On 02.06.21 14:03, Olaf Hering wrote:
> Am Wed, 2 Jun 2021 09:29:08 +0200
> schrieb Juergen Gross <jgross@suse.com>:
> 
>>> +    ctx->restore.m = malloc(sizeof(*ctx->restore.m));
>>> +    if ( !ctx->restore.m ) {
>>
>> ... this case might trigger without the full series applied, due to
>> allocating zero bytes (same for the save side below).
> 
> Such bisection point with a libc that returns NULL would be just bad luck.

Sure, but sending a patch which is known to break bisecting is bad
behavior. You could even add a dummy element (with a comment indicating
its purpose) which could be removed when the first "real" structure
element is being added.

> See git-bisect(1) "Avoiding testing a commit" how to deal with it, in the unlikely case it actually triggers.

It can be avoided, yes, but you need to search for the reason the
failure occurred first. And this debugging effort should be avoided
if possible.


Juergen