« first day (3112 days earlier)      last day (1833 days later) » 

Kiwy
Kiwy
8:15 AM
@RuiFRibeiro Thanks for the edit : https://unix.stackexchange.com/a/108527/53092
:D
 
Rui F Ribeiro
Rui F Ribeiro
8:32 AM
@Kiwy lol....doing it in semi-automatic mode, did not even notice it was you. back to work, enjoy your day.
 
Kiwy
Kiwy
@RuiFRibeiro you too have a nice enjoyable day
 
 
2 hours later…
Rui F Ribeiro
Rui F Ribeiro
10:14 AM
@terdon used Mandrake, do not miss it ;)
@terdon I left some Mandrake servers in central bank of Mozambique...they are not probably using them anymore (I hope)
 
terdon
terdon
10:59 AM
@RuiFRibeiro Ha! :)
 
Kiwy
Kiwy
Apparently Faheem definitely had a problem with that legacy Nvidia driver
 
 
3 hours later…
Pandya
Pandya
2:02 PM
During sudo apt upgrade, I see:
> apt-listchanges: News
---------------------

wpa (2:2.4-1+deb9u3) stretch-security; urgency=high

This release backports changes to help mitigate EAP-pwd security issues
(CVE-2019-9495, CVE-2019-9497, CVE-2019-9498, CVE-2019-9499).
Unfortunately, the complete fix heavily depends on the code added after
wpa 2.4 release, so porting it is not practical. Consider using strong
passwords to prevent dictionary attacks.

For more information about the issues and their impace, see the Debian
Security Advisory DSA-4430-1.
What is this ?
 
Jesse_b
Jesse_b
2:44 PM
A security vulnerability documented:
https://nvd.nist.gov/vuln/detail/CVE-2019-9495
https://nvd.nist.gov/vuln/detail/CVE-2019-9497
https://nvd.nist.gov/vuln/detail/CVE-2019-9498
https://nvd.nist.gov/vuln/detail/CVE-2019-9499
https://w1.fi/security/2019-2/eap-pwd-side-channel-attack.txt
https://w1.fi/security/2019-4/eap-pwd-missing-commit-validation.txt
 
Pandya
Pandya
@Jesse_b Should I consider to do anything on my system? I am not familiar with this.
seen first time.
 
Jesse_b
Jesse_b
Without knowing much about it I would follow the instruction to "Consider using strong
passwords to prevent dictionary attacks. "
 
Pandya
Pandya
ok
 
Jesse_b
Jesse_b
3:04 PM
HAH
-1
Q: How to become a self-taught professional?

Teka i need to kick start a carrier as a software developer. so what are my options and what are some demanding and trending languages .. (im also little bit familiar wit java and c++ .sort of you know .. all these stuff i learned all by self so now i need kinda some guidance)

linux
"Teach me how to teach myself"
 
sourcejedi
sourcejedi
This is the second time someone's posted an unrelated answer on unix.stackexchange.com/questions/384364/… . I'd quite like to get my own answer back above -1 votes :-). I've tried to improve my answer a bit. Anyone have any further comments about my answer? IMO the situation is quite unambiguous, but people are missing the point. (It doesn't help that the question seems to have been abandoned by the OP).
 
Stephen Kitt
Stephen Kitt
3:21 PM
@sourcejedi installing Debian 9 does add an online source to sources.list, except when it can’t connect to the network
 
sourcejedi
sourcejedi
Thanks, I will amend accordingly.
@StephenKitt oh, that would be a possible reason why there wasn't an online source in this question. It doesn't explain the claim in the TecMint article, that the same thing happens when you use a netinstall cd. You can't install from a netinstall cd if you can't connect to the network :-). I guess it is more helpful to mention your point for this question, but maybe leave it open that there are other possible reasons for it.
 
Stephen Kitt
Stephen Kitt
@sourcejedi I often find it helpful to take TecMint articles with a bowlful of salt...
 
sourcejedi
sourcejedi
3:39 PM
I already started putting in caveats :-). A third-party source, it doesn't show full proof, and I haven't rigorously tested what situations cause it myself.
 
