nnpfs_lock on FreeBSD 5.4 with Arla-0.40

[Marcus Crestani]

Hi,

trying to build my current XEmacs workspace on FreeBSD always leads to
nnpfs_lock, probably caused by arla.

Here are the details:

entropy[31] uname -a
FreeBSD entropy 5.4-RELEASE FreeBSD 5.4-RELEASE #0: Sun May  8 10:21:06   
UTC 2005     root at harlow.cse.buffalo.edu:/usr/obj/usr/src/sys/GENERIC     
i386

Arla-0.40 (./configure --without-krb4).

My workspace compiles and links fine, then XEmacs is building a dump file
(launching XEmacs, writing out the current heap to a file, load file
upon later launches instead of re-initializing the heap again to
improve startup speed). Afterwards some updated lisp source files are
recompiled.

Reproducably somewhen during these steps the build process hangs
(though not always at the same point). This time it locks here:
Removing old or spurious .elcs in directory tree
`/afs/informatik.uni-tuebingen.de/home/crestani/src/xemacs/newgc-
xemacs-21.5/src/../lisp'...

Hitting CTRL-T gives this information:
load: 0.43   cmd: xemacs 983 [nnpfs_lock] 1.88u 0.15s 2% 16320k

The output of `ps -o wchan -auxo state | grep xemacs' shows:
nnpfs_ crestani   983  0.0  6.4 17264 16304  p0  DL+   3:31PM   
0:02.04   ./xemacs -no-pac DL+

Compiling my workspace on other systems (like MacOSX and Solaris) on
the same afs works fine without causing this deadlock. This problem
only occurs on FreeBSD. The lock also occured on FreeBSD 5.2.1 with an
older arla (sorry, I don't know what version).

I put the tarball of my workspace online, in case you want to try to
reproduce the failure: http://www.crestani.de/newgc-xemacs-21.5.tgz

It is configured with `./configure --enable-newgc --enable-mule'. To
cause XEmacs to link again and resume the build process which then
later causes the nnpfs_lock, you may use `touch src/gc.c' prior to
`make'. Then wait and see...

If you need more information, please contact me.
And please CC me, I'm not subscribed to this list.

Thank you,
--
Marcus