On Thu, Feb 23, 2012 at 10:25:44AM -0800, Ronald S. Bultje wrote:
> Hi,
> 
> On Thu, Feb 23, 2012 at 10:08 AM, Kostya Shishkov
> <[email protected]> wrote:
> > On Wed, Feb 22, 2012 at 04:49:59PM -0800, Ronald S. Bultje wrote:
> >> From: "Ronald S. Bultje" <[email protected]>
> >>
> >> Fixes crashes for extremely large resizes (several 100-fold).
> >>
> >> Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
> >> CC: [email protected]
> >> ---
> >>  libswscale/x86/swscale_mmx.c |   38 ++++++++++++++++++++++++++++++++++++++
> >>  1 files changed, 38 insertions(+), 0 deletions(-)
> >>
> >> diff --git a/libswscale/x86/swscale_mmx.c b/libswscale/x86/swscale_mmx.c
> >> index 764472e..64d5f0f 100644
> >> --- a/libswscale/x86/swscale_mmx.c
> >> +++ b/libswscale/x86/swscale_mmx.c
> >> @@ -117,6 +117,44 @@ void updateMMXDitherTables(SwsContext *c, int dstY, 
> >> int lumBufIndex, int chrBufI
> >>          const int16_t **chrUSrcPtr= (const int16_t **) chrUPixBuf + 
> >> chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
> >>          const int16_t **alpSrcPtr= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? 
> >> (const int16_t **) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + 
> >> vLumBufSize : NULL;
> >>          int i;
> >> +
> >> +        if (firstLumSrcY < 0 || firstLumSrcY + vLumFilterSize > c->srcH) {
> >> +            const int16_t **tmpY = (const int16_t **) lumPixBuf + 2 * 
> >> vLumBufSize;
> >> +            int neg = -firstLumSrcY, i, end = FFMIN(c->srcH - 
> >> firstLumSrcY, vLumFilterSize);
> >> +            for (i = 0; i < neg;            i++)
> >> +                tmpY[i] = lumSrcPtr[neg];
> >> +            for (     ; i < end;            i++)
> >> +                tmpY[i] = lumSrcPtr[i];
> >> +            for (     ; i < vLumFilterSize; i++)
> >> +                tmpY[i] = tmpY[i-1];
> >
> > looks okay, though those tails (;i < vXxxFilterSize; i++) can be optimised
> > since they set the same value
> 
> This code will soon disappear (once I finish my yasmification), thus
> any effort spent on it is pointless.

ok, commit as is, without any warranties (except that it compiles of course)
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to