Re: [AD] update bitmap drawing after addition of transformations |
[ Thread Index |
Date Index
| More lists.liballeg.org/allegro-developers Archives
]
On 2010-08-15, Elias Pschernig <elias.pschernig@xxxxxxxxxx> wrote:
> On Mon, 2010-08-16 at 00:34 +1000, Peter Wang wrote:
> > On 2010-08-14, Elias Pschernig <elias.pschernig@xxxxxxxxxx> wrote:
> > >
> > > I think I solved all the issues now though and committed it. Now we
> > > should have all possible blitting combinations work again, for once.
> >
> > ex_scale is broken in 'wide' mode.
>
> Yes, there was a memory overwrite since clipping didn't work properly
> for negative scaling (that must have been the earlier black lines issue
> as well). I think it is fixed as of revision 13574.
Ah, sorry, I was in a hurry.
It's broken with hflip or vflip enabled, with negative y.
> > I think we *really* need to start automated testing.
> > One way would be to adapt ex_scale, ex_rotate, etc. to generate the
> > output image for some given parameter values, then compare the hash of
> > the output with known good hashes. This would save on storing the
> > good output images in SVN.
> >
>
> I agree. Maybe also make it even more automated. Move all such tests
> into "tests/" instead of "examples/" and print the result to stdout.
> Then we could make a script which simply runs all the tests/test_*.exe
> and checks if the output says "FAILED" or "PASSED" (like, once we have
> 100 such tests and running them by hand would get cumbersome...).
>
> Or maybe have some simple output format, like:
>
> test1: PASSED
> test2: FAILED
>
> Then a single executable could run more than one test and print one line
> for each.
Yes. I'm thinking that the tests should be defined in a config file,
one section for each set of test parameters and its expected hash.
Some tests could 'extend' other other tests to reduce redundancy.
I'll work on it tonight.
Peter