re: Tests and Valgrinding

[Dan Kegel]

Zachary Goldberg wrote:
> test.winehq.org ... is fantastic, but it doesn't give a very good feel for progress
> that all of these recent patches have been making.

I would love to see a feature added to test.winehq.org
which showed one tests's results across time (and platform).
That would let you see regressions (and progress) at a glance.
> Also, Dan Kegel has been extraordinarily diligent in valgrinding of
> late (/applause) and I just wanted to ask if we're doing anything to
> support him?

A few people are starting to run Valgrind themselves.
How about this: if you ever post a patch that causes a
valgrind warning, you send yourself to valgrind jail for a month,
and valgrind every patch you write before you send it in...
> perhaps making Valgrind-passing a requirement (along
> with not breaking tests) for being committed.  I think perhaps
> we should take a look at this and seriously consider implementing it.

As a first step, James and I could (along with posting the errors
to wine-devel) also post patches to revert the guilty changes.
(And if we do it quickly enough, perhaps Alexandre can
do the revert on the same day...)
Then the author could fix and resubmit.
How's that sound?
- Dan

Tests and Valgrinding

[Alexandre Julliard]

"Dan Kegel"  writes:
> As a first step, James and I could (along with posting the errors
> to wine-devel) also post patches to revert the guilty changes.
> (And if we do it quickly enough, perhaps Alexandre can
> do the revert on the same day...)
> Then the author could fix and resubmit.
> How's that sound?

I'm not going to revert patches just because they introduce a valgrind
warning, that would make a big mess of the history. If you can spot the
error before the patch goes in, great, otherwise you need to submit a
fix, not a revert.

Tests and Valgrinding

[Rob Shearman]

2008/7/2 Dan Kegel :
> Zachary Goldberg wrote:
> > Also, Dan Kegel has been extraordinarily diligent in valgrinding of
> > late (/applause) and I just wanted to ask if we're doing anything to
> > support him?
>
> A few people are starting to run Valgrind themselves.
> How about this: if you ever post a patch that causes a
> valgrind warning, you send yourself to valgrind jail for a month,
> and valgrind every patch you write before you send it in...

While I find Valgrind a very useful tool, not all of us are on
platforms that Valgrind supports. However, I'm happy to receive emails
about new Valgrind warnings from either Dan himself or from an
automated tool directed to a mailing list.
> > perhaps making Valgrind-passing a requirement (along
> > with not breaking tests) for being committed.  I think perhaps
> > we should take a look at this and seriously consider implementing it.
>
> As a first step, James and I could (along with posting the errors
> to wine-devel) also post patches to revert the guilty changes.
> (And if we do it quickly enough, perhaps Alexandre can
> do the revert on the same day...)
> Then the author could fix and resubmit.
> How's that sound?

Some Valgrind warnings are already present in base code and are only
triggered by new code or new tests. Reverting these type of patches
would be wrong IMHO, although every effort should be made to correct
the warning.