Bug 19257 - X starts and stops quickly followed by a "there is a problem starting your graphical display" text message
Summary: X starts and stops quickly followed by a "there is a problem starting your gr...
Status: RESOLVED DUPLICATE of bug 19258
Alias: None
Product: Mageia
Classification: Unclassified
Component: Installer (show other bugs)
Version: Cauldron
Hardware: i586 Linux
Priority: Normal normal
Target Milestone: ---
Assignee: KDE maintainers
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2016-08-28 03:54 CEST by Renato Dali
Modified: 2016-12-17 20:19 CET (History)
1 user (show)

See Also:
Source RPM: sddm
CVE:
Status comment:


Attachments
Xorg log on first boot just after Mageia 6 sta 1 installation with Plasma 5. (5.42 KB, text/plain)
2016-08-29 05:45 CEST, Renato Dali
Details
Journal on first boot just after Mageia 6 sta 1 installation with Plasma 5. (201.39 KB, text/plain)
2016-08-29 05:46 CEST, Renato Dali
Details
Xorg log after running "startx" (just that) as a common user (not root). (28.96 KB, text/plain)
2016-08-29 05:50 CEST, Renato Dali
Details

Description Renato Dali 2016-08-28 03:54:13 CEST
Description of problem:

The graphical system starts but stops immediately. The traditional paragraph stating "Sorry, there was a problem starting your graphical display" is shown.

Version-Release number of selected component (if applicable):

Mageia 6 sta 1

How reproducible:

Tried only once to check compatibility.

Steps to Reproduce:
1. Boot a DVD with Mageia 6 sta 1 (i586, 32-bits, classical installer).
2. Install KDE/Plasma.
3. Reboot after installation.
4. Notice the above described behavior.
Comment 1 Renato Dali 2016-08-28 04:14:34 CEST
First of all, there is another bug 19216 (no graphical display, "good luck" error), which right now I cannot tell whether it is related to the present one.

Also, I remembered a similar discussion last year on bug 16065. By then, I was trying to install Mageia just with IceWM, so chose no DE.

The present situation mirrored more or less exactly those events:

1. Initially, there was a message on X.0.log informing drm failed to start, screens found but not usable => error: no screens found.

2. Just like back then, I installed xterm (with urpmi after a textmode login as root, from the DVD).

3. Just by installing xterm (or some dependency), "startx" managed to make X start (duh...) and it ended immediately because it wasn't told to serve anyone.

4. "startx xterm" afterwards successfully opens a terminal window -- Nouveau is apparently working well.

5. In that window, I typed "startkde" which gives a lot of output -- and specially worth of notice stating that "a CPU with the SSE2 instruction is required". For Qt5, I suppose (which was what I was originally trying to find).

For information, the hardware used is the same of last year, a Sempron 2300+ with enough RAM (it's a 686 CPU with PAE and SSE, but not SSE2).

Please advise me if any other info is needed.
Comment 2 Marja Van Waes 2016-08-28 09:05:38 CEST
Do you mind attaching your matching
   Xorg log
and journal.txt, that is the result of running, as root:
   journalctl -ab > journal.txt

Do you only have this problem using SDDM and Plasma5, or also using e.g. LightDM and XFCE?

(As you know, you can switch to a different instlled DM with drakdm, even in a text tty.)

Keywords: (none) => NEEDINFO
CC: (none) => marja11

Comment 3 Renato Dali 2016-08-28 16:54:05 CEST
I will reinstall it and post the info later.

Also, please notice I'm opting for the Nouveau driver, which I deem better supported than the proposed proprietary driver.
Comment 4 Renato Dali 2016-08-29 00:37:54 CEST
I just installed and will try to post the logs you asked, Marja.

But I think I can answer your question immediately.

> Do you only have this problem using SDDM and Plasma5, or also using e.g. LightDM and XFCE?

Sorry, but I actually forgot about Plasm 5 replacing KDM with SDDM and this link (for instance), shows SDDM does require SSE2:

https://github.com/sddm/sddm/issues/609

So, in simple terms, I cannot even get a login screen with Plasma 5 because of SDDM.

After reporting this bug I went back to the machine and installed Mageia 6 sta 1 with Xfce as DE. Everything went fine, so I think the only reasonable explanation is that it uses another DM (e.g. LightDM) and it worked well.

In other words, this is all by design and this bug possibly can be closed as invalid, too. I don't know what is the current Mageia view on the issue, but I'm curious.
Comment 5 Renato Dali 2016-08-29 05:45:34 CEST
Created attachment 8374 [details]
Xorg log on first boot just after Mageia 6 sta 1 installation with Plasma 5.
Comment 6 Renato Dali 2016-08-29 05:46:42 CEST
Created attachment 8375 [details]
Journal on first boot just after Mageia 6 sta 1 installation with Plasma 5.
Comment 7 Renato Dali 2016-08-29 05:50:44 CEST
Created attachment 8376 [details]
Xorg log after running "startx" (just that) as a common user (not root).
Comment 8 Renato Dali 2016-08-29 06:05:00 CEST
I uploaded the files you asked obtained just after reboot (after the installation).

Then I ran startx and the Xorg log grew with video modes etc. Of course, just running startx makes X start and then stop immediately.

I then wanted to change to LightDM, as you asked. So I installed it and changed from SDDM to it using drakdm (thanks for the hint, BTW).

