[MacRuby-devel] copied Proc objects cause crash when used as Objective-C blocks

Thibault Martin-Lagardette thibault.ml at gmail.com
Fri Nov 19 04:20:23 PST 2010


I see.

It would be ideal if you could provide a small test sample, so that we can reproduce the bug :-)

-- 
Thibault Martin-Lagardette



On Nov 19, 2010, at 12:25, Alan Skipp wrote:

> Whoops, sorry about that. I find that I can get away without assigning to self in 'init' when doing a quick hack, but certainly not the recommended approach, especially when attempting to track down another bug.
> I've amended the initializer now and I still encounter the same problem.
> 
> Al
> 
> On 19 Nov 2010, at 10:53, Thibault Martin-Lagardette wrote:
> 
>> Hi Alan,
>> 
>> I didn't look in further details yet, but clearly your initWithBlock method is wrong, it should be:
>> 
>> - (id)initWithBlock:(void (^)())aBlock;
>> {
>> 	if ((self = [super init])) {
>> 		block = [aBlock copy];
>> 		NSLog(@"Block: %@", block);
>> 		block();
>> 	}
>> 	return self;
>> }
>> 
>> You are not assigning `self` to be equal to what `[super init]` returns, and this is, even without macruby or blocks, prone to crashes :-)
>> 
>> Can you try to fix this and then tell us if it's still crashing?
>> 
>> -- 
>> Thibault Martin-Lagardette
>> 
>> 
>> 
>> On Nov 19, 2010, at 10:14, Alan Skipp wrote:
>> 
>>> I've been attempting to get an objective-c framework to work with macruby and I believe I've found a bug in the way ruby Proc objects are copied when used as objective-c blocks.
>>> The copied block doesn't seem to persist correctly beyond the scope in which it was copied. It isn't deallocated, but calling it results in a crash. Typical error messages are:
>>>  wrong type NSCFSet (expected Proc) (TypeError)
>>>  wrong type NSRectSet (expected Proc) (TypeError)
>>> I'm guessing that there's a pointer to the wrong memory location?
>>> 
>>> Here's the Objective-C implementation:
>>> 
>>> @implementation TestBlock
>>> 
>>> - (id)initWithBlock:(void (^)())aBlock;
>>> {
>>> 	[super init];
>>> 	block = [aBlock copy];
>>> 	NSLog(@"Block: %@", block);
>>> 	block();
>>> 	return self;
>>> }
>>> 
>>> - (void)callBlock;
>>> {
>>> 	NSLog(@"block: %@", block);
>>> 	block();
>>> }
>>> 
>>> @end
>>> 
>>> 
>>> Within 'initWithBlock:', the copied block can be invoked without error. Attempting to do so from 'callBlock', results in a crash. The test framework can be used without error when using objective-c.
>>> 
>>> Here's the ruby controller code:
>>> 
>>> @b = TestBlock.alloc.initWithBlock Proc.new { puts "hello from ruby"} 
>>> 
>>> # this next line is called from a different scope and causes the crash
>>> @b.callBlock
>>> 
>>> 
>>> 2010-11-19 08:41:06.620 CallObjectiveCBlocks[7046:a0f] Block: <__NSAutoBlock__: 0x200be74a0>
>>> hello from ruby
>>> 
>>> 2010-11-19 08:41:20.011 CallObjectiveCBlocks[7046:a0f] block: <__NSAutoBlock__: 0x200be74a0>
>>> 2010-11-19 08:41:20.012 CallObjectiveCBlocks[7046:a0f] /Users/alan/Documents/programming/macruby/CallObjectiveCBlocks/build/Debug/CallObjectiveCBlocks.app/Contents/Resources/Controller.rb:21:in `call:': wrong type Array (expected Proc) (TypeError)
>>> 	from /Users/alan/Documents/programming/macruby/CallObjectiveCBlocks/build/Debug/CallObjectiveCBlocks.app/Contents/Resources/rb_main.rb:23:in `<main>'
>>> _______________________________________________
>>> MacRuby-devel mailing list
>>> MacRuby-devel at lists.macosforge.org
>>> http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel
>> 
>> _______________________________________________
>> MacRuby-devel mailing list
>> MacRuby-devel at lists.macosforge.org
>> http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel
> 
> _______________________________________________
> MacRuby-devel mailing list
> MacRuby-devel at lists.macosforge.org
> http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.macosforge.org/pipermail/macruby-devel/attachments/20101119/dcc44f09/attachment.html>


More information about the MacRuby-devel mailing list