Re: Crash in lockFocus?

FROM : j o a r
DATE : Sun Jan 13 21:06:28 2008

Hello,

Could this be a thread-safety problem? If you're using multiple 
threads, are you sure that you're playing by the (barely documented) 
rules?

j o a r


On Jan 13, 2008, at 11:52 AM, glenn andreas wrote:

> I've got code that makes a small thumbnail of another image by 
> creating a new NSImage, lockFocus on it, and then drawing, 
> unlocking, and autoreleasing it:
>
>     NSImage *retval = [[NSImage alloc] initWithSize:newSize];
>     [retval setCacheMode:NSImageCacheNever];
>     [retval lockFocus];
>     [preview drawInRect:NSMakeRect(0,0,newSize.width,newSize.height)
>           fromRect:NSMakeRect(0,0,curSize.width, curSize.height)
>           operation:NSCompositeCopy
>           fraction:1.0];
>     [retval unlockFocus];
>     return [retval autorelease];
>
>
> Unfortunately, occasionally the lockFocus will crash:
> Program received signal:  “EXC_BAD_ACCESS”.
> [Switching to process 19156 thread 0x3f23]
> (gdb) bt
> #0  0xfffeff10 in objc_msgSend_rtp ()
> #1  0x9141bc20 in _NXAllocateImageCache ()
> #2  0x9141b31c in -[NSCachedImageRep 
> _initWithSize:depth:separate:alpha:allowDeep:] ()
> #3  0x9141b0a0 in -[NSImage lockFocus] ()
> #4  0x001330b4 in -[NSDictionary(QutilCorpse) previewOfSize:] 
> (self=0x67e050, _cmd=0x174ee0, size=128) at /Volumes/Sources/qui/
> NSDictionary_QutilCorpse.mm:50
> (gdb) p newSize
> $1 = {
>  width = 128,
>  height = 128
> }
> (gdb) po retval
> NSImage 0x687e70 Size={128, 128} Reps=(
> )
>
>
> This only rarely happens, but it does happen, and seems to only 
> happen on Leopard (10.5.1) - I tried adding the NSImageCacheNever 
> line to try to see if it would go away based on some searching of 
> the archives, but it didn't help. I really don't see what could 
> possibly be wrong - it allocates the image of the correct size, and 
> it really doesn't seem like lockFocus should crash (maybe throw an 
> exception if, for some reason it couldn't lock focus, but not 
> crash).  Worse, this problem is very intermittent, so not a lot to 
> test (I may see it every few days, sometimes a couple of times a day)
>
> Any ideas of how to work around this problem?