flex-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alex Harui <aha...@adobe.com>
Subject Re: Jenkins build is still unstable: flex-sdk_test #432
Date Fri, 18 Dec 2015 16:02:58 GMT


On 12/18/15, 3:05 AM, "Mihai Chira" <mihai.chira@gmail.com> wrote:

>> Also note that the compiler shouldn't care as much about the order in
>>the
>> library path as the compile time for the definitions in the SWCs.  So if
>> class Foo was compiled into the as3 SWC and then later into the flex
>>SWC,
>> the definition in the flex SWC should always win regardless of order.
>> Maybe the SWCs were built in a different order on your computer.
>
>I know that, BUT: I tried locally with
>flexunit-4.3.0-20140410-as3_4.12.0.swc and it fails with exactly the
>same errors as on the server. Then I tried with
>flexunit-4.3.0-20140410-flex_4.12.0.swc and it passes (except for a
>debug assert in
>FLEX_34625_Tests.test_focus_skin_with_zero_focus_thickness() - but
>different one, anyway, from the one currently failing on the server
>[1].
>
>So I don't know why, but it seems very likely that in the end our CI
>server is using the as3 version of flexunit instead of the flex
>version. I do see that they're included two times each (so 4 times in
>total) on each run, so maybe the compiler might have a limit of how
>many times it's overriding a class in a more recent swc over a
>previous one? Don't know.
>
>Regardless, how can we ensure that we use the flex version only? I
>couldn't really tell where I need to tweak ant scripts or other code.

I don't know FlexUnit very well.  Is it normal to have both the as3 and
flex version in the library-path so their definitions compete or is it
normal to only have one or the other?  I'm trying to understand whether
the current situation where both are in the library-path is the default
and our tests have to compensate or whether the tests are misconfigured
and only one or the other should be used.

-Alex

Mime
View raw message