Merge branch '8.10'
This commit is contained in:
commit
0be1370661
|
@ -17,6 +17,7 @@
|
|||
- check for null memory sources [kleisauke]
|
||||
- revise ppmload, fixing a couple of small bugs
|
||||
- signal error on EOF in jpegload more reliably [bozaro]
|
||||
- better error detection in spngload [randy408]
|
||||
|
||||
18/12/20 started 8.10.5
|
||||
- fix potential /0 in animated webp load [lovell]
|
||||
|
|
|
@ -373,11 +373,25 @@ vips_foreign_load_png_header( VipsForeignLoad *load )
|
|||
png->ihdr.bit_depth < 8 )
|
||||
png->fmt = SPNG_FMT_G8;
|
||||
|
||||
/* Try reading the optional transparency chunk. This will cause all
|
||||
* chunks up to the first IDAT to be read in, so it can fail if any
|
||||
* chunk has an error.
|
||||
*/
|
||||
error = spng_get_trns( png->ctx, &trns );
|
||||
if( error &&
|
||||
error != SPNG_ECHUNKAVAIL ) {
|
||||
vips_error( class->nickname, "%s", spng_strerror( error ) );
|
||||
return( -1 );
|
||||
}
|
||||
|
||||
/* Expand transparency.
|
||||
*
|
||||
* The _ALPHA types should not have the optional trns chunk (they
|
||||
* always have a transparent band), see
|
||||
* https://www.w3.org/TR/2003/REC-PNG-20031110/#11tRNS
|
||||
*
|
||||
* It's quick and safe to call spng_get_trns() again, and we now know
|
||||
* it will only fail for no transparency chunk.
|
||||
*/
|
||||
if( png->ihdr.color_type == SPNG_COLOR_TYPE_GRAYSCALE_ALPHA ||
|
||||
png->ihdr.color_type == SPNG_COLOR_TYPE_TRUECOLOR_ALPHA )
|
||||
|
|
Loading…
Reference in New Issue