Running Live installation from desktop icon. Set up: Laptop with hard disk (currently a dual boot mageia/microsoft, booted on one Mageia Live USB stick, with intention to do a Live install to another USB stick. Selecting target device there are two problems, may be related: 1) In the dropdown to select target device, the stick is not listed by name, but as a long number. This can definitely be improved to show product name instead like IsoDumper do. (The spinning disk is fortunately displayed as name, at least in this set-up). 2) I selected the USB, proceeding with "use existing partition" the dialogue suggest mount points for / and /home on sda = ON THE *DISK* !! It also list one partition on the stick (sdc), but no mount point selected. THIS IS BACKWARDS! IMO, user can easily be fooled, as he did choose a particular target drive in the dropdown, then installer suggest to put all on the other! Also it use naming on form of "sda", and nowhere a description of which disk "sda" is. It should a) default to populate the selected drive, here only one partition, so olny mount point to be defaulted is /. b) By default not put anything on a non selected drive. Display the names of the drives it presented to the user in the drop down. 3) Only when selecting custom partitioning user is reminded to back up data. 4) When Live started copying, trying to close its window is not graceful. First click, user is asked to really close, if user proceed window is still up, process continues. On second attempt window disappear, but process is not killed. Version-Release number of selected component (if applicable): Mageia 8 RC Live xfce i586 with persistence, updated just now. Point 2: Installing on wrong drive make me set this critical. It do prompt user about accepting partitions on sda, but hey this is supposed to be the newbie way, and it in goes against what user selected in the previous dialogue.
Rising for point 2 (and possibly 1)
Assignee: bugsquad => mageiatoolsPriority: Normal => High
Thank you for the report. I have changed for the moment the architecture to 32-bit, because we are currently seeing ISO problems just for that; this is another. If you know the same problem arises on 64-bit, please report that, and adjust it back. If you have now a running Mageia on the machine, please post the output of: $ inxi -MSDpxx to summarise its characteristics.
CC: (none) => lewyssmithHardware: All => i586
I realised today that running stick was actually beta2 based, but running with persistence and all updates so should not matter in this case.
Not sure about that for an installation. I would have thought it is better to have the real latest ISO. CC'ing Martin for his view on that. My request in comment 2 stands.
CC: (none) => mageia
Yea, will retest when i get to it, reloaded sticks now so cant do now.
Multiple bugs on one ticket is hard to manage... Items 1, 2, and 3 apply to both classical and Live installers. Item 4 is not a partitioning issue. Item 2 is the installer working as designed. There has been discussion about this before, although I can't find it now. Although I largely agree with your comments, I know how complex the partitioning wizard code is already, and how hard it is to verify it works for all use cases. We don't have any regression tests for the installer (or indeed, for any of the drakx tools), more's the pity.
Running now on 8 Live xfce i586 second internal RC: Same problems. "Working as designed", yes if not that is a compiler or hardware issue ;) Working as intended, i hope not... Instead of creating 6 bug reports i put them here listed for now, no time to fix i think to release this time anyway, and as you say many are old ones, from traditional installer and diskdrake. Adding: 5) ... from pint 3: if user in that dialogue about backing up data select to abort, instead of returning user to previous screen the dialogue is just greyed and in front pops the partitioner up... 6) the old problem with last installer question about where to put the boot loaders; sda is preselected. Like for point 2, that could be acceptable if it not were the user already in the beginning did select another drive... 7) When aborted, i.e when i aborted it due to missing grub2 Bug 28153, it did not unmount the LVM vg. 8) I know that becase now when retrying i see the LVM vg listed in the dropdown of drives... A bit weird, but i stil dont list it as abug, as it do help in such cases...) 9) After simply unplugging - replugging that stick so vg is "unmounted" i try again: in partitioning i select that encrypted LVM PP, enter password and select that new tab. Then i see the F2FS and swap partitions now both are recognised by diskdrake as ext4... 10) deleted that "ext4" (former swap) and created a new: lvcreate failed as that name already exist. So it was not properly deleted. Have to remove whole LVM. But i cant remove LVM as there is no way to remove the nondeleted nonvisible LVs... unplug/replug and this time not unlocking LVM, i can remove it. If it had not been encrypted there would probably be need for user to use another tool to remove it. 11) If user is in the dialog to create partition, and realise he need to switch to expert mode to select i.e F2FS type, he close that dialogue and... instead of partitioner he is back at installer start dialogue. ...bedtime...
Whiteboard: (none) => 8RCSummary: 8RC Live installation partitioning issues => Some Live installer issues
... had just to see last try complete successfully, and it did :) Now i was so tired i forgot to correct localisation of the bootloader on that dialogue... Fortunately there is a back button on next dialogue and it had not a that stage actually written to sda :) --- Noticed from log that it executes urpme --auto --auto-orphans (removed 22 pkgs) Which brings up that old discussion... --- Also in log, maybe more important, or is it to be expected: During the installer unmounts all partitions there are *many* buffer I/O errors on dm-0 and dm-3 so i wonder if all got flushed out to the device correctly...?
