On 31 July 2013 20:41, Kenneth Graunke <kenn...@whitecape.org> wrote:
> On 07/31/2013 05:16 PM, Ian Romanick wrote:
>
>> On 07/30/2013 03:16 PM, Kenneth Graunke wrote:
>>
>>> This code is getting ugly. It looks like the idea is to allocate two
>>> (now three) separate arrays, but somebody decided to be "clever" and
>>> batch them into a single calloc/free big enough to hold all of them.
>>>
>>
>> Yeah, that was me. Since way back, I hate dipping into malloc/free for
>> a bunch of small, short-lived allocations. What I really want to use is
>> alloca, but alas.
>>
>
> In C++ you can just do:
>
> struct gl_shader *vert_shader_list[prog->**NumShaders];
>
> which basically does alloca() as far as I can tell. The problem is, any
> out of bounds writes smash the stack, and can't be caught by tools like
> Valgrind.
>
> --Ken
>
Actually that's a C99 feature ("variable length arrays"), although GCC
allows it as a C++ extension (try compiling with "-pedantic" and you'll get
a warning). As far as I can tell from some internet research, MSVC doesn't
support the feature at all, so unfortunately we can't use it in core mesa.
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
