple.WebCore              	0x015efb84 WTF::StrHash<WebCore::StringImpl*>::equal(WebCore::StringImpl const*, WebCore::StringImpl const*) + 100 (StringHash.h:42)
2   com.apple.WebCore              	0x015f11b4 WTF::IdentityHashTranslator<WebCore::StringImpl*, std::pair<WebCore::StringImpl*, int>, WTF::StrHash<WebCore::StringImpl*> >::equal(WebCore::StringImpl* const&, WebCore::StringImpl* const&) + 56 (HashTable.h:257)

[Geoffrey Garen]

<rdar://problem/5352570>

[Geoffrey Garen]

7/22/07 12:31 PM Anders Carlsson:
Darin checked in a fix for this, revision 24514 as a followup fix to rdar://problem/5298816.  

We should verify with TOT.