Ali Bahrami <Ali.Bahrami at Sun.COM> writes:

> Hello,
>
>    I've been working on getting GNU emacs 22.1 integrated into
> Solaris, and have run into a problem that I need some help to
> solve.
>
> Emacs can be built in three different ways:
>
>       1) Pure tty application (runs in a terminal emulator)
>       2) Using Athena (libXaw) widgets, which are an ancient
>               toolkit from MIT X11 days, but which works well
>               enough with emacs.
>       3) With GTK
>
> There is an issue with libXaw that means that a binary
> built with (2) won't run under OpenSolaris:
>
>       http://www.opensolaris.org/jive/thread.jspa?threadID=65995&tstart=0
>
> That's going to be fixed at some point, but until then, a binary that
> links to it won't work on OpenSolaris.
>
> I have experimented with statically linking libXaw into emacs, and although
> that does work, doing so opens a whole raft of legal and technical issues:
>
>       legal - I'd have to talk with lawyers for an extended period,
>               after which they'll probably say no anyway, which is
>               a waste of time I don't have for something that is a
>               short term hack.
>
>       technical - Gatekeepers frown on crufty code bloat, and along
>               with agreeing with them, I fear them... :-)
>
> What this means that as things stand today, the GNU emacs that appears in
> OpenSolaris will only be able to run as a tty application for some unknown 
> period
> of time until the libXaw situation gets resolved. This may well be what 
> happens,
> but I'd like to do better...
>
> Which brings me to the GTK version. Possibly, I could ship a GTK emacs
> that would work on OpenSolaris, and which looks better in the bargain.
> However, I have a problem to solve first.
>
> I can build emacs 22.1 with GTK, and the 32-bit version does work. However,
> it spews a series of assertions to stderr that worry me (the stderr output
> is attached). (I've seen this with emacs 22.2 also, don't know about any other
> versions).
>
> It's possible (likely even) that this is a simple error in emacs, although I
> don't understand GTK well enough to fix that efficiently. If you're deep in
> GTK, and have a moment to download and build emacs 22.1, and could send me 
> patches
> for that, you'd be doing a service to the whole community:
>
>       http://ftp.gnu.org/pub/gnu/emacs/emacs-22.1.tar.gz
>       <unpack, and then>
>       % ./configure --with-gtk
>       % make
>       % src/emacs             <<< SHould start quietly
>
> There's more to this than just debugging emacs, and I'd appreciate
> any information you might be able to provide. I've been corresponding with
> Rich Lowe, who has been building and using GTK emacs on snv_91 without
> seeing these errors. We've been trying to find out how our builds differ,
> but without much luck.
>
> I sent him a copy of my emacs-gtk binary (that produces the assertions).
> When he ran it on his system, it ran cleanly (no asserts). I've tried
> snv_95, and Open Solaris, and both produce them. When he ran snv_93 under
> Virtual Box, he was also able to produce them.
>
> This is either a difference in the GTK bits between the Solaris versions,
> or some sort of environmental issue. It's almost as if there's a 'debug mode'
> enabled for GTK. Google brings up nothing other than the GTK_DEBUG environment
> variable, which unfortunately does not seem to have a "be quiet" option.
>

With some help from Glynn Foster, I'm pretty sure I now see what the
difference is.  In Nevada, I'm not using the Nimbus theme, in the
Indiana VM I was (it was the default, and I hardly ever use that VM)

If I switch my Nevada desktop to use that theme, I see these
assertions, with the theme I was using previously (clearlooks) I do
not.

Narrowing it down, it appears that the Nimbus GTK theme/style/whatever
(the "controls" bit in the prefs->appearences dialog), is what causes
this to start happening, switching away from it causes it to stop.

That would seem to suggest this isn't emacs' doing at all.

-- Rich

Reply via email to