Bug 15178 – REGRESSION: Unreproducible ASSERT removing icon from Icon Database

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

[David Kilzer (:ddkilzer)]

* SUMMARY
I've gotten an assertion failure twice so far in a debug build of WebKit r25488 with Safari 3 Public Beta v. 3.0.3 (522.12.1) on Mac OS X 10.4.10 (8R218).  Haven't figured out how to reproduce it yet.  Happens when I click on a link (apparently at the wrong time).

* CONSOLE

ASSERTION FAILED: iconID
(/path/to/WebKit/WebCore/loader/icon/IconDatabase.cpp:1898 void WebCore::IconDatabase::removeIconFromSQLDatabase(const WebCore::String&))
Segmentation fault

* CRASH LOG

Exception:  EXC_BAD_ACCESS (0x0001)
Codes:      KERN_INVALID_ADDRESS (0x0001) at 0xbbadbeef

Thread 0:
0   libSystem.B.dylib              	0x9000b348 mach_msg_trap + 8
1   libSystem.B.dylib              	0x9000b29c mach_msg + 60
2   com.apple.CoreFoundation       	0x907ddba8 __CFRunLoopRun + 832
3   com.apple.CoreFoundation       	0x907dd4ac CFRunLoopRunSpecific + 268
4   com.apple.HIToolbox            	0x9329bb20 RunCurrentEventLoopInMode + 264
5   com.apple.HIToolbox            	0x9329b1b4 ReceiveNextEventCommon + 380
6   com.apple.HIToolbox            	0x9329b020 BlockUntilNextEventMatchingListInMode + 96
7   com.apple.AppKit               	0x937a1ae4 _DPSNextEvent + 384
8   com.apple.AppKit               	0x937a17a8 -[NSApplication nextEventMatchingMask:untilDate:inMode:dequeue:] + 116
9   com.apple.Safari               	0x00006770 0x1000 + 22384
10  com.apple.AppKit               	0x9379dcec -[NSApplication run] + 472
11  com.apple.AppKit               	0x9388e87c NSApplicationMain + 452
12  com.apple.Safari               	0x0000244c 0x1000 + 5196
13  com.apple.Safari               	0x0004f1b0 0x1000 + 319920

Thread 1 Crashed:
0   com.apple.WebCore              	0x01383438 WebCore::IconDatabase::removeIconFromSQLDatabase(WebCore::String const&) + 248 (IconDatabase.cpp:1898)
1   com.apple.WebCore              	0x01383a0c WebCore::IconDatabase::writeIconSnapshotToSQLDatabase(WebCore::IconSnapshot const&) + 360 (IconDatabase.cpp:1938)
2   com.apple.WebCore              	0x01386bf0 WebCore::IconDatabase::writeToDatabase() + 740 (IconDatabase.cpp:1478)
3   com.apple.WebCore              	0x01388a00 WebCore::IconDatabase::syncThreadMainLoop() + 296 (IconDatabase.cpp:1285)
4   com.apple.WebCore              	0x0138a708 WebCore::IconDatabase::iconDatabaseSyncThread() + 1596 (IconDatabase.cpp:975)
5   com.apple.WebCore              	0x0138a760 WebCore::IconDatabase::iconDatabaseSyncThreadStart(void*) + 40 (IconDatabase.cpp:881)
6   libSystem.B.dylib              	0x9002bd08 _pthread_body + 96

[David Kilzer (:ddkilzer)]

Created attachment 16256 [details]
Two crash logs

[David Kilzer (:ddkilzer)]

<rdar://problem/5474001>

[Mark Rowe (bdash)]

I think we may need some more uppercase letters in the title.

[David Kilzer (:ddkilzer)]

(In reply to comment #3)
> I think we may need some more uppercase letters in the title.

SORRY.  :)

[Brady Eidson]

Yes, the title drove me nutso enough to change it.

Dave, were you running old Safari side-by-side with ToT WebKit?  That case is known to have problems because the old WebKit overwrites the DB schema causing ToT to fail in it's DB operations.

If that is not the case, we need something more reproducible, I fear...  =/

[David Kilzer (:ddkilzer)]

(In reply to comment #5)
> Dave, were you running old Safari side-by-side with ToT WebKit?  That case is
> known to have problems because the old WebKit overwrites the DB schema causing
> ToT to fail in it's DB operations.

Define "old Safari".  Do you mean a 2.0.x-vintage Safari or a 3.0.x-beta-vintage Safari?  Was there not enough config info in Comment #0?

> I've gotten an assertion failure twice so far in a debug build of WebKit r25488
> with Safari 3 Public Beta v. 3.0.3 (522.12.1) on Mac OS X 10.4.10 (8R218). 

[David Kilzer (:ddkilzer)]

(In reply to comment #5)
> If that is not the case, we need something more reproducible, I fear...  =/

Is Bug 15179 related?

[Brady Eidson]

What I meant was were you running Safari with ToT WebKit at the SAME TIME as a "stock" Safari - in this case Safari 3.0.3 Beta?

Two Safari's at once.

Is what I meant.

:)

[David Kilzer (:ddkilzer)]

(In reply to comment #8)
> What I meant was were you running Safari with ToT WebKit at the SAME TIME as a
> "stock" Safari - in this case Safari 3.0.3 Beta?
> 
> Two Safari's at once. Is what I meant. :)

Oh, yes, I may have started up Safari 2.0.4 to test something at the same time.  Will close for now and reopen if I see this again without launching/quitting Safari 2.0.4 at the same time.

[David Kilzer (:ddkilzer)]

I had another instance of this crash happen again.  Still don't know what the trigger point is.  Will bad things happen in a release build when iconID is null?

[David Kilzer (:ddkilzer)]

Created attachment 16263 [details]
Third crash log

[David Kilzer (:ddkilzer)]

(In reply to comment #10)
> I had another instance of this crash happen again.  Still don't know what the
> trigger point is.  Will bad things happen in a release build when iconID is
> null?

No, the current code will LOG() and return early.  It would be nice if I could enable some kind of icon database consistency checking in a debug build to catch the problem earlier.

[David Kilzer (:ddkilzer)]

I believe this happens if scrolling occurs (mouse wheel only?) at the "wrong" time after a page load.

I haven't figured out how to reproduce it reliably, though.

[David Kilzer (:ddkilzer)]

Fixed by Brady in r25604.