Re: [MACTCL] File dialog crashes under Leopard

[Andy C. <an...@ac...>]

Hi Daniel,

I did a workaround/hack which replaces the system dialogs with the non- 
system Tk versions. (the color picker also displayed the same  
crashes).  This has made some of our Mac users a lot happier.

I just took some stuff from tk.tcl - you can see the script here:

http://www.inivis.com/forum/showthread.php?p=20348

One side effect of this workaround is that under 8.4.16, the dialogs  
sometimes popup under the main window(?) (or at least they won't  
display until the main window is clicked) - this doesn't seem to  
happen under 8.4.14.   The dialog titlebar is also different.  I  
assume these differences may be because of your changes you mentioned?  
Tk 8.4.14 appears to handle the old dialogs better.

Another thing - the crash on displaying the help menu (with the new  
apple search field added) doesn't happen with 8.4.14 but always  
happens under 8.4.16.  (sorry, I never tried 8.4.15).  Might this  
indicate some problem with the help menu/focus was added to Tk after  
8.4.14?

Many thanks for your help,

Andy



On 1 Nov 2007, at 13:00, Daniel A. Steffen wrote:

> Hi Jon,
>
> On 01/11/2007, at 22:59, Jon Guyer wrote:
>
>> Is this thread on carbon-dev relevant?
>>
>>   http://lists.apple.com/archives/Carbon-dev/2007/Oct/msg00407.html
>
> saw that, thanks, it doesn't look particularly related, they have a  
> deadlock in UI drawing code, whereas our crash looks like an  
> unexpected exception in Cocoa-Carbon focus handling.
>
> Given that the exact same crash also happens when drawing the help  
> menu with its new Coco-based search field (Andy sent me the  
> backtrace for that off-list), I don't think it has to do with the  
> file dialog specifically, I suspect that we trigger some general  
> issue in Cocoa-Carbon integration (all the crashes occur at the same  
> location: -[NSView unlockFocus] -> -[NSFocusStack unfocusView:] ->  
> NSRemoveAltHandler2 -> CFDoExceptionOperation).
>
> Possibly we need some new Leopard magic setup that normal Carbon  
> startup or RAEL does and we skip with our somewhat non-standard  
> startup and event handling.
> The only documented such magic is calling NSApplicationLoad and  
> setting up an autorelease pool, Jeff kindly tested yesterday that  
> adding both of these has no effect on the crash...
>
> I plan to post a question about this issue on carbon-dev tomorrow,  
> if anybody has filed a radar on this yet, please let me know the  
> report number so that I can mention it. Thanks.
>
> Cheers,
>
> Daniel
>
> -- 
> ** Daniel A. Steffen                   **
> ** <mailto:da...@us...>  **
>
>