Memory is not released after a failed call to realloc()

[philcostin]

Please see attached patch.

[waddlesplash]

That's not a memory leak. Also, do you realize you're literally calling free(NULL) there?

[philcostin]

I see what I've done there now yes, disregard the patch. However, the issue (although small) remains.

The first parameter to realloc() is not freed when the function returns null. The patch needs changing.

[Janus]

There is a memory leak.

[Janus]

I can't remove "has a patch". If somebody has the proper permissions please remove it.

[philcostin]

Please review this. Thanks.

[Janus]

Hi philcostin, you should use git format-patch to submit patch. (https://dev.haiku-os.org/wiki/CodingGuidelines/SubmittingPatches#Preparingtocreatepatchfile).

[Janus]

For the Haiku guidelines:

sockaddr *resized -> sockaddr* resized

You can remove the else

if (resized == NULL) { 
	free(address); 
	return NULL; 
} 
address = resized; 

The patch with this changes looks good to me. Has somebody more to say?

[philcostin]

patch with suggestions from Janus

[Janus]

Hi philcostin, I saw the full code of InterfaceAddress maybe this is not a memory leak. I leave the decision to someone who know the network stack better. The freed address is a member of the Interface class and could be used after the invalidation.

[pulkomandy]

Hi,

Janus: you can't remove "has a patch" directly. The idea is that you mark the patch as "obsolete" and the "has a patch" will be removed automatically.

[axeld]

Patch looks good, although I would remove the blank line between the allocation, and the check for its success. Thanks philcostin!

[Janus]

Hi Axel,

I'm afraid that the code that invoke the Prepare doesn't check if the realloc worked. If the realloc fail, I think there are other problems other than the memory leak. The address passed as a parameter is still used. Am I wrong?

[axeld]

I can't follow you there. There are three calls to Prepare(), one checks the result against NULL directly, the other two pass their result to the address module's set_to_defaults() which does check the parameters against NULL, too.

[Janus]

Thanks Axel, that's all I want to know.

[Barrett]

Just for the sake of not keeping one line patches open I'm going to apply it in the next days, if philcostin want can submit an updated patch :-) so that I don't have to add a follow-up fix.

[Barrett]

Patch applied in hrev49987.

[philcostin]

Not sure what happened there but so long as everyone agreed what the fix was, if indeed one ended up being needed at all then that's fine by me - I just re-found this ticket by accident :-)