Re: [Evms-devel] evms_activate : Still a bug ?

[Steve D. <st...@us...>]

kev...@us... wrote on 01/31/2005 08:09:40 AM:

> Hi Michel,
>
> On Saturday 29 January 2005 3:03 am, Michel Bouissou wrote:
> > Le Vendredi 28 Janvier 2005 17:26, Steve Dobbelstein a =E9crit :
> > > Please try the attached patch and see if it cleans up the empty
> > > directories.
> > >
> > > (See attached file: discover.c.patch)
> >
> > I'm afraid it makes no difference. Still the initial evms_activate
creates
> > empty directories, and calling an EVMS UI immediately removes them.=

>
> I've been unable to reproduce this so far, so I'll need to get some
> additional
> info from you.

Looks like I missed a lot of action on this thread over the weekend. :)=


Question for Michel:  Did you update your initrd with the patched and
rebuilt libevms.so?  The patch is supposed to do the cleanup the first =
time
the Engine is run.  If your initrd hasn't been updated then it won't cl=
ean
up the empty directories.

Kevin, I was able to "reproduce" the problem by creating an extra direc=
tory
in /dev/evms and putting a dummy block dev node in it.  the first run o=
f
the Engine will empty out the node but will not clean up the empty
directory.  A subsequent run of the Engine will clean up the empty
directory.

My guess is that somehow the compatibility dev node is created (don't k=
now
how) but later (don't know when) it doesn't really it doesn't exist bec=
ause
the EVMS volume was discovered.  At the end of evms_commit_changes() it=

calls cleanup_dev_evms_tree() to delete all the undiscovered nodes.  At=
 the
end of evms_commit_changes() it calls cleanup_dev_evms_tree() which del=
etes
the dev node for the compatibility volume, but does not cleanup the emp=
ty
directory.  That patch is supposed to fix the cleanup of the empty
directory.

Steve D.=