Re: [hatari-devel] One bug, one warning, one :)

[David Savinkoff <dsavnkff@xxxxxxxxx>]

Wednesday 14 November 2012 01:22 :
> Le 14/11/2012 02:16, David Savinkoff a écrit :
> > On Tuesday 13 November 2012 16:03, David Savinkoff wrote:
> >> Nov 13, 2012 03:40:31 PM, hatari-devel@xxxxxxxxxxxxxxxxxxx wrote:
> >>> As above things work, did you try also a Hatari built without
> >>> optimizations (-O0)?
> >>>
> >>> (Just to make very it's not some optimization bug in your very old
> >>> GCC version.)
> >>>
> >>> - Eero
> >>
> >> Solved.
> >>
> >> I explicitly used -O0 and the problem is gone.
> >>
> >> I did compile without optimizations, or so I thought,
> >> by not specifying optimizations to cmake and I had the
> >> problem.
> >>
> >> Sorry for the trouble.
> >>
> >> David
> >
> > This one is a mystery. -O0 did not make a difference, it was a
> > coincidence. After I explicitly compiled with -O0 things worked
> > 'too good', ... even when I compiled it again with optimizations
> > the same way I had before! So I rebooted my computer,
> > and none of it works!!!
> >
> > The code that suddenly started working, stopped working after
> > a reboot.
> >
> > Looks like something is not initialized properly on my computer
> > (maybe even hatari).
> >
> > David
>
> Just to be sure, did you try a "make clear + ./configure" at one point
> to be sure everything is re-compiled up to date. Or try a fresh pull
> from the hatari repo and try to build it ?
>
When I compiled with -O0, I did a fresh compile in a new directory.
When I'm comparing compilations I use separate directories.
>
> It could be a missing dependancy in your dev env, and a .c that was not
> properly compiled/linked, but that's very strange.
>
My opinion points toward hatari because I compile everything from
xorg to seamonkey to TORCS SDL mesa (-march=pentium3 -O3)
without problems (seamonkey takes 9 hrs and works)
>
> Nicolas
>
> You sure it's the same binary, run in the same way?
>
Yes.
>
> Are you e.g. running Hatari with full path (e.g. /usr/local/bin/hatari,
> as sometimes shell path handling can do funny things)?
Yes. I checked my .bash_history to be sure.
>
> > Looks like something is not initialized properly on my computer
> > (maybe even hatari).
>
> You could try running Hatari e.g. Valgrind to see whether that
> changes anything or does Valgrind report anything for you.
>
I'm familiar with text editors and compilers, less so with Valgrind.
I can try Valgrind if there is something specific.
>
>     - Eero

I just compiled hatari again with -O0 and new directories; I get
the same problem as before. The only glitch I've had in all of
this is that hatari started 'working' until the reboot; otherwise,
recent changesets continue to have the aforementioned problem.

I don't think my old compiler is the problem because I can point
to the last changeset that works, and compiler optimizations
don't make a difference.

David