On Thu, Aug 6, 2009 at 4:09 PM, Dan Kegel <d...@kegel.com> wrote:
> Despite having had several valgrind/purify fixit weeks,
> our bug tracker is still full of 'em.
Hence my email last night.
i.e. Next time we hold a valgrind fixit,
> how about we just delete the non-upstream-bugs
> suppressions files?
I think this would be a mistake.
There is a lot of accumulated "knowledge" in the valgrind suppresions file.
Doubtless some of it is obsolete (suppressions that no longer apply).
However, both erg and phajdan.jr have removed things like this (LayoutTest
cruft for erg) recently, and that can be done in a
change-one-variable-at-a-time controlled fashion.
By contrast, dumping the whole file would have two effects:
(1) If you're really serious about it, the tree will be closed for at least
several days. This is an extremely high cost.
(2) You will be getting intermittent red far more frequently for weeks as
the rarest of the suppressions in that file occasionally rear their head.
Flakiness is very bad.
Compare this to today, where the valgrind bots went red, I put hclam on it,
and he had a leak fix inside of an hour. Non-flaky tests mean we can fix
new failures fast.
For better or worse, I believe we simply have to get people to clean things
up themselves. Hence my directive last night for every Googler to put
something on his or her OKRs. If we start fixing flaky LayoutTests, memory
bugs, etc., the hard-to-track-down valgrind suppressions will be removable
too.
PK
--~--~---------~--~----~------------~-------~--~----~
Chromium Developers mailing list: chromium-dev@googlegroups.com
View archives, change email options, or unsubscribe:
http://groups.google.com/group/chromium-dev
-~----------~----~----~----~------~----~------~--~---
