more problems with 'mount -t 9p' fs

[GoogleCodeExporter]

In #150 you have mentioned that you're using pyxp and wmiir to control wmii. 
Are you interested in maintaining/keeping the kernel-mounted fs interface? If 
not, 
I'll just start using wmiir or smth., otherwise I have 2 issues to report (and 
will continue testing, 'cause I like the mounted interface).

1. echo '_anything_' >> colrules
    and
    echo '__anything__' | wmiir write /colrules
    behave differently (the latter overwrites, the former appends).

2. This one looks like a race and might be difficult to reproduce, but this 
line triggers it for me:
   [[ writing in a terminal on tag '2' ]]
   $ sleep 1; echo '1' >> client/sel/tags; rm lbar/*
   [[ switch to some other tag (e.g. 3) before the shell wakes up ]]
   $ ls -l client/|grep sel
   ls: cannot access client/sel: No such file or directory
   d????????? ? ? ? ?                ? sel

Should I open separate 'Issues' next time?
Versions: wmii-hg2582, libixp-hg116, Arch Linux x86_64, Linux 2.6.30.9

Original issue reported on code.google.com by rogu...@gmail.com on 2 Nov 2009 at 12:11

[GoogleCodeExporter]

For issue 2.,
$ echo '1' >> client/sel/tags; rm lbar/*
must be executed when there's no selected client.

Original comment by rogu...@gmail.com on 2 Nov 2009 at 12:15

[GoogleCodeExporter]

I'm interested in supporting any complant np interface.

The first problem, I'll fix after I release 3.9. It might get a bit thorny, and 
I
don't want to delay the release that long.

The latter problem is not a wmii bug at all. Unlike 9p/Plan 9, the posix readdir
doesn't return full stat information, so ls has to stat each file that comes 
back in
the readdir list when you use the -l flag. In the interim between getting the 
list
and stating the files, the 'sel' file has disappeared. If anything, this is a 
ls(1)
but that should probably be reported to the GNU people (although I recommend 
just
switching to a BSD userland).

Original comment by maglion...@gmail.com on 2 Nov 2009 at 12:32

[GoogleCodeExporter]

Could you please confirm that the following is caused by a bug in GNU userland 
(because I wasn't sure why you were talking about 'ls')?

[[ open a terminal in tag '1' ]]
sleep 1; cat .wmii/fs/client/sel/ctl; cat .wmii/fs/client/sel/ctl
[[ switch to an empty tag '2' before the shell wakes up ]]
[[ switch back after a second, with the terminal open and focused ]]
cat .wmii/fs/client/sel/ctl
[[ cat: .wmii/fs/client/sel/ctl: No such file or directory ]]

...and it starts working again after some time (perhaps less than a minute)...

Original comment by rogu...@gmail.com on 2 Nov 2009 at 2:07

[GoogleCodeExporter]

Sorry, I misunderstood your point.

This looks very much to be a bug in v9fs. It seems to cache the error it gets 
walking
to /client/sel the first time. From the logs, it never even tries to find the 
file
after that first error, at not least until after some delay (which is clearly 
wrong).

Original comment by maglion...@gmail.com on 2 Nov 2009 at 3:34

[GoogleCodeExporter]

Should I inquire upstream, then? Or will you do it, since you're better 
equipped?

Original comment by rogu...@gmail.com on 2 Nov 2009 at 10:07

[GoogleCodeExporter]

I'll talk to the maintainers.

Original comment by maglion...@gmail.com on 13 Nov 2009 at 12:12

[GoogleCodeExporter]

Closing this as invalid since it's not a wmii bug.

Original comment by maglion...@gmail.com on 30 May 2010 at 7:52

• Changed state: Invalid