[hatari-devel] Latest Hatari HG version gets stuck sometimes

[npomarede at corp.free.fr]

On Mon, 6 Jul 2009, Thomas Huth wrote:

> On Sat, 04 Jul 2009 16:09:48 +0200
> Matthias Arndt <marndt at asmsoftware.de> wrote:
>
>> Am Samstag, den 04.07.2009, 01:15 +0300 schrieb Eero Tamminen:
>>> Hi,
>>>
>>> Nicolas, could this be related to your recent video.c changes?
>>>
>>> For some reason Hatari seems now to sometimes get stuck looping here
>>> (doesn't update its screen, uses 100% CPU etc):
>>> [...]
>>
>> I don't know if this is related to Nicolas's patch.
>> I can reproduce thsi lockup witha  totally blank screen (and even F12
>> not working anymore) with the freshly released "Just when" Demo by XIA
>> (http://files.dhs.nu/party/sommarhack2009/justwhen.zip)
>>
>> HG version from about 10mins ago, STE+4MB TOS 1.62de
>
> I also can't reproduce this problem. No matter whether I've got "Patch
> timer-d" enabled or not... I slowly get a really bad feeling about this
> bug... Could you please also try to uncomment the fprintf line right
> before the Int_AddRelativeInterrupt(... INTERRUPT_VIDEO_ENDLINE) line
> in the Video_InterruptHandler_EndLine() function, to see whether
> "LineTimerBCycle - LineCycles + nCyclesPerLine" behaves strange as soon
> as the problem occurs?
> By the way, which GCC version do you use?
>
> Thomas

Same for me, can't reproduce the 2 reported problems :(

Maybe a variable that would not be correctly initialized under some gcc 
version or some kind of side effect, I don't know.

What is sure is that the new video.c code with its "dynamic" adjustment of 
timer B/HBL interrupt will give this kind of "results" if something goes 
wrong in the internal handling of interrupts (if one timer B int is 
skipped/ignored, this can create this kind of loop).

I saw such bugs earlier with Int_AddAbsoluteInterrupt not reordering the 
int list before adding a new int, but this was already fixed in devel 
version.


Nicolas