BUGZID:12336
Optimizations and cleanups that piled up during the cvs-svn transition while
trying to make a KDE uniconf plugin go more smoothly: wvqtstring.cc had some
leftover log messages. unimountgen::whichmount didn't actually work right.
uniconfd didn't know its version number to give to WvDaemon. Several "i()."
changed to "i->" in unilistgen. UniListGen now passes set() along to *all*
its child generators, not just the first one, so make sure you make the
right ones read-only. (This is critical because it's often *not* the first
one that you want to set, and set() has no return value for latency
reasons.) Made UniFileSystemGen actually more-or-less work, and replaced a
lot of its plain-C file accesses with more WvStreams-friendly bits. Wrote a
new UniAutoMountGen, which, combined with the filesystemgen and inigen, can
read ~/.kde/share/config. UniConfDaemon doesn't yield() *quite* so often,
because it was horribly slow. Fixed bugs in the UniIniGen parser when you
have corrupt lines in the file (eg. unclosed braces), so it can try to
recover without losing the remainder of the file. UniIniGen is *not*
initially dirty, because that's weird. Reduced priority of "file hasn't
changed" log message. Link libwvbase.so with libxplc.so, since it depends
on it. Add a rm_rf() function to wvfileutils to go with mkdirp(). Make
WvStream use outbuf.optgettable() instead of outbuf.used() for flushing its
outbuf, which goes *much* faster when you have a huge outbuf. WvDaemon
should set "die" signals back to SIG_DFL after it receives them the first
time, so CTRL-C CTRL-C "really" kills it even if your daemon is broken.
WvPrint's job.t.cc had a valgrind memory leak.
git-svn-id: file:///home/apenwarr/alumnit-svn/public/trunk@10135 6619c942-59fa-0310-bf43-a07aaa81d9ac