flex-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alex Harui <aha...@adobe.com>
Subject Re: Mustella Test Repairs
Date Mon, 27 Aug 2012 17:05:33 GMT

On 8/27/12 7:35 AM, "Carlos Rovira" <carlos.rovira@codeoscopic.com> wrote:

> Hi Peter,
> yes, I'm fully synchronized with develop branch.
> The differences in number of test reported is already scaring me
> since. I saw this before in other folders and is nothing good...
> I think you are ok with getting some method to work over since I think
> the team working right now on this is not already using the same
> rules. For example I saw Alex said to Jose to not run -createImages,
> but I think you and I consesuated to use it...we should something
> unified and all people working on it should get over that rules.
I'm not sure we can write an algorithm for everyone to follow, but my
general flow is this:

1) Run a folder of tests.
2) Fix the non-CompareBitmap failures by modifying the tests.
3) If I see CompareBitmap failures, use ImageDiffAIR to examine the
    a) If the only differences appear the be that the baseline did not use
embedded fonts, then cut a new baseline
    b) Otherwise, try to figure out why there are differences and determine
if the test needs fixing or if it is ok to cut a new baseline.

The last part is a loosely defined.  But I would assume that, other than
embedded font errors, the baselines checked in are accurate and differences
are important.

If there are just a few differences in the antialiasing, you can try to get
around that by bumping up the colorVariance options on CompareBitmap, but
don't just stick in high numbers to make the test pass.

It is great to see all of your efforts and I know we are in a hurry to get
tests passing, but we don't want to skip over important tests to reach that

Alex Harui
Flex SDK Team
Adobe Systems, Inc.

View raw message