Stephen Kitt
Stephen Kitt
@sourcejedi but yes, it’s likely the fix is your answer, minus the “requires” part. There’s not much point in speculating why the OP ended up in that situation... I just re-ran a Debian 9 netinst and ended up with a correct sources.list.
 
Faheem Mitha
Faheem Mitha
Hi folks. Just upgraded to buster. I thought it would be easy - boy, was I wrong. Being at it since 1 am this morning, on and off.
 
sourcejedi
sourcejedi
The TecMint claim seemed to match the first time I saw it linked. (OP posted it the link as "I found the solution", and it seemed like it was a netinstall they had used. unix.stackexchange.com/questions/444531/…)
Faheem's back! Hooray!
 
Faheem Mitha
Faheem Mitha
Finally managed to log in.
sddm was still throwing errors last I checked. And I don't seem to have a KDE desktop for now. But at least I'm back up.
Got lightdm to log me in. Apparently my .Xauthority was owned by root, so that wasn't working right at first, either.
And Nvidia was a freaking nightmare. Jesus.
It looks like one needs to switch glx alternatives manually, because nothing did it for me. I.e. with 
update-glx --config nvidia
Unless I'm missing something.
I guess I should write an upgrade report. Though, the mood I'm in right now, it would have a lot of swear words.
And systemd apparently really wants something called plymouth installed, though I'd never heard of it before. Without it, it kept crashing...
@derobert Around?
 
Stephen Kitt
Stephen Kitt
@FaheemMitha that’s rather unfortunate! plymouth is supposed to be optional, my Buster systems don’t have it :-/
 
Faheem Mitha
Faheem Mitha
3:50 PM
I'm going to give sddm another try now. I kind of need my desktop back.
 
sourcejedi
sourcejedi
Ow. plymouth isn't supposed to be required unless you need to enter a disk password or something, I think.
@FaheemMitha Good luck!
 
Faheem Mitha
Faheem Mitha
@StephenKitt Yes, I would have thought so, but it seemed very definitely to want it. Though the messages I saw on the net suggested that when it said it wanted plymouth, it really had a problem with /etc/fstab. I decided to take the requirement literally, and once I installed it, that particular crash stopped.
 
overexchange
overexchange
Unable to delete inner_folder 
export environ='/home/user/a'
mkdir -p a a@1 a@2 a@1/inner_folder # inner_folder is a source code build workspace
rm -rf ${environ}.*/inner_folder # does not delete a@1/inner_folder
pwd # /home/user
 
Faheem Mitha
Faheem Mitha
But it was hanging a lot at boot too. And when I did Ctrl-C, it would put me in emergency mode.
You folks will be glad to know that I'm learning to hate systemd too.
Does anyone know if I need to do something special to get a desktop with lightdm? Currently it's a blank screen.
Oh, and what retard switched su to no longer include necessary paths by default? I'd like to give him/her a piece of my mind.
 
Kiwy
Kiwy
Good Job @FaheemMitha, it sounds like a lot of pain...
 
Faheem Mitha
Faheem Mitha
3:55 PM
@Kiwy Maybe I was doing something wrong. Dunno.
It's possible my old disks are misbehaving. Or systemd thinks they are. Because it seemed to keep wanting to fsck the LVs, but only the ones on the old disks.
Oh, it looks like I'm running Xfce.
Hang on, let me see if I can switch to KDE.
No, lightdm isn't giving me an option.
 
Kiwy
Kiwy
I gave a chance at XFCE after getting bored of GnomeShell inconsistency and instability, but it really is a pain in the ... Even more than GnomeShell
 
Faheem Mitha
Faheem Mitha
There's an Arch Linux question which mentions lightdm-kde-greeter, but there is not such package here.
Fine, giving sddm another shot. Maybe it had a problem with the .Xauthority permissions too.
Back in a bit. Hopefully.
Hooray, back up and running with KDE.
So it really was the .Xauthority file thing. You'd think a project that has been around for 20 years could produce an intelligible error message for a permissions issue.
 
Faheem Mitha
Faheem Mitha
4:24 PM
Can someone tell me the good way to fsck a ext4 filesystem on top of a LV? And does it have to be mounted to do the fsck?
And today was sobering in one respect. i realised how few things I could do without a computer.
 