Traget system seem to work splendid :) Well yes simple install to a machine with single storage and not fuzzing with partitioning it works without noticeable issues. We should not tear Live installer apart before mga8 release.
Severity: critical => majorPriority: High => Normal
Created attachment 12226 [details] For comment 2: inxi-MSDpxx (while installer was installing)
Created attachment 12227 [details] For Comment 8: buffer errors while installer unmounts
Just to clear this up: - does all the commentary apply equally to 64-bit systems? - does the presence of a persistant USB partition come into it? - Are we talking just about the choice 'use existing partitions'? - if you did Custom partitioning, would these 'problems' occur? It seems to me that anyone who deals with LVMs, encryptation, installing to a USB, should jolly well know what they are doing, and use 'Custom'. We could say in the installation instructions that if you have such complicated setups, do custom partioning to avoid any ambiguities. One can go on forever inventing obscure scenarios that would defeat the disc partitioner. From comment 6, Martin: "Although I largely agree with your comments" are there any particular ones you think the installer really should cover? This bug could become an enhancement request for just those. We overlook that Mageia has been around for years, and the disc partitioning has evolved a lot, for EFI especially. Clearly it works effectively for (almost) everybody; except exotic cases. We should not get carried away.
(In reply to Lewis Smith from comment #12) > Just to clear this up: > - does all the commentary apply equally to 64-bit systems? Probably, i use to get into similar problems with plain classic installer and diskdrake itself, you can see from my many years of bugs in it, some solved some not. If i remember to select correct partition, never undo or change anything, remembered to install grub2 before, etc, it went well, as I reported. > - does the presence of a persistant USB partition come into it? Probably not. And BTW i verified user files, installed programs etc followed thropugh to target system, so OK. > - Are we talking just about the choice 'use existing partitions'? Only my comment0, then i went for Custom. > - if you did Custom partitioning, would these 'problems' occur? Most my hair pulling came from Custom yes > It seems to me that anyone who deals with LVMs, encryptation, installing to > a USB, should jolly well know what they are doing, and use 'Custom'. We > could say in the installation instructions that if you have such complicated > setups, do custom partioning to avoid any ambiguities. Naturally. How would LVM etc else be set up if not using Custom mode? - Except for the wrinkles it is a pleasure to use :) > One can go on forever inventing obscure scenarios that would defeat the disc > partitioner. What obscure scenario did i invent? > From comment 6, Martin: > "Although I largely agree with your comments" > are there any particular ones you think the installer really should cover? Show disk by name, obey given answer, accept abort, list all supported filesystem types? Basically very simple: do not mess it up for the user who knows what he is doing... Note i do not ask for *any* new function. > This bug could become an enhancement request for just those. > We overlook that Mageia has been around for years, and the disc partitioning > has evolved a lot, for EFI especially. Clearly it works effectively for > (almost) everybody; except exotic cases. We should not get carried away. Lets accept the old bugs in diskdrake for now. Problem is that there have been new problems *added*, Like: User gets to select what drive to install to, but installer do not obey it. (defaulting partitions on sda nevertheless, and boot on sda) (And also the problem it shows a number instead drive product name, like IsoDumper can do) In plain diskdrake (at least on mga7) there is a abort button in the dialogue to create partition, not in Live installer as i ran it, so i had to create wrong partition to get out of it, select expert mode to list all, then change it. Combine that with old bugs that diskdrake sometimes screws up such changes especially within LVM.
I just want to add I did not hunt for faults, I just attempted to make a universally bootable stick = bootable on both classic ans EFI, and of course encrypted and with swap, wich implies LUKS and LVM, and by basing on Live easily get wide hardware compatibility. (Another method is to use a Live system with custom partitioning and/or swap file, both have different advantages) I just noted down my findings here when I stumbled on them along that path. Lets let this rest for now. I will probably make a few enhancement requests later.
Assigning to mageiatools; CC'ing Thierry who has most to do with drakxtools SRPM.
Hardware: i586 => AllComponent: RPM Packages => InstallerCC: lewyssmith => thierry.vignaudSource RPM: (none) => drakxtools-18.40-1.mga8.src.rpm
(In reply to Morgan Leijström from comment #7) > "Working as designed", yes if not that is a compiler or hardware issue ;) > Working as intended, i hope not... "Working as designed" means working as the designer intended. Originally "Use free space" also used all disks. That was changed in 2015 - see bug 16055. But "Use existing partitions" still has the original behaviour. > Problem is that there have been new problems *added*, Like: Both of those problems exist in 7.1.