[cairo] pixman "Simplify clipping rule" problems

Kalle Vahlman kalle.vahlman at gmail.com
Sun Jun 28 05:02:14 PDT 2009


2009/6/28 Siarhei Siamashka <siarhei.siamashka at gmail.com>:
> On Sunday 28 June 2009, Kalle Vahlman wrote:
>> 2009/6/28 Siarhei Siamashka <siarhei.siamashka at gmail.com>
>>
>> > The one thing that worries me a bit is that apparently (almost) nobody
>> > else is using 'scaling-test' program. As limited as it is, it could have
>> > prevented a handful of broken commits already. And you can't rely on me
>> > monitoring this mailing list forever, effectively working as a trained
>> > monkey who just runs the test program occasionally and complains when
>> > something gets obviously broken.
>>
>> I'm sure nobody is asking you to do that?-)
>
> You probably misunderstood me. I'm not whining or complaining.

Ok, so you sound sour even without complaining. I'll remember that.
(implicating that people treat you as a monkey sounds like somethings
not right to me, but as said maybe it's just a style issue).

>> That would be nice, but at least it seems that scaling-test as it is
>> can't be used in such target.
>
> It can be easily used as such target. It only needs some return code depending
> on whether it was successful or not (to automatically check it in the scripts
> or makefiles) and fix for the big endian systems to provide exactly the same
> deterministic result as on little endian systems.

So not as-is, but with fixes :)

>> A quick look in it tells me to just run it, but when I do it says that
>> it fails. On x86 with and without all optimizations. I tried with
>> master, 0.15.14, 0.15.12 and 0.14.0. Of these the two first ones say
>> that it fails and the two others crash.
[snip]
> As you can see, every case of scaling-test failure is associated with one
> pixman bug or another.

Right. So I guess the biggest issue was that there was no existing
baseline for the test so testing older versions was not a good idea.

-- 
Kalle Vahlman, zuh at iki.fi
Powered by http://movial.com
Interesting stuff at http://sandbox.movial.com
See also http://syslog.movial.fi


More information about the cairo mailing list