Kusalananda
Kusalananda
4:36 PM
@overexchange You are deleting ${environ}.*/inner_folder. The real folder path is /home/user/a@1/inner_folder. There is no dot (.) in a@1.
@overexchange Remember that the shell uses filename globbing patterns, not regular expressions.
 
Torin
Torin
@FaheemMitha It would be best to unmount it first. You can run fsck.ext4 on the LV path
 
Kusalananda
Kusalananda
@overexchange What you want is rm -rf "$environ"*/inner_folder
 
Faheem Mitha
Faheem Mitha
@Torin Ok. Looking at the man page. There are lotsa flags.
 
Kusalananda
Kusalananda
@overexchange Or rm -rf "$environ"@1/inner_folder
 
Faheem Mitha
Faheem Mitha
Is it safe to run fsck.ext4 with the -f flag?
Hmm, sound is not working. Sigh.
 
sourcejedi
sourcejedi
4:45 PM
@FaheemMitha yes, it is no more dangerous than running fsck normally :-).
 
Faheem Mitha
Faheem Mitha
@sourcejedi Ok, thanks.
 
Faheem Mitha
Faheem Mitha
4:55 PM
Hmm, apparently timidity had grabbed control of my sound card.
Weird, I didn't even know I was running it.
Many are the fun ways to have your computer not work.
Yay, I've got sound back.
 
derobert
derobert
5:53 PM
@FaheemMitha am now
I don't have plymouth installed here either.
 
Faheem Mitha
Faheem Mitha
@derobert I didn't think you did. So this upgrade-reports thing. It doesn't have a template?
And I take it your upgrades are smoother?
 
derobert
derobert
Well, I haven't done many stretch -> buster updates yet since I've been tracking buster for a while...
 
Faheem Mitha
Faheem Mitha
@derobert Oh
 
derobert
derobert
But yeah, that sounds like a lot of stuff breaking.
 
Faheem Mitha
Faheem Mitha
So, upgrade-reports?
 
derobert
derobert
5:58 PM
Running sddm here, on nvidia-legacy even. So it does work....
Or can, at least.
 
Faheem Mitha
Faheem Mitha
@derobert I fixed the issue. .Xauthority was owned by root.
I would probably not have figured it out, but lightdm reported the issue correctly. sddm just crashed.
Probably worth reporting against sddm, actually.
Upgrade reports doesn't seem to have a template. Is that right?
 
derobert
derobert
@FaheemMitha Probably...
 
Faheem Mitha
Faheem Mitha
I guess I'll try writing a draft report. Are you willing to take a look?
 
derobert
derobert
Sure
 
Faheem Mitha
Faheem Mitha
Possibly etherpad. Does anyone have a recommendation for an etherpad server these days? It's hard to keep track.
 
Kusalananda
Kusalananda
6:02 PM
$ gdate -u -d @1555555555
Thu Apr 18 02:45:55 UTC 2019
 
Faheem Mitha
Faheem Mitha
I know I can use Google Docs, but I'm increasingly Google-allergic.
 
derobert
derobert
@Kusalananda That's hardly a significant date. It's not even even, much less a power of two :-P
 
Faheem Mitha
Faheem Mitha
@derobert I guess I could go with pad.riseup.net
At least they ask how long you want the pad for.
 
Kusalananda
Kusalananda
6:40 PM
@derobert No, it's not, but there's a lot of 5's.
 
derobert
derobert
@Kusalananda I mean, I guess a lot of 5s is normal if you check /usr/bin. But that's only when its in octal...
 
Jesse_b
Jesse_b
@Kusalananda: What does the @ signify? Is that just what you need to do to input an epoch?
 
derobert
derobert
... and 07555555555 was a while ago.
 
Jesse_b
Jesse_b
Ah yeah I see it in the gnu web docs, doesn't seem to be mentioned in the man page
 
derobert
derobert
EXAMPLES
       Convert seconds since the epoch (1970-01-01 UTC) to a date

              $ date --date='@2147483647'
... from the man page
(and if you check the info page, it has its own section)
 
Jesse_b
Jesse_b
6:46 PM
mouth <<<"foot"
Who reads the example section of a man page anyway
I'm on mac though so when I do info gdate it just opens the regular man page
 
Faheem Mitha
Faheem Mitha
6:58 PM
@Jesse_b Lots of people. More would, if there were actually examples.
 
