Re: CGContextSelectFont spinlock
Re: CGContextSelectFont spinlock
- Subject: Re: CGContextSelectFont spinlock
- From: "A.M." <email@hidden>
- Date: Thu, 16 Sep 2010 18:26:21 -0400
On Sep 16, 2010, at 5:54 PM, Stevo Brock wrote:
> Great suggestion.
>
> Here's the full list of backtraces from the iPad (3.2.2)... Spin lock is in Thread 1 at the end...
>
> Thread 1 (thread 12035):
> #0 0x33af18a4 in spin_lock ()
> #1 0x33af21aa in pthread_once ()
> #2 0x33d795a2 in CGFontCreate ()
> #3 0x33df21e0 in CGFontCreateWithDataProvider ()
> #4 0x3414cb72 in AddFontsFromURLOrPath ()
> #5 0x3414cc98 in Initialize ()
> #6 0x3414d360 in GSFontInitialize ()
> #7 0x33d78e12 in initialize_font_database ()
> #8 0x33af21c8 in pthread_once ()
> #9 0x33d78d46 in CGFontCreateWithFontName ()
> #10 0x33d78d1e in CGFontCreateWithName ()
> #11 0x33de8950 in CGContextSelectFont ()
> #12 0x00041cf8 in -[SceneTextController drawGL:] (self=0x274b20, _cmd=0x4b77c, offset={x = 0, y = 0}) at .../SceneTextController.m:162
> #13 0x000404ee in -[SIGameHighScoreDisplayController drawGL:] (self=0x2749c0, _cmd=0x4b77c, offset={x = 0, y = 0}) at .../SIGameHighScoreDisplayController.m:109
> #14 0x000128a8 in -[SceneController drawGL] (self=0x259280, _cmd=0x46df0) at .../SceneController.m:196
> #15 0x0001fabe in -[SceneTransitionPush drawGL] (self=0x267d40, _cmd=0x46df0) at /.../SceneTransitionPush.m:152
> #16 0x0001af76 in -[SceneViewController updateAndDraw:] (self=0x220dc0, _cmd=0x46a33, timer=0x259420) at .../SceneViewController.m:674
> #17 0x303b1eb8 in __NSFireTimer ()
> #18 0x314d8d22 in CFRunLoopRunSpecific ()
> #19 0x314d84e0 in CFRunLoopRunInMode ()
> #20 0x3414e0da in GSEventRunModal ()
> #21 0x3414e186 in GSEventRun ()
> #22 0x32241430 in -[UIApplication _run] ()
> #23 0x3223f95a in UIApplicationMain ()
> #24 0x00029712 in main (argc=1, argv=0x2ffff5f0) at .../main.m:10
I can confirm that this is a bug in Apple's font implementation- here is a sample program which exhibits the same characteristics:
#include <pthread.h>
pthread_once_t oncetoken = PTHREAD_ONCE_INIT;
void once()
{
pthread_once(&oncetoken,once);
}
int main()
{
pthread_once(&oncetoken,once);
}
This program gets stuck on __spin_lock() in the same fashion as above. The bug is that a pthread_once()-initiated function is calling through to another method which calls pthread_once with the same once token (because the font database has not yet been created). For what it's worth, Linux also does not like nested pthread_once and will choke in the same fashion.
Unfortunately, due to the nature of closed source software, there is nothing we can do about this...
Cheers,
M_______________________________________________
Cocoa-dev mailing list (email@hidden)
Please do not post admin requests or moderator comments to the list.
Contact the moderators at cocoa-dev-admins(at)lists.apple.com
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden