Bug 23488

At the beginning of the installation the mount command must have been found and used. I don't understand why it isn't found when the packages get installed.

CC: (none) => isobuild, mageiatools, marja11

Package scripts are run from the system being installed, the package calling mount needs a proper dependency to have it installed before .

CC: (none) => pterjan

(In reply to Pascal Terjan from comment #2)
> Package scripts are run from the system being installed, the package calling
> mount needs a proper dependency to have it installed before .

Thanks for the explanation :-)

The error appeared just before makedev was installed. Does that mean that makedev needs 

  Requires(pre): util-linux

Or can't the culprit have been installed with the dependency missing?



CC'ing the kernel maintainers, because makedev is a "System/Kernel and hardware" package. 

Btw, what is the new upstream for makedev? We still have 

  URL:            http://cvs.mandriva.com/cgi-bin/cvsweb.cgi/soft/makedev/

in the spec file :-/

Source RPM: (none) => makedev??

(In reply to Marja Van Waes from comment #3)
> The error appeared just before makedev was installed. Does that mean that
> makedev needs 
> 
>   Requires(pre): util-linux

Looking at the SPEC file, it actually calls mount in %post, not %pre (not sure why the error appears first), so it'd be Requires(post), so you had the right idea.

The other problem with this kind of issue, especially with basesystem packages is dependency loops, so hopefully that fix would be enough, but if there's a dependency loop, it might not be until the loop is broken.

(In reply to David Walser from comment #4)
> (In reply to Marja Van Waes from comment #3)
> > The error appeared just before makedev was installed. Does that mean that
> > makedev needs 
> > 
> >   Requires(pre): util-linux
> 
> Looking at the SPEC file, it actually calls mount in %post, not %pre (not
> sure why the error appears first), so it'd be Requires(post), so you had the
> right idea.
> 
> The other problem with this kind of issue, especially with basesystem
> packages is dependency loops, so hopefully that fix would be enough, but if
> there's a dependency loop, it might not be until the loop is broken.

Thanks, David.

Assigning to the basesystem maintainers (or is makedev, as "System/Kernel and hardware" package, maintained by the kernel maintainers?)

Component: Installer => RPM Packages
Source RPM: makedev?? => makedev
Assignee: bugsquad => basesystem
CC: (none) => kernel

A better fix would be to not call mount (I wonder if I wrote that code, it seems familiar /o\)

The script tries to find out if it should create devices or not in an ugly way:

if ! df /dev | grep -q /dev$ || ! mount | grep -q ' /dev type devtmpfs '; then

It would probably be better to do something like:

if ! df /dev | grep -q '^devtmpfs .* /dev$' ; then

But this seems a bit wrong to create them in all other cases, like if there is a bind mount.

Maybe it could just be something like if [ ! -c /dev/null ].

This may be OT, but nettle is mentioned, and nettle-devel "does not verify."  The SHA256 does not match.

CC: (none) => laidlaws

Or maybe it's time to stop installing it?
Eg, FC stopped installing MAKEDEV a while ago...
chroots users could still manuall installing it.
(or learn to use mount --bind /dev/ /some/chroot/dev -o ro)

I've stopped maintaining both the code & the conf a looong time ago...
(Also we lost the VC sources...)

CC: (none) => thierry.vignaud

One of the users in iurt :)

(In reply to Pascal Terjan from comment #9)
> One of the users in iurt :)

Mock doesn't use makedev, but as it's pulled in and used by basesystem-minimal, it's installed in the chroot/container.

CC: (none) => ngompa13

Further to Comment 7, the problem with nettle-devel was different, and is now fixed. Vide Bug 23842.