[PATCHES BUG 13865] memcache tracking of talloc object sizes

[Jeremy Allison]

On Fri, Mar 29, 2019 at 01:07:42PM -0700, Jeremy Allison wrote:
> On Fri, Mar 29, 2019 at 01:00:33PM -0700, Jeremy Allison via samba-technical wrote:
> 
> > Actually I'm wrong - we can't mix talloc/non-talloc
> > stores here (it crashes, I tested :-). But we still
> > need to take the talloc'ed size into account.
> > 
> > > > 
> > > > 281                                 TALLOC_FREE(ptr);
> > > > 282                         }
> > > > 283                         /*
> > > > 284                          * We can reuse the existing record
> > > > 285                          */
> > > > 286                         memcpy(cache_value.data, value.data, value.length);
> > > > 287                         e->valuelength = value.length;
> > > > 288                         return;
> > > > 289                 }
> > > > 
> > > > Probably need to add a test for this too...
> > > 
> > > Yes, will do. I will be busy with something else this afternoon, but i
> > > should have an update for this early next week.
> > 
> > I'm nearly done with the fix and test. If I can
> > get it working I'll post for your review.
> 
> Here is the updated patchset containing the fix for the
> corner case and additional test for it. I checked that
> with the corner case fix commented out the test fails,
> but passes with it added.
> 
> CI build here:
> 
> https://gitlab.com/samba-team/devel/samba/pipelines/54311307
> 
> Please review and push if happy !

Actually this patchset makes:

make test TESTS=samba3.base.rw1

fail repeatedly. Interesting, I'm trying to track
down why (wouldn't have thought it had an effect,
but it clearly does :-).