W3C home > Mailing lists > Public > public-css-testsuite@w3.org > January 2017

Re: Writing Modes PR & Testsuite

From: Koji Ishii <kojiishi@gmail.com>
Date: Fri, 20 Jan 2017 19:19:05 +0900
Message-ID: <CAN9ydbXreEJhT+nv8nvcL9_VeRcYn8342+koDC3-nhFEswt8jQ@mail.gmail.com>
To: Geoffrey Sneddon <me@gsnedders.com>
Cc: www-style list <www-style@w3.org>, "public-css-testsuite@w3.org" <public-css-testsuite@w3.org>, GĂ©rard Talbot <css21testsuite@gtalbot.org>
2017-01-19 2:07 GMT+09:00 Geoffrey Sneddon <me@gsnedders.com>:

>
> When all browsers pass *modified versions of the tests* (because, by
> definition, the reftests in wpt and csswg-test require pixel-for-pixel
> identity), I don't think we should consider the testsuite good. After
> all, if it were good, nobody would feel the need to modify it!
>

Blink doesn't run "modified versions of the tests", at least in the
anti-alias POV. We do have some modifications during the import process for
other purposes, we're trying to minimize/stop doing so, but nothing for
anti-alias even today.

I don't know about Edge and WebKit, the result of the test harness is from
contributors in Japan running them manually.

So to me, it is only Gecko that is hitting the anti-alias issue. Even for
Gecko, IIUC, setting a flag to run test properly is different from running
"modified versions of the tests". I'm not very familiar but AFAIU, Blink
uses some flags for CSS2 test suites too. Several factors such as UA
stylesheets, system settings, fonts in the platform, anti-aliases, etc. are
obstacles to run tests in a stable manner. Better to normalize them when
running tests.

Does this feel you good?

/koji
Received on Friday, 20 January 2017 10:19:59 UTC

This archive was generated by hypermail 2.3.1 : Friday, 20 January 2017 10:20:01 UTC