Jesse_b
Jesse_b
7:08 PM
 
overexchange
overexchange
7:18 PM
@Kusalananda Does grep also use file name globbing patterns?
but not regex?
 
Jesse_b
Jesse_b
@overexchange There is a good answer here that describes many of the differences:
89
A: Why does my regular expression work in X but not in Y?

GillesUnfortunately, for historical reasons, different tools have slightly different regular expression syntax, and sometimes some implementations have extensions that are not supported by other tools. While there is a common ground, it seems like every tool writer made some different choices. The con...

But grep uses BRE unless -E or egrep is specified, in which case it uses ERE
 
Kusalananda
Kusalananda
@overexchange grep uses regular expressions to do matching of the contents of files, but to match filenames you use filename globbing patterns, regordless of what command you use.
For example: grep 'PATTERN1' PATTERN2 would first match what files to process with grep using PATTERN2 (a glob pattern), and then return all lines from these files that matched PATTERN1 (a regular expression).
 
Jesse_b
Jesse_b
yeah so grep '[0-9].*' file0?.txt is using both regex ([0-9].*) and shell globbing (file0?.txt)
 
Faheem Mitha
Faheem Mitha
@derobert Any idea what this might mean? I was getting it a lot. I think I still am:
 
Kusalananda
Kusalananda
@overexchange In your rm example, you just want to match filenames. This is done by the shell with a glob pattern.
 
Faheem Mitha
Faheem Mitha
7:24 PM
Apr 17 01:41:55 orwell kernel: scsi 10:0:0:1: Wrong diagnostic page; asked for 1 got 8
Apr 17 01:41:55 orwell kernel: scsi 10:0:0:1: Failed to get diagnostic page 0x1
Apr 17 01:41:55 orwell kernel: scsi 10:0:0:1: Failed to bind enclosure -19
So journalctl has 500k lines of logs going back to 31st March.
 
Kusalananda
Kusalananda
Oooh, failed to bind enclosure? Always be sure to bind your enclosures! Preferably to a wall or some other immovable object.
(sorry)
 
derobert
derobert
@FaheemMitha well, first you'd probably want to figure out which device 10:0:0:1 is
@Kusalananda It's a diagnostic page, so presumably the enclosure is a book. Usually you don't bind those to walls. You could chain them to the bookshelf, though.
 
Faheem Mitha
Faheem Mitha
@derobert Ok. Tell me how to, and I will.
 
derobert
derobert
lsscsi ?
(which is part of the lsscsi package, if you don't have it installed). Or look at /sys yourself, of course, though I hear dragons may hide there.
 
Jesse_b
Jesse_b
Faheem the scsi dragon slayer
 
Faheem Mitha
Faheem Mitha
7:33 PM
@derobert [10:0:0:1] enclosu WD SES Device 1007 -
Not sure what that is. And no, I didn't have it installed.
 
Jesse_b
Jesse_b
Sounds like a western digital hard drive
 
derobert
derobert
@Jesse_b Hah! There are no SCSI dragons, only gremlins. They're drawn to SCSI, especially if you have insufficient terminators. Or too many. Or the wrong type. Or its Tuesday. Or the terminators are in the wrong place.
@FaheemMitha so that's an external HDD
 
Jesse_b
Jesse_b
> Western Digital My Passport/My Book hard drives require a special service called SES (SCSI Enclosure Services) between the PC and the hard drive to enable certain features such as password protection, LED control, and access to the drives label, if applicable.
 
Faheem Mitha
Faheem Mitha
@derobert Well, I have one attached, but that's not it.
Huh, it's probably the Bluetooth adapter.
 
derobert
derobert
no, it's surely the external drive
 
Faheem Mitha
Faheem Mitha
7:35 PM
@derobert The Bluetooth adapter is a USB thingy.It couldn't be that?
I don't have that many external USB thingys plugged in.
 
derobert
derobert
I don't think bluetooth adapters get SCSI addresses
 
Faheem Mitha
Faheem Mitha
This is my external drive: 
[10:0:0:0]   disk    WD       My Passport 0820 1007  /dev/sdg
 
derobert
derobert
Nor label themselves WD enclosures
 
Faheem Mitha
Faheem Mitha
@derobert How sad for them.
 
derobert
derobert
@FaheemMitha that's the disk, the :1 is the enclosure
 
Faheem Mitha
Faheem Mitha
7:36 PM
@derobert Hmm, yes. That's Western Digital, right?
@derobert You lost me.
 
derobert
derobert
It's not just a disk. It's a disk inside an enclosure. Both are addressable.
 
