Hi Ken,
----- Original Message -----
> I have 15 different failures for qa/720 across 16 systems.
> [...]
> Is there any technically plausible reason that these should be treated
> as synonymous and magically filtered into a passing test?
Yeah, SASL errors can come from different places depending on the local
SASL (runtime) configuration - from the library code or from the plugin
libraries, which may be activated in different orders for each platform.
> 2. valgrind failures in a blinding assortment of variations around
> _plug_get_password ... I'm currently trying a minimalist valgrind
> suppression of the form
>
> {
> cyrus-sasl callback leak, RHBZ1191183, common case variant
> Memcheck:Leak
> fun:calloc
> ...
> fun:_plug_get_password
> ...
> fun:sasl_client_step
> ...
> }
>
> Does this look OK?
Will give it a go.
> How are others going with this test?
>
Pretty ordinary - I'd begun extending the valgrind filter, but had it
a whole lot more specific than the above (and it wasn't right yet) - I
will try yours instead & get back to you.
I also have a fix pending to make _valgrind_clean_assert (qa/666) work
again for older valgrind versions without vgdb support.
cheers.
--
Nathan
|