On Thu, 2020-10-15 at 08:21 +0200, Pierre Labastie wrote:
> On Sun, 2020-10-11 at 15:26 +0200, Thomas Seeling wrote:
> > Hallo,
> >
> >
> > I've been building 32- and 64-bit versions of LFS 10.0 sysV over
> > the
> > last weeks (with some more automation for my entertainment) and I
> > noticed a persistent problem in all builds.
> > I've been using jhalfs SVN 4205 and (yesterday) 4206.
> >
> > The command "userdel -r tester" from the last script of chapter 08
> > consistently fails with the error message "still in use by process
> > ####". This does not hurt very much so I basically commented out
> > the
> > "userdel" from the script and it continued with the chapter 09
> > scripts.
> > This may be some race condition although it's not quite clear to me
> > what
> > exactly happens since the previous commands finish successfully
> > (some
> > rm
> > commands).
> >
> > Does this happen to anyone else?
> >
>
> Bonjour Thomas,
>
> The last time I've used jhalfs was on september 28, and my log does
> not
> show that, although there is a warning:
> userdel: tester mail spool (/var/mail/tester) not found
>
> But I have not run the tests for that build, so maybe what you see is
> an indication that some test tun as user "tester" has left a zombie
> process that was not terminated properly. In that case, it is rather
> a
> problem with lfs.
>
> I'll try a test build with all the tests and let you.
>
>
There was no errors but... I added "ps -u tester" just before the
userdel command in the script. And a lot of processes were listed.
The reason is that the tester user's id is 1000, and the user running
jhalfs on host ("pierre") has also uid 1000. In chroot, the kernel has
no way to tell whether a process with uid 1000 belongs to tester or
pierre.
Now why userdel does not fail in this case for me is not clear...
Pierre
--
http://lists.linuxfromscratch.org/listinfo/alfs-discuss
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
