[Pan-users] Re: Re: ANN: Pan 0.97 "Atoz and Tanda"

[Duncan]

walt <address@hidden> posted
address@hidden, excerpted below, on  Sat, 13 May 2006 15:28:17
-0700:

> You seem to be running a 'gentoo-unstable' system, am I right?  AFAICT,
> the latest 'stable' AMD-64 compiler is 3.4.5.

Actually, gcc-4.0.3 and 4.1.0 are both still masked for testing, that is
not even ~arch yet, AFAIK. 3.4.6 is the latest unmasked version.

The problem is that gcc 3.x on amd64 is still very unoptimized.  The amd64
arch was a bolt-on half-way thru the series, and until the rewrite of 4.0
that was able to take amd64 into account directly, it remained quite
apparent that amd64 was a bolt-on.  There have been three major
improvements to amd64 support on gcc, the first with gcc-3.2 (I
believe) when it was first supported, the second with 3.3, when one could
actually do a -march=k8, and now 4.1, the first real effective release of
the 4.x series.  4.0 completed the rewrite, but didn't actually deliver
on the promise of that rewrite.  4.1 delivers, and despite its masked
status, seems far more stable here than any 3.x version EVER was.

Anyway, for that reason, gcc 4.1, while it might be a minor improvement
for x86, is a *HUGE* improvement for amd64.

> Thanks to your observations I just built 0.97 on my ~x86 (gentoo
> unstable) box just out of curiosity.  I've noticed some distinct
> differences in pan's behavior between the two machines, so I can't help
> wondering if you might also notice a difference if you built pan on a
> 'gentoo-stable' AMD64 machine.
> 
> Do you have such a machine to play with?  (I wouldn't dream of being
> without both kinds ;o)  I actually have three gentoo boxes, and one of
> them has both 'stable' and 'unstable' gentoo in separate partitions --
> it's trivial to do (as I'm sure you already know.)

I only keep one machine, and altho I do have a second image of everything
critical on my system, it's a snapshot image taken periodically when I
think the system is stable enough to warrant it, and functions as my
backup if anything on ~arch or from my selected hard-masked unmasks, or in
reality, as likely my fat-fingering of something, kills something such
that I find I can't boot the regular working version.

While I don't doubt what you are saying, I'm relatively sure it's mostly
due to the different gtk+ versions you'd be running in stable vs.
unstable.  Also, all of the problems so far have been general bugs,
experienced by others as well, or in a couple cases due to my customizing
-- the colors thing I reported back with 0.93 IIRC, is such an example --
it would have happened given my color scheme (light text on a darker
background, the reverse of most folks, it would seem) regardless of what
system versions of whatever I was running.

The problem here, BTW, as addressed in another thread, appears to be
because Charles always closes using the application close function, while
I (and others experiencing unsaved settings) normally close apps using the
titlebar close button -- which basically results in the window manager
signalling the app to close, rather than using the app's internal close
function.  The internal close function worked, since Charles tested it. 
The window manager signaling didn't, as he'd missed a bit of code and
never realized it since that's not the way he functions and he never
closed the app that way.  =8^P

Anyway, there's a patch for that that I'll apply shortly, now that it has
been traced.  Meanwhile, I'll be careful to close PAN using the internal
close function, and will hopefully not see the issue again.  =8^)

All that said, I'd be interested in reading about your differences, just
because I'm the curious type.  =8^)  Maybe they /will/ cause me to revert.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman