When starting up xfontsel in a en_US.UTF-8 setup, it takes about 3 seconds
during which the CPU load is 100% and eventually loads with the warning:
"Warning: Missing charsets in String to FontSet conversion" output on the
console. I'm guessing during the load it's doing something with the font
server. Starting it without UTF-8 (LANG=en_US xfontsel) brings it up
immediately without warnings.
I'm actually hunting the same thing (slow high-load startup, warnings, incorrect
non-ASCII fonts) for the XEmacs package in FE devel, but found that xfontsel
exhibits the same behaviour and it's a lot smaller thing which could be easier
to use in tracking the problem. Ideas? FWIW, this is on x86_64.
I don't recall seeing any reports like this so far, and there's not a lot
of info to really diagnose the problem without details and reproduceability.
Can you strace/ltrace/gdb it and try to narrow it down? That would potentially
I'd also recommend filing it in X.Org bugzilla, in case there are
corelations there as well.
Yep, I know there's not too much info here yet, but I was hoping it'd be easy to
Anyway, I've run strace and ltrace when starting up xfontsel with and without
UTF-8 $LANG, results at http://cachalot.mine.nu/tmp/xfontsel/ . Those results
don't tell me much, but it appears that the strace outputs (which are pretty
diffable) differ significantly in the amount of -EAGAINs for some read()s.
Also, I've verified that it's (unsurprisingly?) xfs that is consuming the CPU
time when xfontsel is launched in UTF-8 $LANG mode. I'll require some
hand-holding with gdb'ing the process if it's needed.
Will report to x.org bugzilla if this is still and issue and not reproduced
Does the problem occur in FC5t3 still? If so, can you file a report to
X.Org and report back?
I'm running latest Rawhide updated from FC5t2 (haven't installed "vanilla"
FC5t3), and the problem is exactly the same as before.
What product/version/component would you file this against in
Filed at https://bugs.freedesktop.org/show_bug.cgi?id=6138
I can no longer reproduce this with FC5 final, assuming fixed.