#Carbon copy cloner os x 10.3.9 Patch
#Carbon copy cloner os x 10.3.9 code

Time-consuming, even with the help of Oink. To do this (and particularly to test it) will be very difficult and The only way to fix that is to catch them at all the boundary > problem Objective-C exceptions must not propagate through XPCOM > The fix for this bug really needs to be a focused on the core

Just that - stop-gap measure until we can come up with something But (as you'll have noticed) my quick-fix is That use Objective C exception handling, and though I'm currently Likely only degrade the performance of external frameworks and plugins > for all plug-ins? Can you say that for all future OS versions? > exception handling internally for that own needs? Can you say that > How do you know that no frameworks used by Camino don't use The following at a Terminal prompt to use them both at once:ĭefaults write OSCrashReporter 1 Is enabled you only see the Breakpad window on a crash. (By default Breakpad pre-empts Apple's crashreporterd - if Breakpad The MOZ_AIRBAG environment variable for Breakpad to work (e.g. In ones more than a few days old you need to set (By the way, in the most recent Minefield nightlies Breakpad seems toīe on by default. But what testing I've done indicates that it getsĪlong fine with Breakpad and with Apple's crashreporterd. I haven't yet done much testing, so I'm not yet requesting a reviewįor this patch. That can be achieved quickly (with very little reliable information).

#Carbon copy cloner os x 10.3.9 Patch

In the meantime I think my patch is the best solution With luck I'll be able to revise this patch or roll it backĪltogether. Over the next month or two I hope to find out more about this subject. We don't yet know enough about the possible exceptions to determine

#Carbon copy cloner os x 10.3.9 code

But the problem is that we can't (orĭon't yet know that we can) rely on Apple's code to distinguishīetween exceptions that should be fatal and ones that shouldn't. Not every Cocoa exception is fatal, so doing this will cause theīrowser to crash 'too often'. It hooks - and crashes the browser whenever there's Splinter Review Here's a quick-fix patch that does what I outlined in comment #12. In other words, all Obj-C exceptions must be caught within down-calls from XPCOM into Obj-C code, and additional care must be taken in mixed Obj-C/XPCOM code to avoid throwing exceptions that fail to destroy local XPCOM objects. The problem described in this bug has a very well defined scope: Obj-C exception handling must avoid unwinding any stack frame that contains XPCOM code. XPCOM is not (yet) C++ exception aware or friendly. > There is also Wolf Rentzsch's NSXException.Īlso not useful. This uses NSSetUncaughtExceptionHandler, so is also not useful. NSExceptionHandler's delegate and it's delegate methods. Once this gets called, the stack has already been unwound back up the top. > of the old style, from when ObjC++ made it's debut: Old-style NS_DURING and friends and new-style Here's an example > There are three ways of catching and handling all exceptions in ObjC. It would be wonderful if we could use the native exception handling features (I would suggest option 3), rather than having to resort to black magic of some sort.

The new 10.3 exception handling is a language feature, not just a macro. Effectively, each NSIMETHODIMP method needs to handle all exceptions. According to timeless, you're not allowed to longjmp past XPCOM boundaries. It seems that at least on 10.2, it is highly likely that all of these are implemented setjmp and longjmp. 20:53:01.869 frameworkless growl app *** Uncaught exception: *** -: selector not recognized This is all particularly nasty because literally any call to an Objective-C method can generate an exception. There is also Wolf Rentzsch's NSXException. Old-style NS_DURING and friends and new-style Here's an example of the old style, from when ObjC++ made it's debut: Ģ. There are three ways of catching and handling all exceptions in ObjC.ġ.