Bug 21380
| Summary: | many crashes running layout tests under GuardMalloc | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Darin Fisher (:fishd, Google) <fishd> |
| Component: | WebKit Misc. | Assignee: | Nobody <webkit-unassigned> |
| Status: | RESOLVED INVALID | ||
| Severity: | Normal | CC: | mrowe |
| Priority: | P2 | ||
| Version: | 528+ (Nightly build) | ||
| Hardware: | Mac (Intel) | ||
| OS: | OS X 10.5 | ||
Darin Fisher (:fishd, Google)
many crashes running layout tests under GuardMalloc
$ ./WebKitTools/Scripts/run-webkit-tests -g -c debug fast/dom
fast/dom/DOMParser-assign-variable.html -> crashed
fast/dom/ImageDocument-image-deletion.html -> crashed
fast/dom/NamedNodeMap-setNamedItem-crash.html -> crashed
fast/dom/XMLSerializer-doctype2.html -> crashed
...
In fact, it would appear that they are all crashing.
I have not been successful at running DumpRenderTree under GDB when GuardMalloc is enabled. DRT hits an EXC_BAD_ACCESS signal really early on (too early on to correspond to the crash observed while not running under GDB).
Under GDB w/ GuardMalloc enabled, I get this backtrace:
#0 0x936d5a21 in _LSCopyInfoForNode ()
#1 0x936f28cb in LSCopyItemInfoForURL ()
#2 0x951a9590 in CGSNewConnection ()
#3 0x951a558f in _CGSDefaultConnection ()
#4 0x90f3c48f in _RegisterApplication ()
#5 0x90f3bc5a in GetCurrentProcess ()
#6 0x96014750 in GetSystemUIMode ()
#7 0x960146ff in IsMenuBarVisible ()
#8 0x92d69271 in _NSInitializeAppContext ()
#9 0x92d68bfa in -[NSApplication init] ()
#10 0x92d6860b in +[NSApplication sharedApplication] ()
#11 0x00006582 in main (argc=3, argv=0xbffff878) at /Users/darin/webkit-trunk/We
bKit/WebKitTools/DumpRenderTree/mac/DumpRenderTree.mm:535
So, I think there is just some problem with running GuardMalloc under GDB that makes it hard for me to determine why 'run-webkit-tests -g' is crashing all the time.
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Mark Rowe (bdash)
You're being bitten by a bug that was briefly introduced to LaunchServices in an early Leopard software update that was only visible under guard malloc. It was fixed in either 10.5.4 or 10.5.5, so please do a software update.
Closing as invalid as this is not a WebKit bug.