Re: [Gnustep-cvs] gnustep/dev-libs/gsweb/GSWeb.framework GSWDynam...

[David Ayers]

David Wetzel schrieb:
> CVSROOT:      /cvsroot/gnustep
> Module name:  gnustep
> Branch:       
> Changes by:   David Wetzel <address@hidden>   05/09/30 15:30:00
> 
> Modified files:
>       dev-libs/gsweb/GSWeb.framework: GSWDynamicURLString.m 
> 
> Log message:
>       added - (NSData*) dataUsingEncoding: (NSStringEncoding)encoding 
> allowLossyConversion: (BOOL)flag
> 
> CVSWeb URLs:
> http://savannah.gnu.org/cgi-bin/viewcvs/gnustep/gnustep/dev-libs/gsweb/GSWeb.framework/GSWDynamicURLString.m.diff?tr1=1.14&tr2=1.15&r1=text&r2=text
> 

Even though this is the right thing to do for GSWDynamicURLString, I
believe this is may be covering up a bug in the generic implementation
of dataUsingEncoding:allowLossyConversion: of NSString.  .... and as I'm
looking at the code to explain what the issue is, the bug jumps at me:

The NSData was created with the freed (larger) buffer and not the one
returned by the realloc for UTF8 conversions.

Could you locally update -base and revert your fix just to test that
this was the issue?  (Well I'm about 99.9998% sure that this is the
issue but there may still be another issue hiding.)  But we do want to
keep your implementation of GSWDynamicURLString.

Cheers,
David

PS: A test for the test suite would also be great so if someone has the
time to implement a minimal custom string class...