CFJohnston
CFJohnston
Any SSIS experts in here?
 
Faheem Mitha
Faheem Mitha
@derobert News to me. What do I conclude from that, if anything?
@CFJohnston Isn't that an MS thing?
 
derobert
derobert
@FaheemMitha well, you could probably make the messages stop by unloading the scsi enclosure module (ses). The messages are harmless, though, and probably don't indicate any real issue.
 
Faheem Mitha
Faheem Mitha
@derobert Ok.
Moving on, then.
 
CFJohnston
CFJohnston
7:39 PM
@FaheemMitha yup
 
derobert
derobert
but see @Jesse_b's message above about what that's for
 
Faheem Mitha
Faheem Mitha
@derobert Which one?
 
derobert
derobert
the one that highlights, his last message
 
Faheem Mitha
Faheem Mitha
Ah, you must mean:
5 mins ago, by Jesse_b
> Western Digital My Passport/My Book hard drives require a special service called SES (SCSI Enclosure Services) between the PC and the hard drive to enable certain features such as password protection, LED control, and access to the drives label, if applicable.
So this is some sort of obscure SCSI bug?
 
derobert
derobert
[12:0:0:0]   disk    WD       easystore 25FB   3005  /dev/sdl
[12:0:0:1]   enclosu WD       SES Device       3005  -
... I have one too. Don't think I'm getting those errors. Different kernel or drive, probably.
 
Faheem Mitha
Faheem Mitha
7:41 PM
@derobert Is that yours?
@derobert So not worth reporting?
 
derobert
derobert
shrug, it's at least an annoying error message. Maybe you have something polling the enclosure, no idea what would by default.
 
Faheem Mitha
Faheem Mitha
BTW, journalctl shows errors in red, right?
Something called ntop is sure complaining a lot.
 
derobert
derobert
which was removed after jessie
 
Faheem Mitha
Faheem Mitha
@derobert That would explain it, then.
 
derobert
derobert
ntopng is a replacement
 
Faheem Mitha
Faheem Mitha
7:45 PM
Purging now.
Seeing this one a fair amount, but I think I just need to install the firmware:
 
derobert
derobert
I suspect you missed debian.org/releases/stretch/amd64/release-notes/… (and the equivalent section in the Buster notes)
 
Faheem Mitha
Faheem Mitha
Apr 17 01:57:43 orwell kernel: bluetooth hci0: firmware: failed to load brcm/BCM20702A1-0a5c-21e8.hcd (-2)
Apr 17 01:57:43 orwell kernel: firmware_class: See https://wiki.debian.org/Firmware for information about missing firmware
@derobert I tend not to read notes a lot.
Probably a character flaw.
Oh, and the other thing is that the nouveau kernel models starting loading by default, at some point, even with nvidia in xorg.conf. Shouldn't that be automatically be blacklisted by default?
 
derobert
derobert
@FaheemMitha yeah, possibly bluez-firmware
 
Faheem Mitha
Faheem Mitha
I had to manually blacklist it.
That took me some time to figure out.
 
derobert
derobert
@FaheemMitha It should be automatically blacklisted by having the nvidia drivers installed
 
Faheem Mitha
Faheem Mitha
7:50 PM
It was mentioned in dmesg, otherwise I wouldn't have spotted it.
@derobert I definitely had the legacy drivers installed.
Blacklisting should live in /etc/modprobe,d, right?
 
derobert
derobert
nvidia-kernel-common provides an /etc/modprobe.d/nvidia-kernel-common.conf here
actually I guess it's nvidia-blacklists-nouveau.conf
 
Faheem Mitha
Faheem Mitha
@derobert Yes, I'm seeing that too.
 
derobert
derobert
via alternatives /etc/alternatives/glx--nvidia-blacklists-nouveau.conf
 
