Hi. On Thu, May 5, 2011 at 2:42 PM, Alex Converse <alex.conve...@gmail.com> wrote: > On Thu, May 5, 2011 at 11:39 AM, İsmail Dönmez <ism...@namtrac.org> wrote: >> Hi; >> >> 2011/5/5 Aℓex Converse <aconve...@google.com> >>> >>> 2011/5/5 Aℓex Converse <aconve...@google.com>: >>> > On Thu, May 5, 2011 at 11:15 AM, Ronald S. Bultje <rsbul...@gmail.com> >>> > wrote: >>> >> Hi, >>> >> >>> >> 2011/5/5 Aℓex Converse <aconve...@google.com>: >>> >>> 2011/5/5 Jindřich Makovička <makov...@gmail.com> >>> >>>> >>> >>>> Hi, >>> >>>> >>> >>>> although I don't know much about wma voice, I think this piece of >>> >>>> code >>> >>>> doesn't do what is expected (sizeof returns the size of pointer in >>> >>>> this case, i.e. 4 or 8 based on arch, not 25). >>> >>>> >>> >>>> decode_vbmtree(GetBitContext *gb, int8_t vbm_tree[25]) >>> >>>> { >>> >>>> ... >>> >>>> memset(vbm_tree, 0xff, sizeof(vbm_tree)); >>> >>>> >>> >>> >>> >>> Thanks. >>> >> >>> >> Did you confirm that the code indeed doesn't do what it's supposed to >>> >> do? >>> >> >>> >> (In that case, I'm surprised it works at all, but that's probably just >>> >> me...) >>> >> >>> > >>> > I confirmed that I know that detail of C. >>> > >>> >>> #include <stdio.h> >>> void x(char y[25]) { printf("%zu\n", sizeof(y)); } >>> int main(void) { char z[26]; x(z); } >>> >>> $ ./a.out >>> 8 >> >> I think the question is, does FATE pass? If not, then we have to update the >> reference, if yes then wma tests are unsurprisingly incomplete. >> Regards, >> ismail >> > > $ make fate-wmavoice-11k fate-wmavoice-19k fate-wmavoice-7k > HOSTCC tests/tiny_psnr.o > HOSTCC tests/tiny_psnr > TEST wmavoice-11k > TEST wmavoice-19k > TEST wmavoice-7k > > > wma tests are unsurprisingly incomplete.
No, the WMA Voice tests don't test this particular error condition (codec referencing an uninitialized vmbtree entry). 0xff=-1=interpreted as an error in the codec, default=av_mallocz() return=0 means that if the tree entry was uninitialized (broken file), codec acts as if it was a valid file and probably outputs noise. Nothing particularly bad. Oink. Patch OK. Ronald _______________________________________________ libav-devel mailing list libav-devel@libav.org https://lists.libav.org/mailman/listinfo/libav-devel