Re: Building GNUstep from source - failed tests

[Richard Frith-Macdonald]

On 13 Jan 2013, at 17:15, Fred Kiefer wrote:

> 
> I just run the base tests on my 64bit OpenSuse machine and had a failed file 
> here as well (and also a few expected failed tests from NSLocale, we need to 
> mark those as hopes). The failing file in create.m in NSPointerArray and 
> running it with a debugger gives this backtrace:
> 
> 
> #0  0x00007ffff7421731 in objc_msg_lookup () from /usr/lib64/libobjc.so.4
> #1  0x00007ffff78bcf18 in pointerFunctionsAssign (value=<optimized out>, 
> addr=0x644960,
>    PF=<optimized out>) at NSConcretePointerFunctions.h:142
> #2  pointerFunctionsAssign (value=<optimized out>, addr=0x644960, 
> PF=<optimized out>)
>    at NSPointerArray.m:455
> #3  pointerFunctionsMove (new=0x644960, PF=0x643030, old=<optimized out>)
>    at NSConcretePointerFunctions.h:172
> #4  -[NSConcretePointerArray insertPointer:atIndex:] (self=0x643028, 
> _cmd=<optimized out>,
>    pointer=0x608740 <_OBJC_INSTANCE_8>, index=0) at NSPointerArray.m:467
> #5  0x00000000004056c3 in main () at create.m:33

Thanks for the report ... it wasn't happening on my system, but I think that 
was just luck of the memory layout.
The problem was with some uninitialised memory.
I changed the pointer array code to clear new memory when the array is 
allocated/grown, so we never think an unused location has anything in it.