Installer repeadedly failed to load from desktop icon. Version-Release number of selected component (if applicable): MGA5 RC live DVD 64 bit installer (sorry, can't recall name on binary) How reproducible: Attempt to start installer from desktop icon. (from "live" account) Steps to Reproduce: 1.click icon 2.Window (empty) opens for a second or so 3.Window closes with no errors or any text seen. Opened konsole, su root, opened icon properties to find binary, and ran listed binary with full path to see any error msgs... Woked flawlessly run as root, so most likely a permissions issue. Machine details: Lenovo X300 core2 duo (gen1?) 1.4 GHz (ULV) intel GM3000 onboard gfx 8GB DDR2 ram 256gb Samsung SSD 512G WD SSHD in cd tray. I hope this is isolated incident, marked major as new folks would likely simply give up when the installer blew rasperries at them. Reproducible: Steps to Reproduce:
Target Milestone: --- => Mageia 5
Can you look at the icon's properties, copy the command line to console and run it as a normal user?
Source RPM: MGA5 RC KDE installer live dvd (rpm unknown) => draklive-install
(In reply to Samuel VERSCHELDE from comment #1) > Can you look at the icon's properties, copy the command line to console and > run it as a normal user? It was done already > Opened konsole, su root, opened icon properties to find binary, and ran > listed binary with full path to see any error msgs... > Woked flawlessly run as root, so most likely a permissions issue. I'll see if I can reproduce.
CC: (none) => ennael1, lmenut, neoclust, tmb
(In reply to Rémi Verschelde from comment #2) > (In reply to Samuel VERSCHELDE from comment #1) > > Can you look at the icon's properties, copy the command line to console and > > run it as a normal user? > > It was done already > was done as root.
Does running (as root) journalctl -f 2>&1 | tee journalctl.txt show anything interesting if you first start that in a konsole and then click the icon (so starting install as user)? if so, or if you don't know, then please attach journalctl.txt
CC: (none) => marja11
OK, this is weird--- went to duplicate it so I could better document it (same machine, same install media) and it worked perfectly fine. I was previously overwriting a Mint 17.1 test install, but it doesn't seem like it should matter vs. what it was doing. Tried it multiple times. Tried to follow exact procedure used, brought up wireless first etc.[shrugs] The only thing I can see that was different was the partitioning has changed.
Adding NEEDINFO keyword until can be reproduced.
Keywords: (none) => NEEDINFO
Created attachment 6623 [details] Extract from journal showing error messages from draklive-install I saw this once before (using the GNOME 64-bit Live DVD), but could never reproduce it. However it has just occurred again, using the round 4 GNOME 64-bit Live DVD. The attached file contains the error messages from draklive-install extracted from the journal.
CC: (none) => mageia
I should have added that when I tried again, it worked without a hitch.
@ pterjan Should I have assigned to you?
Keywords: NEEDINFO => (none)CC: (none) => pterjan, thierry.vignaudAssignee: bugsquad => tmbSummary: MGA5 RC KDE livedvd, 64 bit, installer fails to launch from icon => MGA5 RC KDE livedvd, 64 bit, installer fails to launch from icon (INTERNAL ERROR: unknown device sda1)
I'll reinstall Mint17.1 using the same partition scheme this weekend and see if that was the issue.
I hope I haven't hijacked Greg's report with a different problem - I thought the first time I saw this it failed shortly after clicking on the icon, but whilst trying to trace the fault further, all the failures I've seen have come after clicking Next in the first window that appears (the one that just tells you it's going to run the installer). Anyway, the failures I'm seeing are caused by something early in the install process causing the sda partitions to be temporarily removed from /dev. Occasionally they don't reappear in time for the next stage in the installer to find them when it goes looking for them. In the failing cases, entries in the journal look like this: draklive-install[5717]: test_for_bad_drives(/dev/sda on sector #62) kernel: sda: sda1 sda2 sda3 sda4 sda5 sda6 sda7 sda8 sda9 sda10 sda11 sda12 kernel: sda: sda1 sda2 sda3 sda4 sda5 sda6 sda7 sda8 sda9 sda10 sda11 sda12 draklive-install[5717]: found a gpt partition table on /dev/sda at sector 0 draklive-install[5717]: error: INTERNAL ERROR: unknown device sda1 <trace info as in previously attached file> kernel: sda: sda1 sda2 sda3 sda4 sda5 sda6 sda7 sda8 sda9 sda10 sda11 sda12 (sometimes without the "INTERNAL ERROR" message) Non-failing cases look like this: draklive-install[5319]: test_for_bad_drives(/dev/sda on sector #62) kernel: sda: sda1 sda2 sda3 sda4 sda5 sda6 sda7 sda8 sda9 sda10 sda11 sda12 kernel: sda: sda1 sda2 sda3 sda4 sda5 sda6 sda7 sda8 sda9 sda10 sda11 sda12 kernel: sda: sda1 sda2 sda3 sda4 sda5 sda6 sda7 sda8 sda9 sda10 sda11 sda12 draklive-install[5319]: found a gpt partition table on /dev/sda at sector 0 (sometimes with only one of the "kernel:" lines, not three).
