On Thu, Sep 24, 2009 at 10:10 AM, Mark Mentovai <[email protected]> wrote: > The Valgrind test step doesn't turn red for test failures, only for > unsuppressed leaks.
That's not quite true. It also turns red for nasty memory errors like use-after-free and uninitialized memory reads. The memory leaks are the bulk of the warnings, but they're not really why valgrind is so important; if all we cared about were memory leaks, we could find them faster with other tools. > I wanted to ask dank about that last week, but he > wasn't around when it initially upset me, so I moved on. Let's ask > him now. dank? I think ignoring test errors under Valgrind was a temporary expedient while we were bringing it up. I agree that test failures under valgrind should count as errors now. But before we can switch that on, we first need to clear up the tests that don't pass under valgrind. (They're easy to find; just look at green valgrind runs like http://build.chromium.org/buildbot/waterfall/builders/Chromium%20Mac%20(valgrind)/builds/2071/steps/valgrind%20test:%20unit/logs/stdio or http://build.chromium.org/buildbot/waterfall/builders/Linux%20Tests%20(valgrind)(3)/builds/165/steps/valgrind%20test:%20ui/logs/stdio and look for test failures at the bottom.) Seems like a good thing for the stability task force to tackle, but if anyone else wants to have a look, please do. - Dan p.s. I'm out on vacation until Wed Sept 30. --~--~---------~--~----~------------~-------~--~----~ Chromium Developers mailing list: [email protected] View archives, change email options, or unsubscribe: http://groups.google.com/group/chromium-dev -~----------~----~----~----~------~----~------~--~---
