Re: Interesting error in GWorkspace

[Eric Wasylishen]

On 2011-09-11, at 10:24 AM, Riccardo Mottola wrote:

> Hi Eric,
> 
> On 09/09/11 21:53, Eric Wasylishen wrote:
>> Hi Riccardo,
>> I think I found the problem. Sorry about this, it was a sloppy mistake on my 
>> part.
>> 
>> When I modified _doImageCache:, I was assuming the rep parameter would be 
>> non-nil, so I sent [rep size] to it and used that when setting up the cache. 
>> But in fact, the rep parameter is nil when you lock focus on an image with 
>> no representations (which the Dock.m code was doing.) I guess we should add 
>> a test case for that in the gui test suite, since it would have caught this 
>> mistake.
> It works fine for me again!

Great!

> Very good. I wonder why it happened only on certain computers I had... 
> actually only on my laptop.

I think it was just a coincidence - the code with the bug was able to handle 
[rep size] returning (0, 0), so if rep was nil on platforms where [nil size] is 
(0, 0), it would silently work.

I was just thinking a bit how we can detect and avoid mistakes like this in the 
future. According to 
http://www.gnustep.org/resources/documentation/Developer/Base/ProgrammingManual/manual_2.html#Messages
 , the return value of a message send to nil is only defined when the selector 
return types is an object. 

I wonder if the clang static analyzer can find places where we send a message 
returning something other than an object to something that might be nil?

Eric

> My workstation runs a different OS, but still, the same backend and about the 
> same setup, with desktop and backdrop.
> 
> Good work. With GWorkspace running again I can test the service stuff much 
> better now.
> 
> Riccardo
> 
> _______________________________________________
> Discuss-gnustep mailing list
> Discuss-gnustep@gnu.org
> https://lists.gnu.org/mailman/listinfo/discuss-gnustep