Hit this with round 4 of Gnome x64 Live DVD on real EFI hardware booted from DVD, installing directly from the boot menu (because Live mode failed). Agree with comment 11 that this arises after clicking 'Next' on the Install Wizard start dialogue. INTERNAL ERROR: unknown device sda1 MDK::Common::Various::internal error() called from /usr/lib/libDrakX/devices.pm:131 devices::entry() called from /usr/lib/libDrakX/devices.pm:146 devices::make() called from /usr/lib/libDrakX/fs/type.pm:254 fs::type::call_blkid() called from /usr/lib/libDrakX/fs/type.pm:262 fs::type::type_subpart_from_magic() /usr/lib/libDrakX/fsedit.pm:281 fsedit::get_hds() called from /usr/lib/libDrakX/fs/any.pm:14 fs::any::get_hds() called from /usr/sbin/draklive-install:134 (eval)() called from /usr/sbin/draklive-install:134 main::init_hds() called from /usr/sbin/draklive-install:64 main::install_live() called from /usr/sbin/draklive-install:32 Will upload the journal.
CC: (none) => lewyssmith
Created attachment 6626 [details] Full journal installing M5 Final round 4 Live Gnome DVD re Comment 12 Sorry it is so big, I never know where to trim it.
That may be a different bug, or different manifestation. I never got to the screen to hit "next": A blank window popped up for a second.
Do you want me create a new bug report for the issue Lewis and I are seeing?
Right now it looks as if is possibly related, so no. I reinstalled Mint 17.1 again, recreated partitions etc and reinstalled from the same image/medi as before but still cannot duplicate this end, its truly flake..
commit 62645b8ddc107a13151f2a30297aefda4c16f790 Author: Thomas Backlund <tmb@...> Date: Thu May 28 10:21:30 2015 +0200 ignore special 'Packard_Bell' and 'Push_Button_Reset' partitions (mga#15999) --- Commit Link: http://gitweb.mageia.org/software/drakx/commit/?id=62645b8ddc107a13151f2a30297aefda4c16f790
(In reply to Martin Whitaker from comment #11) > Anyway, the failures I'm seeing are caused by something early in the install > process causing the sda partitions to be temporarily removed from /dev. > Occasionally they don't reappear in time for the next stage in the installer > to find them when it goes looking for them. In the failing cases, entries in > the journal look like this: We might be missing a "udevadm settle" call...
Created attachment 6657 [details] run udevadm settle before reading partition table Does this patch helps? (to be applied in /usr/lib/libDrakX)
Keywords: (none) => PATCH
@Olivier: why draklive-install neither load disk/ide|scsi|... category nor run udevadm settle?
Created attachment 6658 [details] drakx: wait for udev to create nodes (1/2) Another try
Created attachment 6659 [details] draklive-install: probe hds & wait for udev to create nodes (2/2)
@Olivier, Pascal: WDYT? I think that's the saner way
Sorry, neither of these attempts fixes it :-( What seems to be triggering the fault is something that opens the raw device (i.e. calls open on /dev/sda). The first culprit I've found is test_for_bad_drives. If I comment out the call to this function, I move the point of failure. I'm trying to trace exactly where it next fails to confirm my diagnosis.
Created attachment 6663 [details] drakx: run udevadm settle after actions that trigger device reload I wasn't able to track down exactly what was triggering the device reload (apart from test_for_bad_drives), but the attached patch seems to fix the issue for me - I've run draklive-install::init_hds 200 times without seeing a failure, whereas without the patch I'm seeing a failure rate of between 1 in 5 and 1 in 20 (it is highly variable).
commit f3c5ce01ee71ba2b210ea4a5d4447232ae1d2fe9 Author: Thomas Backlund <tmb@...> Date: Fri May 29 10:56:19 2015 +0200 run udevadm settle after actions that trigger device reload (Martin Whitaker, mga#15999) --- Commit Link: http://gitweb.mageia.org/software/drakx/commit/?id=f3c5ce01ee71ba2b210ea4a5d4447232ae1d2fe9
Should be fixed in next live ISO
Status: NEW => RESOLVEDResolution: (none) => FIXED