[hatari-devel] Code-review, unit tests and bug-tracker

Deniz Turkoglu turkoglu.deniz at gmail.com
Wed May 11 22:57:25 CEST 2011


On Wed, May 11, 2011 at 10:19 PM, Eero Tamminen <eerot at users.berlios.de> wrote:
> Hi,
>
> On keskiviikko 11 toukokuu 2011, Deniz Turkoglu wrote:
>> > The screenshots & sound output are files saved by Hatari.  The control
>> > socket is a local unix socket, not a tcp/ip one.  I don't understand
>> > what you mean by "hashes"?
>>
>> We hash the images (md5) and we keep a list of possible passes and
>> fails (regressions). In screenshots
>
> Why not just do cmp?  The PNG screenshots are pretty small size wise,
> so it's not even a question of storage and you want the real image so that
> you can see where the difference is.
>
> (you could also use imagemagick to produce a "diff" of the image.)

We use hashes for a few reasons, we use tcp/ip and run these on a
variety of OSes and devices. The file io is quite slow, esp. on
devices, and when the number of tests grow, it becomes a bottleneck in
integration, hashing the image within the buffer in mem is faster, and
seems to work fine so far. The main problem with this approach was
with the fonts and rendering (web pages change a lot when a font
changes) which can be solved by ref tests [1] though I am not sure
if/how this can be applied to hatari.

[1] https://developer.mozilla.org/en/Creating_reftest-based_unit_tests

>
>> and audio files this can help us
>> to identify some things quickly (such as blank screenshot, empty audio
>> file, etc...).
>
>
>        - Eero
> _______________________________________________
> hatari-devel mailing list
> hatari-devel at lists.berlios.de
> https://lists.berlios.de/mailman/listinfo/hatari-devel
>



More information about the hatari-devel mailing list