In a new Mageia 5 cinnamon only install, with updates installed during the
install, after reboot into the installed system, I tried to run "urpmi xsane".
It failed with the message ...
Use of uninitialized value in subroutine entry at /usr/lib/perl5/vendor_perl/5.20.1/x86_64/linux-thread-multi/Net/DBus/Binding/Connection.pm line 257
Running rpmdrake, it complained that it could not access Core Release, so
I edited the media, and removed the usb stick repositories, even though it
was still plugged in. Using rpmdrake, I was then able to install xsane.
Adding Thierry as leuhmanu posted on irc that it should be assigned to you,
not jquelin, even though :maint on irc shows jquelin.
Oops. Sorry for the double posting.
CC'ing Martin too as it might be related to recent fixes in bug 5690.
I've lowered the priority, as there is a work around that works, so I'll be
voting to go for release in todays council meeting.
Please attach (not paste) your /etc/urpmi/urpmi.cfg and your /root/drakx/report.bug.xz
Created attachment 6756 [details]
Created attachment 6757 [details]
Dave, looking at your log, this appears to be a UEFI install, with you having selected the first menu option in the grub2 menu, "Install from DVD", not the third menu option, "Install from USB". Is that correct?
I've just tried a fresh install like this, Cinnamon desktop only, and couldn't reproduce the problem - 'urpmi xsane' worked flawlessly. I left the USB stick plugged in all the time. The original urpmi.cfg entries (in report.bug.xz) look right to me.
It's on a dos style mbr drive. It has had a uefi install on it before, which
is why there is an efi/esp partition, but this install was done in non-uefi
mode. With my firmware, I have to go into the settings every time I want
to boot in uefi mode, which I only do when specifically testing uefi installs
which I now have on a third drive (sdc) which uses a gpt style partition
I suspect the usb stick is somehow being turned off when rebooting after
installing, as the led light on it goes out, so is not accessible. I'll
repeat the test later today after I catch up on other things. To recreate
the bug on your system, I suspect the stick will have to be unplugged.
Ah, fooled by the /boot/EFI entry in your fstab!
I've reproduced this (both with legacy and UEFI boot). Without the USB stick plugged in, urpmi first prints the message
Please insert the medium named 'Core Release'
then immediately gives the error message you reported. With the USB stick plugged in, it works as expected.
This only affects urpmi - drakrpm works as expected.
I tried adding the same media entries in my main mga5 installation, but didn't see this fault there. Maybe this hints at something missing in the base installation.
N.B. This is not related to the fixes for bug 5690, as here the USB stick is being treated as a cdrom, not a hard disk.
(In reply to Martin Whitaker from comment #11)
> I tried adding the same media entries in my main mga5 installation, but
> didn't see this fault there. Maybe this hints at something missing in the
> base installation.
No, this is because my main installation was installed from disk, so the perl-Hal-Cdroms package wasn't installed. Installing that package causes urpmi to behave in the same way as my test installation.
For me, this error message isn't fatal - urpmi continues to run. The problem I'm seeing is that urpmi is waiting for a CD to be inserted, and a USB stick won't do! If after inserting the USB stick, I insert a CD (any old CD will do), urpmi then resumes and installs the package from the USB stick.
Please urgently add a small note to the Errata page. Mageia 5 is out already.
Added to errata:
Dave, can you test whether the workarounds I describe work for you.
Wouldn't removing the CD/DVD installation sources from drakrpm-edit-media be a simpler workaround?
(In reply to Samuel VERSCHELDE from comment #15)
> Wouldn't removing the CD/DVD installation sources from drakrpm-edit-media be
> a simpler workaround?
Not for people with a slow internet connection, who want to install from media they've already downloaded. But I've updated the errata to add that as a third alternative workaround.
is it still valid ?