Before rebooting, I thought about running startx with an application which wouldn't require Qt5. I tried "startx audacity" and it worked well considering there was no window manager.

Dolphin surprisingly started -- but without icons and having no means to close it.

Next, I rebooted and saw the login screen came up as expected; I could even login normally, but again there was no window manager. The Mageia welcome application started, but I couldn't do anything afterwards.

Ctrl-Alt-Del caused a ksmserver error dialog, but X stopped anyway.

Hope that helps, but feel free to ask for more info if needed.
Comment 9 Marja Van Waes 2016-08-29 13:35:05 CEST
@ Renato

Thanks for all the feedback and attachments :-)

(In reply to Renato Dali from comment #4)
> 
> Sorry, but I actually forgot about Plasm 5 replacing KDM with SDDM and this
> link (for instance), shows SDDM does require SSE2:
> 
> https://github.com/sddm/sddm/issues/609
> 
> So, in simple terms, I cannot even get a login screen with Plasma 5 because
> of SDDM.
> 

So bug 19278 is a duplicate of this one, correct?
I'm confused, because that one is about qtdeclarative5 and this one about sddm 

Leaving it to the KDE team to sort this out :-)

Keywords: NEEDINFO => (none)
See Also: (none) => https://bugs.mageia.org/show_bug.cgi?id=19258, https://github.com/sddm/sddm/issues/609
Assignee: bugsquad => kde
Source RPM: (none) => sddm

Comment 10 Renato Dali 2016-08-29 14:36:09 CEST
Well, when I created both I had different intentions on each one.

This one was too similar to the situation we faced last year with Mageia 5, so I was even imagining something was missing in Plasma installation. This time I observed a graphical splash screen was shown for a very brief moment (the starry sky with the Mageia 5 logo) prior to the "Sorry, there has been... " text message.

If I observed it more accurately I'd describe the situation a little differently (I did edit the bug title now).

I'm not sure they're duplicates, and in the initial Xorg logs there's no mention to sddm; we suspect now the cause is sddm stopping because SSE2 is missing in my CPU -- but not even the journal mentions that (though there are lots of messages about sddm).

Maybe both have the same cause, and if it is something by design then there's really very little that can be done... maybe both can have the same solution (e.g. a warning to not install Plasma on such CPus or a greyed out Plasma option in the classical installer, or still just an Errata about such uncommon situation). Having the same cause does not mean they're duplicates, but if the same solution solves both, that'd be convenient. If both bugs are seen as a Qt5 dependency problem, I suppose we could tag both as duplicates, even if the effects are different. Again, problem solved anyway.

As an aside, I seem to remember having tested SDDM once (for my personal needs, not in connection with Mageia testing) and that it worked (maybe it didn't require Qt5 at that time).

As the test with LightDM shows, it is possible to solve this bug (e.g. by replacing SDDM with LightDM) and for the other bug about Plasma 5 to remain (the "solution" for that would be installing and using Xfce instead of Plasma).

Thanks, Marja.

Summary: X fails to start after installation ("there is a problem starting your graphical display" message) => X starts and stops quickly followed by a "there is a problem starting your graphical display" text message

Comment 11 Renato Dali 2016-08-30 03:12:18 CEST
Sorry, my memory is uncertain, I'm not sure if I tested Slim and not actually SDDM back then.

Also there are other bugs about SDDM (like bug 19216). On a first sight, it appears to be a different issue (though the results are the same).
Comment 12 Renato Dali 2016-09-02 15:38:30 CEST
Hi, everyone,

After reading this post on a Mageia forum:

https://forums.mageia.org/en/viewtopic.php?f=4&t=10947

... and reflecting on the issue, I cam e to the following conclusions:

- this bug actually has two issues (thus should be broken in two):
        1. SDDM related ones and
        2. Plasma related ones.
- SDDM might not be working, and the logs I provided might help (or not);
- SDDM itself is suggested by KDE, making it indirectly a Plasma related issue;
- Mageia may decide to replace it with LightDM, and that won't help solve the Plasma dependency on SSE2.

Because of all that, I chose to do the following things:
- suggest that the aspect of Plasma (or SDDM) dependency be ignored from the present bug, because as Marja noticed, it already is mentioned in bug 19258 (so it's not useful to open a new one), itself tagged as UPSTREAM and
- maintain the present bug open in the hope that its data helps someone sort the SDDM issue, which happens in other bugs (please see comment #1).

On the course of a more detailed analysis, one might find the remaining part of the present bug is a duplicate of one of the other SDDM-related ones (but not bug 19258!) and as such mark the present one as a DUPLICATE. It might be considered UPSTREAM, too, but I'm not certain about that at the present point in time.

Thank you all.
Comment 13 Renato Dali 2016-12-17 20:19:51 CET
After some research, I concluded both SDDM and Plasma depend on the SSE2 instruction (which is unavailable in the processor on which this bug appeared).

So, despite the two aspects of the bug (discussed above), both were found to depend on the same cause, which is the same one in bug 19258.

So I guess we can consider the present bug as a duplicate of bug 19258.

*** This bug has been marked as a duplicate of bug 19258 ***

Status: NEW => RESOLVED
Resolution: (none) => DUPLICATE


Note You need to log in before you can comment on or make changes to this bug.