Faheem Mitha
Faheem Mitha
dlocate nvidia-blacklists-nouveau.conf
nvidia-kernel-support: /etc/nvidia/current/nvidia-blacklists-nouveau.conf
nvidia-legacy-390xx-kernel-support: /etc/nvidia/legacy-390xx/nvidia-blacklists-nouveau.conf
But at one point it was definitely loading. And I can see it in the logs.
Maybe I didn't have the kernel-support packages installed?
 
derobert
derobert
Maybe. Do you have the -driver metapackage installed?
nvidia-legacy-390xx-driver
 
Faheem Mitha
Faheem Mitha
7:53 PM
@derobert I think I installed xserver-xorg-video-nvidia-legacy-390xx.
But that one too, as well.
 
derobert
derobert
And probably nvidia-legacy-390xx-kernel-dkms or some other way of getting the kernel module
 
Faheem Mitha
Faheem Mitha
@derobert Oh, that one I definitely had installed.
 
derobert
derobert
BTW, depending on how many obsolete packages you have failed to purge... even left them in config-only state... that could have been part of what was annoying systemd.
 
Faheem Mitha
Faheem Mitha
Does nvidia-graphics-drivers-legacy-390xx pull in nvidia-legacy-390xx-kernel-support? Don't see it here.
Apparently yes: 
apt-get -s purge nvidia-legacy-390xx-kernel-support
The following packages will be REMOVED:
  nvidia-legacy-390xx-driver* nvidia-legacy-390xx-kernel-dkms* nvidia-legacy-390xx-kernel-support*
I guess I'll chalk that one up to user error.
@derobert I was seeing stuff like:
pr 17 01:57:43 orwell kernel: nvidia: loading out-of-tree module taints kernel.
Apr 17 01:57:43 orwell kernel: nvidia: module license 'NVIDIA' taints kernel.
Apr 17 01:57:43 orwell kernel: Disabling lock debugging due to kernel taint
Apr 17 01:57:43 orwell kernel: nvidia: module verification failed: signature and/or required key missing - tainting kernel
Apr 17 01:57:43 orwell kernel: nvidia-nvlink: Nvlink Core is being initialized, major device number 243
Apr 17 01:57:43 orwell kernel: NVRM: The NVIDIA probe routine was not called for 1 device(s).
 
Kusalananda
Kusalananda
@derobert Well, it does sound like a kid's excuse for not doing their homework. "I looked for page 1, but I only got page 8. I failed to get page 1. The binding on this book is crap."
 
derobert
derobert
8:05 PM
@FaheemMitha You could check the script files from your upgrade to see what happened. (Provided of course you made them, which is also in the release notes, I think...)
 
Faheem Mitha
Faheem Mitha
@derobert No, unfortunately I didn't.
 
derobert
derobert
@Kusalananda LOL, yes it does.
 
Faheem Mitha
Faheem Mitha
And shortly afterwards: 
Apr 17 01:57:44 orwell kernel: NVRM: The NVIDIA GeForce GT 610 GPU installed in this system is
                               NVRM:  supported through the NVIDIA 390.xx Legacy drivers. Please
                               NVRM:  visit nvidia.com/object/unix.html for more
                               NVRM:  information.  The 410.104 NVIDIA driver will ignore
                               NVRM:  this GPU.  Continuing probe...
Apr 17 01:57:44 orwell kernel: NVRM: No NVIDIA graphics adapter found!
@derobert I think i see the possible problem: 
lrwxrwxrwx   1 root root   53 Apr 17 18:48 nvidia-blacklists-nouveau.conf -> /etc/alternatives/glx--nvidia-blacklists-nouveau.conf
lrwxrwxrwx   1 root root   43 Apr 17 18:48 nvidia.conf -> /etc/alternatives/glx--nvidia-modprobe.conf
But the /etc/alternatives was pointing in the wrong place. I had to manually change it to legacy.
update-glx --config nvidia
There are 2 choices for the alternative nvidia (providing /usr/lib/nvidia/nvidia).

