Mercurial > mplayer.hg
changeset 29610:c35359578f8d
Rename CHECKED_ALLOC(Z) to FF_ALLOC(Z)_OR_GOTO and add context and label
parameters.
author | ramiro |
---|---|
date | Sun, 06 Sep 2009 00:07:35 +0000 |
parents | 9b0c934c78b5 |
children | 998475fffe0c |
files | libswscale/swscale.c |
diffstat | 1 files changed, 20 insertions(+), 20 deletions(-) [+] |
line wrap: on
line diff
--- a/libswscale/swscale.c Sat Sep 05 23:35:59 2009 +0000 +++ b/libswscale/swscale.c Sun Sep 06 00:07:35 2009 +0000 @@ -1451,12 +1451,12 @@ #endif // NOTE: the +1 is for the MMX scaler which reads over the end - CHECKED_ALLOC(*filterPos, (dstW+1)*sizeof(int16_t)); + FF_ALLOC_OR_GOTO(NULL, *filterPos, (dstW+1)*sizeof(int16_t), fail); if (FFABS(xInc - 0x10000) <10) { // unscaled int i; filterSize= 1; - CHECKED_ALLOCZ(filter, dstW*sizeof(*filter)*filterSize); + FF_ALLOCZ_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail); for (i=0; i<dstW; i++) { filter[i*filterSize]= fone; @@ -1467,7 +1467,7 @@ int i; int xDstInSrc; filterSize= 1; - CHECKED_ALLOC(filter, dstW*sizeof(*filter)*filterSize); + FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail); xDstInSrc= xInc/2 - 0x8000; for (i=0; i<dstW; i++) { @@ -1481,7 +1481,7 @@ int i; int xDstInSrc; filterSize= 2; - CHECKED_ALLOC(filter, dstW*sizeof(*filter)*filterSize); + FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail); xDstInSrc= xInc/2 - 0x8000; for (i=0; i<dstW; i++) { @@ -1520,7 +1520,7 @@ if (filterSize > srcW-2) filterSize=srcW-2; - CHECKED_ALLOC(filter, dstW*sizeof(*filter)*filterSize); + FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail); xDstInSrc= xInc - 0x10000; for (i=0; i<dstW; i++) { @@ -1608,7 +1608,7 @@ if (srcFilter) filter2Size+= srcFilter->length - 1; if (dstFilter) filter2Size+= dstFilter->length - 1; assert(filter2Size>0); - CHECKED_ALLOCZ(filter2, filter2Size*dstW*sizeof(*filter2)); + FF_ALLOCZ_OR_GOTO(NULL, filter2, filter2Size*dstW*sizeof(*filter2), fail); for (i=0; i<dstW; i++) { int j, k; @@ -1738,7 +1738,7 @@ // Note the +1 is for the MMX scaler which reads over the end /* align at 16 for AltiVec (needed by hScale_altivec_real) */ - CHECKED_ALLOCZ(*outFilter, *outFilterSize*(dstW+1)*sizeof(int16_t)); + FF_ALLOCZ_OR_GOTO(NULL, *outFilter, *outFilterSize*(dstW+1)*sizeof(int16_t), fail); /* normalize & store in outFilter */ for (i=0; i<dstW; i++) { @@ -2599,7 +2599,7 @@ if (!dstFilter) dstFilter= &dummyFilter; if (!srcFilter) srcFilter= &dummyFilter; - CHECKED_ALLOCZ(c, sizeof(SwsContext)); + FF_ALLOCZ_OR_GOTO(NULL, c, sizeof(SwsContext), fail); c->av_class = &sws_context_class; c->srcW= srcW; @@ -2840,10 +2840,10 @@ c->chrMmx2FilterCode = av_malloc(c->chrMmx2FilterCodeSize); #endif - CHECKED_ALLOCZ(c->lumMmx2Filter , (dstW /8+8)*sizeof(int16_t)); - CHECKED_ALLOCZ(c->chrMmx2Filter , (c->chrDstW /4+8)*sizeof(int16_t)); - CHECKED_ALLOCZ(c->lumMmx2FilterPos, (dstW /2/8+8)*sizeof(int32_t)); - CHECKED_ALLOCZ(c->chrMmx2FilterPos, (c->chrDstW/2/4+8)*sizeof(int32_t)); + FF_ALLOCZ_OR_GOTO(c, c->lumMmx2Filter , (dstW /8+8)*sizeof(int16_t), fail); + FF_ALLOCZ_OR_GOTO(c, c->chrMmx2Filter , (c->chrDstW /4+8)*sizeof(int16_t), fail); + FF_ALLOCZ_OR_GOTO(c, c->lumMmx2FilterPos, (dstW /2/8+8)*sizeof(int32_t), fail); + FF_ALLOCZ_OR_GOTO(c, c->chrMmx2FilterPos, (c->chrDstW/2/4+8)*sizeof(int32_t), fail); initMMX2HScaler( dstW, c->lumXInc, c->lumMmx2FilterCode, c->lumMmx2Filter, c->lumMmx2FilterPos, 8); initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmx2FilterCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4); @@ -2877,8 +2877,8 @@ goto fail; #ifdef COMPILE_ALTIVEC - CHECKED_ALLOC(c->vYCoeffsBank, sizeof (vector signed short)*c->vLumFilterSize*c->dstH); - CHECKED_ALLOC(c->vCCoeffsBank, sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH); + FF_ALLOC_OR_GOTO(c, c->vYCoeffsBank, sizeof (vector signed short)*c->vLumFilterSize*c->dstH, fail); + FF_ALLOC_OR_GOTO(c, c->vCCoeffsBank, sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH, fail); for (i=0;i<c->vLumFilterSize*c->dstH;i++) { int j; @@ -2914,23 +2914,23 @@ // allocate pixbufs (we use dynamic allocation because otherwise we would need to // allocate several megabytes to handle all possible cases) - CHECKED_ALLOC(c->lumPixBuf, c->vLumBufSize*2*sizeof(int16_t*)); - CHECKED_ALLOC(c->chrPixBuf, c->vChrBufSize*2*sizeof(int16_t*)); + FF_ALLOC_OR_GOTO(c, c->lumPixBuf, c->vLumBufSize*2*sizeof(int16_t*), fail); + FF_ALLOC_OR_GOTO(c, c->chrPixBuf, c->vChrBufSize*2*sizeof(int16_t*), fail); if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat)) - CHECKED_ALLOCZ(c->alpPixBuf, c->vLumBufSize*2*sizeof(int16_t*)); + FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize*2*sizeof(int16_t*), fail); //Note we need at least one pixel more at the end because of the MMX code (just in case someone wanna replace the 4000/8000) /* align at 16 bytes for AltiVec */ for (i=0; i<c->vLumBufSize; i++) { - CHECKED_ALLOCZ(c->lumPixBuf[i+c->vLumBufSize], VOF+1); + FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf[i+c->vLumBufSize], VOF+1, fail); c->lumPixBuf[i] = c->lumPixBuf[i+c->vLumBufSize]; } for (i=0; i<c->vChrBufSize; i++) { - CHECKED_ALLOC(c->chrPixBuf[i+c->vChrBufSize], (VOF+1)*2); + FF_ALLOC_OR_GOTO(c, c->chrPixBuf[i+c->vChrBufSize], (VOF+1)*2, fail); c->chrPixBuf[i] = c->chrPixBuf[i+c->vChrBufSize]; } if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) for (i=0; i<c->vLumBufSize; i++) { - CHECKED_ALLOCZ(c->alpPixBuf[i+c->vLumBufSize], VOF+1); + FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i+c->vLumBufSize], VOF+1, fail); c->alpPixBuf[i] = c->alpPixBuf[i+c->vLumBufSize]; }