Created attachment 9854 [details] screenshot I tried to do a fresh install of Mageia 7 using Mageia-Cauldron-netinstall-nonfree-x86_64.iso, and it keeps failing when trying to download RPMs, see the screenshot. I tried with mageia.unige.ch and mirror.math.princeton.edu, and I got the same error message (with a different list of RPMs). I also tried with kernel.org, but it's unable to read the synthesis file. Is there no way to make things more reliable? Something like atomic sync (all or nothing instead of partial sync)?
(In reply to Frédéric Buclin from comment #0) > Created attachment 9854 [details] > screenshot The bad rpms message is bogus, I don't know what's going on, only that this was already reported in bug 21886 > > I tried to do a fresh install of Mageia 7 using > Mageia-Cauldron-netinstall-nonfree-x86_64.iso, and it keeps failing when > trying to download RPMs, see the screenshot. I tried with mageia.unige.ch That mirror is outdated, because it synced with distrib-coffee http://mirrors.mageia.org/status > and mirror.math.princeton.edu, That's a good mirror > and I got the same error message (with a > different list of RPMs). > > I also tried with kernel.org, but it's unable to read the synthesis file. > > Is there no way to make things more reliable? Something like atomic sync > (all or nothing instead of partial sync)? I'll keep this report assigned to sysadmin team and infrastructure, in case the problem is indeed there. Keeping bug 21886 open, too, though.
See Also: (none) => https://bugs.mageia.org/show_bug.cgi?id=21886CC: (none) => marja11
Re: bug#21886 , my tests were done with the princeton mirror which is first-tier and has been updating regularly.
CC: (none) => ftg
As reported on bug 21886, there is something wrong with the squashfs file. $ unsquashfs -l mdkinst.sqfs Parallel unsquashfs: Using 4 processors Read on filesystem failed because EOF read_uids_guids: failed to read id index table FATAL ERROR:failed to uid/gid table With the m6 version of mdkinst.sqfs ... $ unsquashfs -l mdkinst.sqfs Parallel unsquashfs: Using 4 processors 3806 inodes (5019 blocks) to write squashfs-root squashfs-root/bin squashfs-root/etc squashfs-root/etc/brltty <snip rest> Is the file system where the squashfs output being created full, and the build ignoring the error?
CC: (none) => davidwhodgins
In /usr/lib64/drakx-installer-stage2/misc/mdkinst_stage2_tool it has ... mksquashfs $LIVE_DIR $COMPRESSED_IMAGE -all-root -noappend >/dev/null || { echo "mksquashfs failed"; exit 1; } chmod 755 $COMPRESSED_IMAGE echo foo > $STAGE2_DIR/.room if [ -s $STAGE2_DIR/.room ]; then rm -f $STAGE2_DIR/.room [ -n "$CLEAN" ] && rm -rf $LIVE_DIR else echo "not enough space" rm -f $COMPRESSED_IMAGE exit 1 fi The use of the .room file looks to me like it's trying to ensure there is room, which implies the mksquashfs does not set a return code on running out of space. I'm thinking the mdkinst.sqfs is being truncated, but when it fails there is still room for a three byte file.
commit e9dd64112cbb6582a931546d9f451a5f91c84601 Author: Pascal Terjan <pterjan@...> Date: Thu Jan 4 04:16:43 2018 +0000 Really exit on error Incorrect code noticed in a comment on mga#22266 --- Commit Link: http://gitweb.mageia.org/software/drakx/commit/?id=e9dd64112cbb6582a931546d9f451a5f91c84601
CC: (none) => pterjan
CC: (none) => thierry.vignaud
CC: (none) => thomas.bigot
*** Bug 22746 has been marked as a duplicate of this bug. ***
No more error reported, solution submitted. Thus I close this report.
Resolution: (none) => FIXEDStatus: NEW => RESOLVEDCC: (none) => yves.brungard_mageia