Bug 49212 – WebKitPluginHost is losing its connection to fontd

• Component: Tools / Tests
• Status: NEW

[Eric Seidel (no email)]

Created attachment 73291 [details]
crash log

WebKitPluginHost is losing its connection to fontd

See Radar 8630760.

Radar 8630760 is about general ATS-related crashes for all apps on the system after running run-webkit-tests enough times on Snow Leopard.

This bug is specifically about WebKitPluginHost crashing, as it seems to be the most common process to crash, and may be what is causing fontd to go down for all apps (which is the current working theory in the Radar).

This may be related to quicktime?   Maybe this should move to Radar?  Please advise.

[Eric Seidel (no email)]

Created attachment 73292 [details]
Similar QtKitServer crash log.

[Eric Seidel (no email)]

Thread 0 Crashed:  Dispatch queue: com.apple.main-thread
0   ...ple.ApplicationServices.ATS	0x94a6b060 _eFOGetFontFamilyTextEncoding + 89
1   ...ple.ApplicationServices.ATS	0x94aca239 FOGetFontFamilyTextEncoding + 80
2   com.apple.HIToolbox           	0x95e1d6e7 GetFontScriptHL + 36
3   ...ple.CoreServices.CarbonCore	0x92d9224a GetFontScript + 73
4   ...ple.CoreServices.CarbonCore	0x92d9238b _IntlScript + 56
5   ...ple.CoreServices.CarbonCore	0x92d923f7 IntlScript + 11
6   com.apple.QuickTime           	0x92139017 PrivateGetMoviePublicMovieComplete_priv + 324

Thread 0 Crashed:  Dispatch queue: com.apple.main-thread
0   ...ple.ApplicationServices.ATS	0x94a6b060 _eFOGetFontFamilyTextEncoding + 89
1   ...ple.ApplicationServices.ATS	0x94aca239 FOGetFontFamilyTextEncoding + 80
2   com.apple.HIToolbox           	0x95e1d6e7 GetFontScriptHL + 36
3   ...ple.CoreServices.CarbonCore	0x92d9224a GetFontScript + 73
4   ...ple.CoreServices.CarbonCore	0x92d9238b _IntlScript + 56
5   ...ple.CoreServices.CarbonCore	0x92d923f7 IntlScript + 11
6   com.apple.QuickTime           	0x92139017 PrivateGetMoviePublicMovieComplete_priv + 324
7   com.apple.QuickTime           	0x92128903 NewMovieFromUserProcPriv + 693
8   com.apple.QuickTime           	0x9220bee8 NewMovieFromUserProc_priv + 56
9   com.apple.QuickTime           	0x9236be59 NewMovieFromUserProc + 66
10  ...le.QuickTimeMPEG4.component	0x985dab9f EatMPEG4DataRef + 2581
11  ...ple.CoreServices.CarbonCore	0x92d31237 callComponentStorage_4444444444 + 84
12  ...ple.CoreServices.CarbonCore	0x92d2556c CallComponentFunctionCommonWithStorage(char**, ComponentParameters*, long (*)(), unsigned long) + 54
13  ...le.QuickTimeMPEG4.component	0x985da10b EatMPEG4ComponentDispatch + 113
14  ...ple.CoreServices.CarbonCore	0x92d1dce5 CallComponentDispatch + 29
15  com.apple.QuickTime           	0x9237c11c MovieImportDataRef + 91
16  com.apple.QuickTime           	0x9220cf51 newMovieFromDataRefFromComponent + 655
17  com.apple.QuickTime           	0x92126132 NewMovieFromDataRefPriv_priv + 11366
18  com.apple.QuickTime           	0x92122c7f NewMovieFromProperties_priv + 6213
19  ...ple.QuickTime Plugin.plugin	0x0051e20f MyNewMovieFromProperties + 733
20  ...ple.QuickTime Plugin.plugin	0x0051e3a9 instantiateMovieWithQTURLDataHandler + 334
21  ...ple.QuickTime Plugin.plugin	0x0052ba0a LoadMovieURL + 1875
22  ...ple.QuickTime Plugin.plugin	0x005122b9 oop_loadFirstMovieURL + 198
23  ...ple.QuickTime Plugin.plugin	0x00509d66 NPP_GetValue + 91
24  ...ple.QuickTime Plugin.plugin	0x0051105f WebHelper_Private_GetValue + 61
25  com.apple.WebKit.PluginHost   	0x6a67b872 0x6a673000 + 34930

[David Kilzer (:ddkilzer)]

<rdar://problem/8643916>

[David Kilzer (:ddkilzer)]

(In reply to comment #0)
> This bug is specifically about WebKitPluginHost crashing, as it seems to be the most common process to crash, and may be what is causing fontd to go down for all apps (which is the current working theory in the Radar).
> 
> This may be related to quicktime?   Maybe this should move to Radar?  Please advise.

I imported it into Radar for tracking purposes and related it to <rdar://problem/8630760>.

[Eric Seidel (no email)]

My current suspicion is that these crashing may be triggered by us feeding invalid font data to CG during one of the layout tests.  That's just a theory.  I certainly know we have @font-face tests.  Some of which may register all sorts of nasty fonts.

It's also possible that CG or fontd is leaking in some form and that eventually enough @font-faces causes it to get into a bad state.

The full system death happened again this evening and I updated radar 8630760.