Safe allocation wrappers #754
Conversation
|
Just quoting the last parts of the conversation in a merged PR from @michael-grunder
Thanks... and I think the changes in this PR would work for me. 👍 |
|
Would be great to get some more eyes on this. :) |
No, but |
|
Merged, thanks everyone! |
| @@ -57,7 +58,7 @@ static unsigned int callbackHash(const void *key) { | |||
|
|
|||
| static void *callbackValDup(void *privdata, const void *src) { | |||
| ((void) privdata); | |||
| redisCallback *dup = malloc(sizeof(*dup)); | |||
| redisCallback *dup = hi_malloc(sizeof(*dup)); | |||
There was a problem hiding this comment.
Can still result in a SEGFAULT,
If hi_malloc returns NULL -> next call to memcpy will SEGFAULT
There was a problem hiding this comment.
Just waking up here but how can hi_malloc return NULL if HIREDIS_OOM_HANDLER is called? (eg. _exit?)
There was a problem hiding this comment.
@lamby yeah, I forgot to redo the comment, talked about it already with @michael-grunder
either way personally I'm against doing abort() if we can handle it, for example when doing callback-allocation, we can return error instead of abort, remember that this is a 3rd-party library for most users, and crashing someone's application due to internal errors is not the way to do things IMHO
Adds allocation wrappers that invokes a defined OOM handler.
My goal was the smallest surface area of changes possible. Once we merge this it might be prudent to replace every
malloc/freecall with our wrappers.