Re: [Rosegarden-devel] rosegarden-4-1.2.3rc2 now available for testing

[Pedro Lopez-C. <ped...@gm...>]

On Saturday 11 February 2006 08:17, Chris Cannam wrote:
> Pedro:
> > On Friday 10 February 2006 20:12, Chris Cannam wrote:
> > > Did my subsequent emails announcing that rc2 had
> > > already been superseded by rc3 not make it out?
> >
> > Seems that Sourceforge is losing messages again.
>
> Turned out it had bounced it because the mail server
> I was using was in a blacklist.  Funny, the same
> server had worked for my previous messages
> yesterday.  It'll be very annoying if that continues.
>
> > There is a regression on sysex messages recording.
>
> Bugger.  Well, I'm going to be away until Monday
> probably - if you should happen to find a fix...

Fixed and committed to CVS HEAD, now.

Long report:

There were some QFile.open() calls inside assert() statements, in
sound/MappedEvent.cpp. The bug was there for the last 2 years and 5 months.

You are asking yourself: how this could pass all over the 1.0 release, 
developers' reviews and many users' hands without a single report?

First, because if you compile Rosegarden with debug activated, the code 
works right. Many developers, and contributors like me, usually compile RG 
with debug flags activated to help in finding bugs. This turned out to be the 
opposite: you needed to switch off debugging to find the bug. Nasty.

Second, because the code was working well in Rosegarden 1.0, even on 
compilations without debug activated, because the NDEBUG symbol was never 
defined in our old build system, using autotools. Since bksys/scons has 
changed this situation, the bug has bitten us. I'm sure that someone would 
make a good joke about candid souls shooting themselves on their feet.

And there was at least one related bug report, by b.ohnsorg at freenet.de, 
many thanks to him, again.

Regards,
Pedro