27_io/filebuf-members.exe

Ken Moffat ken at kenmoffat.uklinux.net
Wed May 5 17:52:34 PDT 2004


(Anybody who doesn't understand the title can ignore this!)

 Ever since I started playing with gcc-3.3.3, I've had fun with this
member of the testsuite, culminating in a patch to revert the changes.
The main problem was the build hanging, but only from a recent (LFS)
host.  Now, all of my recent builds have used ash for /bin/sh (it seemed
a good idea at the time, and exposed bashisms).

  I'd already noticed that with ash the expect testsuite all passes,
whereas with bash there are several failures, but this weekend I started
to suspect that ash might be the problem.  At the moment, I've got a ppc
build of 5.1-pre2 running - it's now in ch6 autotools - only this time I
symlinked /bin/sh to bash.  Chapters 5 and 6 gcc have completed, so I'm
starting to think the shell might have a bearing on this.  I know others
have seen this - could ash be the common factor, or am I off "not
waving, but drowning" again ?

 The subsidiary problem (filebuf-members.exe is still there, and will
prevent a clean umount) remains.  At the moment I see no point in
guessing whether fuser will identify the process to kill, or if I'll
need to build lsof.

Ken
-- 
 das eine Mal als Tragödie, das andere Mal als Farce




More information about the lfs-dev mailing list