Selection Path Priority Status
------------------------------------------------------------
0 /usr/lib/nvidia/current 410 auto mode
1 /usr/lib/nvidia/current 410 manual mode
* 2 /usr/lib/nvidia/legacy-390xx 390 manual mode
But it was pointing to 0 till I changed it. And that was after I removed the regular nvidia driver package.
If there is something that is supposed to change it, it didn't happen in my case.
 
derobert
derobert
@FaheemMitha odd... 
# update-glx --config nvidia
There is only one alternative in link group nvidia (providing /usr/lib/nvidia/nvidia): /usr/lib/nvidia/legacy-390xx
Nothing to configure.
 
Faheem Mitha
Faheem Mitha
@derobert Huh?
 
derobert
derobert
8:16 PM
Odd yours gives three choices, when you only have one installed.
Mine doesn't even give a choice.
 
Faheem Mitha
Faheem Mitha
Well, I guess I still have some of the 410 packages installed.
Just not the drivers, I think. 
root@orwell:/etc/modprobe.d# dpkg -l | grep 410
ii  libxnvctrl0:amd64                             410.104-2                            amd64        NV-CONTROL X extension (runtime library)
ii  linux-patch-grsecurity2                       3.0+3.14.22-201410250026-1           all          grsecurity kernel patch
ii  nvidia-alternative                            410.104-3                            amd64        allows the selection of NVIDIA as GLX provider
ii  nvidia-egl-common                             410.104-3                            amd64        NVIDIA binary EGL driver - common files
 
derobert
derobert
# dpkg -l | grep 410
ii  libxnvctrl0:amd64                                        410.104-2                                               amd64        NV-CONTROL X extension (runtime library)
ii  nvidia-egl-common                                        410.104-3                                               amd64        NVIDIA binary EGL driver - common files
ii  nvidia-libopencl1:amd64                                  410.104-3                                               amd64        NVIDIA OpenCL ICD Loader library
... same here.
some of those are weird, like the vdpau one
 
Faheem Mitha
Faheem Mitha
I guess the question is then where update-glx is getting its information from.
 
derobert
derobert
but probably you need to purge some of that
 
Faheem Mitha
Faheem Mitha
Perhaps there is no switching mechanism. I just depends on the other driver going away?
@derobert Which ones?
 
derobert
derobert
8:20 PM
@FaheemMitha you probably have some other nvidia packages in config files only state
 
Faheem Mitha
Faheem Mitha
Hmm, you don't have `nvidia-alternative` installed.
`
@derobert I purge everything.
 
derobert
derobert
personally I search for all packages in config-file-only state and purge them before and after the upgrade. ... though if you purge everything, probably not that
@FaheemMitha Indeed not. I guess one nvidia version is good enough for me :-/
 
Faheem Mitha
Faheem Mitha
@derobert That might be what is causing the issue.
Try installing it and see if the update-glx --config nvidia output changes.
 
derobert
derobert
Or I could not try and break my machine :-P
But yeah, that's probably at least part of why you have a couple of options.
 
Faheem Mitha
Faheem Mitha
@derobert Well, you could delete it immediately afterwards.
I remember I grepped for nouveau in /etc/modprobe.d.
But if there was a symlink pointing to nothing, that wouldn't have given a hit.
If I had done an ls -lah, I would immediately have realised there was something wrong.
 
derobert
derobert
8:29 PM
So, probably the release notes need to tell you how to properly switch to nvidia-legacy...
Although I'm not sure even perfect directions in the release notes would have helped (!)
 
Faheem Mitha
Faheem Mitha
8:52 PM
@derobert Do the release notes not currently do so?
Having thought about it, I don't think the alternatives system makes much sense here.
This is to allow having multiple versions of something be installed, and enabling the user to switch the default among that group, correct?
But in the case of drivers, the only really good reason for having multiple versions installed is if the idea is to have them usable simultaneously.
Possibly by different cards.
But the alternatives system is clearly not allowing for that. So what is the point?
Allowing the user to switch between different versions of the driver without reinstalling? But what is really the point of that? People don't exactly change their cards often. And if they do, they can always install a different version, as necessary.
@derobert Also note there is a /usr/share/doc/nvidia-legacy-390xx-driver/README.alternatives.
Which is where I learned about this. But I think it kind of violates the principle of least surprise.
 

« first day (3112 days earlier)      last day (1833 days later) »