* TITLE/SUMMARY [PP] Viewer suicides after 10 seconds if URL text left selected * STEPS TO REPRODUCE 0) Launcher viewer.app 1) Select any portion of the text of the default URL. (I tried "res", and also tried the entire URL.) 2) Wait 10 seconds. (YMMV if you have a faster or slower Mac) * RESULT - What happened Viewer quits. No error. Doesn't happen if you select text in the content region; only if you leave text selected in the URL field. - What was expected <#insert a random wise-ass comment here. Suggestions welcome for future bugs, Steve. ;> * REGRESSION - Occurs On viewer (1.5.98 build for Mac OS) viewer (1.4.98 build for Mac OS) - Doesn't Occur On viewer (1.5.98 build for Win32) * CONFIGURATIONS TESTED - PowerMac 8500/150 (233 Mhz 604e), 64 MB RAM, Mac OS 8.5.1

[Note: Bug does not occur on 11.4.98 debug build, as provided by Steve Dagley.]

When I broke at ExitToShell, here is the stack trace I get: 05B01DA0 PPC 05930708 nsMacMessageSink::DispatchMenuCommand(EventRecord&, long)+004FC 05B01CC0 PPC 05930BD8 nsMacMessageSink::DispatchMenuCommand(EventRecord&, long)+009CC 05B01C60 PPC 04980540 LPeriodical::DevoteTimeToRepeaters(const EventRecord&)+00048 05B01C00 PPC 05E9B6D8 TimerPeriodical::SpendTime(const EventRecord&)+00048 05B01BA0 PPC 05E9B5DC TimerPeriodical::RemoveTimer(TimerImpl*)+0002C 05B01B50 PPC 049762AC LArray::SetComparator(LComparator*, unsigned char)+ 00080 05B01B00 PPC 04976098 LArray::Sort()+0005C 05B01A90 PPC 049AEDE4 StPointerBlock::StPointerBlock(long, unsigned char, unsigned cha r)+00088 05B01A40 PPC 078C6F00 __throw+0009C 05B01860 PPC 078C6DB8 std::bad_exception::~bad_exception()+0039C 05B017E0 PPC 078C6004 __unregister_fragment+002CC 05B017A0 PPC 078C704C std::terminate()+00010 05B01760 PPC 078C6FF0 __destroy_global_chain+00060 05B01720 PPC 078C7BA0 __del_hdl+00060 05B016E0 PPC 078C7C58 __del_hdl+00118 Looks like an exception is being thrown and we're not catching it.

Joe francis (jfrancis@netscape.com) looking at this (but I don't have a Bugzilla account yet)

ignore previous comment - obviously, i DO have an account.

joe, any progress?

yes - this bug is a memory leak. I should have a fix checked in within 24 hours. This bug also revealed problems with having Powerplant in our build, which is why I've been removing it.

We were leaking every GraphicState that was ever made on behalf of an nsRenderingContextMac. And then we were runing out of mem in PowerPlant code, which threw an exception which caused immediate program termination (since we aren't catching excpetions - since we aren't supposed to be using them).

We were leaking every GraphicState that was ever made on behalf of an nsRenderingContextMac. And then we were runing out of mem in PowerPlant code, which threw an exception which caused immediate program termination (since we aren't catching excpetions - since we aren't supposed to be using them).

To solve part be of this problem I removed all Powerplant usage (see bug 2246). Solving the first part was a pair of simple changes in the destructor of nsRenderingContextMac.

Verified fixed using 1.20.99 build. Specifically: - highlighted both a portion of a URL ("res") and observed no crash after a minute - opened a few background windows, highlighted an entire URL (www.pcweek.com) and observed no crash after 5 minutes. Thanks, Joe!

Inserting Milestone info.