Possible bug in apr_rmm_realloc.   When calculating the value of
blk, it looks like it should be:

blk = (rmm_block_t*)((char*)rmm->base + old - RMM_BLOCK_SIZE);

instead of

blk = (rmm_block_t*)((char*)rmm->base + old);

otherwise blk is pointing at the caller's data.

Keith

Reply via email to