Do you find that you have time to figure out if rebaselining a test is the
right thing to do while you're actively gardening? Maybe I just work too
slowly, but I often find that if I'm trying to rebaseline on the fly, it
requires that I do at least *some* investigation of the test failure to make
sure I'm not rebaselining in an error (or rebaselining a test that is merely
flaky) which slows me down enough that I fall behind and inevitably am
crushed by the WebKit juggernaut.

-atw

On Fri, Jan 8, 2010 at 4:33 PM, Dimitri Glazkov <[email protected]>wrote:

> Before heading out for the weekend, I just want to mention this:
> rebaseline tool really, really rocks. And yesterday I discovered an
> option that I, to my shame, hadn't seen before: -w. This option pulls
> baselines from the canary. It's like getting test expectations from
> the future!
>
> In other words, there are no more excuses for you, dear WebKit
> gardeners, to commit those ghastly BUG_SOMENAME entries in
> test_expectations.
>
> The workflow is mind-numbingly simple:
>
> 1) Identify tests that need rebaselining prior to rolling
> 2) Add these tests to test_expectations.txt as if you were to commit
> them -- except add a REBASELINE flag next to BUG_SOMENAME
> 3) Run rebaseline -w
> 4) Make sure that the tool ran and removed these entries from
> test_expectations.txt
> 5) Create CL -- you will notice how new expected result files are
> conveniently added for you.
> 6) Enjoy regression-free WebKit roll.
>
> Big thanks to Victor Wang for this amazing instrument of everlasting
> harmony. Send him your accolades. Or cash. If you find bugs/quirks,
> fix them.
>
> :DG<
>
> P.S. As part of gardener/sheriff process overhaul, I will be adding
> these and other helpful tips/hints to our gardening doc.
>
> --
> Chromium Developers mailing list: [email protected]
> View archives, change email options, or unsubscribe:
>    http://groups.google.com/group/chromium-dev
>
-- 
Chromium Developers mailing list: [email protected] 
View archives, change email options, or unsubscribe: 
    http://groups.google.com/group/chromium-dev

Reply via email to