Re: [Gap-dev-discuss] Cynthiune: Further Evidence on Old Problem

[Sebastian Reitenbach]

Hi,

didn't got around earlier to try, see below...

On 06/10/15 03:42, Richard Stonehouse wrote:
> Hi Sebastian,
>
> I've tested playing MP3s on 32-bit and it seems to be fine (as it was
> before on 32-bit).
>
> However, I have another problem, which seems to be 64-bit only. I did
> see it in the previous version of Cynthiune (SVN rev 2647) but assumed
> it was just another feature of the MP3 bug - but it's still there and
> perhaps more consistent.
>
> The following steps reproduce it consistently for me on 64-bit, and do
> not reproduce it on 32-bit:
>
> 1. Put some files in the Playlist (any type).
>
> 2. Play one or more files (any type).
>
> 3. Select all the files in the Playlist, using the mouse.
Here I clicked on the first in the list, scrolled to the last
pressed ctrl, and then clicked the last one so that all got highlighted.

> 4. Click on the 'Remove Selection' button.

I'm unable to reproduce.
When I do that, then all songs disappear, and Cynthiune stops playing.

Tried on OpenBSD amd64. I compile with
clang, and make use of libobjc2. Otherwise, using all the latest
releases.

What versions of the libraries are you running, and compiler using?
Do you can compile all objc stuff with debugging information, and
run Cynthiune in gdb and try getting a backtrace when it hangs?

I don't really think so, but maybe its a backend problem, do you can 
reproduce it with a different audio output backend?

Sebastian

> Cynthiune hangs. It doesn't look like a loop - no CPU is being
> consumed.  If you double-click on the title bar of the upper Cynthiune
> window to shrink the window to title bar only, then double-click again
> to re-expand it, the upper window reappears with no content. It isn't
> possible to miniaturise the window.
>
> It doesn't go into this state if you omit step 2, playing a file.
>
> It doesn't if, at step 3, you select only some of the files in the
> playlist. However it does occur even if there is just one file in the
> playlist, and you try to select and then remove it.
>
> I also hit another problem when, while messing about trying to reproduce
> the above, clicking on the 'Next' button crashed Cynthiune. It produced
> a lot of output to the terminal - the first bit of this was lost but the
> attached is as much was retrievable. I haven't found a way of
> reproducing this again.
>
> On Tue, Jun 09, 2015 at 09:03:03AM +0200, Sebastian Reitenbach wrote:
> > Hi,
> >
> > Am 6/9/2015 um 2:52 AM schrieb Richard Stonehouse:
> > > Hi Sebastian,
> > >
> > > On Mon, Jun 08, 2015 at 07:15:58PM +0200, Sebastian Reitenbach wrote:
> > >
> > >
> > > > For me on OpenBSD, trying to play any MP3 segfaulted Cynthiune, on
> > > > amd64 and on i386.
> > > > I exchanged a memcpy with a memmove, and that segfaulting now was gone
> > > > on both platforms.
> > > > Afterward, playing that test file, sounds exactly the same for me on
> > > > both. It's in SVN now.
> > > >
> > > > Do you can check, whether that might have fixed the noise problem
> > > > for you too?
> > > > svn co svn://svn.savannah.nongnu.org/gap/trunk/user-apps/Cynthiune
> > >
> > > Yes thanks, that fixes it for me on 64-bit. I've tried several MP3s
> > > and they all play fine. Great!
> >
> > thanks for feedback.
> >
> > > I'll build the 32-bit version and try it again tomorrow - but that
> > > used to work anyway, it would be just a regression test.
> >
> > At least it works for me, but please let us know. I guess it's time
> > then for a new bugfix release.
> >
> > > > Otherwise, what audio output backend do you use? I use OpenBSD specific
> > > > sndiod. If you still get that noise problem, do you can try another
> > > > audio backend?
> > >
> > > I'm now using the Pulse Audio server but the same problem occurred
> > > with output direct to alsa.
> >
> > Yeah, I was also more expecting the input drive to be the problem,
> > than those getting the music out.
> > > > While there I fixed two other minor compilation warnings also in SVN.
> > > >
> > > > Sebastian
> > > >
> > > > > Also attached are the /proc/cpuinfo from both machines.