Re: [hatari-devel] Hatari screen test

[Eero Tamminen <oak@xxxxxxxxxxxxxx>]

Hi,

On 5/29/20 2:52 PM, Thomas Huth wrote:
> On 29/05/2020 13.41, Eero Tamminen wrote:
> > On 5/29/20 1:12 PM, Thomas Huth wrote:
> > > I think I've found now a nice solution: The test program now checks the
> > > TOS version at the end of the screen, and if it's TOS 0.01 a.k.a
> > > Hatari's FakeTOS, it jumps to a routine that takes a screenshot via
> > > XBIOS #20 and then exits. Thus, no more command-FIFO polling in the
> > > test script, no more races, the test should still run on an original
> > > ST, and the test runs visibly faster in "make test".
> >
> > I agree, this looks the best solution.
> >
> > "Hatari doesn't provide just a hammer, it includes
> >   also a selection of screen^H^Hewdrivers, and
> >   a chainsaw."
>
> LOL, yes. But maybe we should still add some separate tests for the
> command fifo and the advanced debugger magic, just to make sure that
> there are no further regressions in that area in the future...

Conditional breakpoints and few other debugger
things are pretty well covered.

Chaining traced breakpoint isn't.  Maybe you
could add test-case for that using your
fullscreen test and the patch I mailed? :-)


Profiler has large amount of functionality which
isn't covered, but I think doing good tests for
that would be hard because profiling results
depend a lot on emulator cycle-accuracy which
changes over time.


Command FIFO support is not much code in itself,
it relies on code shared by debugger, socket API
etc, so test coverage can be increased by testing
those other things too.

Hconsole scripting is probably easiest way to
test them.


	- Eero