Hi, On Thu, Jan 10, 2013 at 11:33 AM, Luca Barbato <[email protected]> wrote: > On 10/01/13 20:12, Dale Curtis wrote: >> From: "Paweł Hajdan, Jr" <[email protected]> >> >> Memory passed to av_realloc cannot be allocated using memalign. >> >> From realloc(3): >> >> The realloc() function changes the size of the memory block pointed to >> by ptr to size bytes. (...) Unless ptr is NULL, it must have been returned >> by an earlier call to malloc(), calloc() or realloc(). >> > > The patch itself is fine (and queue locally), I'm trying to figure out > what's the best to avoid such issues since few malloc implementation > support aligned realloc. > > - in debug mode mark malloc and realloc memory differently and assert > accordingly in realloc, and make quite clear we have 2 kinds of memory. > > - support aligned realloc (means wasting 8 bytes per allocation if we > want to keep the same signature, or introduce av_realloc2 that takes in > account the size to be preserved)
Why so complex? Is there currently a need for an aligned realloc? If not, don't add one. Wait until it becomes necessary. Likely, it will never be necessary. (Incidentally, if there were a real use case for aligned realloc, Linux and Mac would probably support it natively. They don't. Probably - again - because there is no need for it.) Ronald _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
