bug in make install

[Sam Halliday]

hi there,

i know reporting bugs in glibc is real live flaimbait, but hey, i got
stung by this one and at least it will be archived now...

make install seems to be trying to do too much. i recently upgraded an
i586 optimised glibc runngin on GNU/Linux (i also tested an i386
optimised glibc) to an i686 optimised glibc, and 'make install' bailed
on a segfault. what was happening was that 'make install' installed
/lib/libc-2.3.2.so first, then did a whole load of funky shit, and was
only then going to install /lib/libpthread-2.3.2.so...

but it never got that far, because after installing /lib/libc-2.3.2.so,
make install went on to try and use a program linked to
/lib/libpthread-2.3.2.so, which of course was an old version (and in
this case, incompatible), resulting in a segfault and bailing make. from
then on any app linked to pthread was dud. luckily cp is NOT linked to
pthread and i recovered my system by running this...

make install
if [ ! $? = 0 ]
then
 /bin/cp -f linuxthreads/libpthread.so /lib/libpthread-0.10.so
 make install
 echo 'GLIBCBUG encountered, and tried to workaround.'
fi

but, my point is this... i think make install is trying to do too much.
all it should do is check if everythign has been compiled (if not, run
standard 'make' first), and then copy over all the new binaries, with
all the glibc workarounds for overwriting a running c library etc etc. I
really dont think it should be trying to make a whole load of locales
and regenerating a lot of files, and running a lot of unnecessary
programs which use new and old libs all at once. at the very least, the
new lib files should be copied over all at the same time, explicitly,
from a top-level makefile.

cheers,
Sam
-- 
Disobedience:  The silver lining to the cloud of servitude.
-- Ambrose Bierce