bug#36865: Guix gc breaks grub

bug#36865: Guix gc breaks grub

[Xavier Montillet]

[-- Attachment #1: Type: text/plain, Size: 1575 bytes --]

*Symptoms*

After running guix gc, and rebooting, I got the following errors (which stay something like half a second on the screen):

error: no such device: /gnu/store/0fijjh7qw8ppsks0ba3q9m873ljhj797-grub-2.04/share/grub/unicode.pf2
error: file `/gnu/store/0fijjh7qw8ppsks0ba3q9m873ljhj797-grub-2.04/share/grub/unicode.pf2' not found.
error: no suitable video mode found.
error: no video mode activated.

I then get to the grub menu where I can pick which version of Guix I want to boot on, but all of them lead to getting stuck on the following text (with nothing moving after that for at least several minutes):

error: no suitable video mode found.
Booting in blind mod

*Fix
*

Fortunately nckx was on IRC and saved me by telling me to press C to get the grub command line when I got to the grub menu and run setup_gfxterm.

Running guix system reconfigure config.scm removed the need for setup_gfxterm for subsequent boots.

*Cause*

guix gc thinks that grub (and a lot of other stuff that it shouldn't think is dead) is dead, as shown by the following commands (given by nckx):


$ grep unicode /boot/grub/grub.cfg
search --file --set /gnu/store/0fijjh7qw8ppsks0ba3q9m873ljhj797-grub-2.04/share/grub/unicode.pf2
if loadfont /gnu/store/0fijjh7qw8ppsks0ba3q9m873ljhj797-grub-2.04/share/grub/unicode.pf2; then

$ guix gc --list-dead | grep 0fijjh7qw8ppsks0ba3q9m873ljhj797
finding garbage collector roots...
determining live/dead paths...
/gnu/store/0fijjh7qw8ppsks0ba3q9m873ljhj797-grub-2.04

I was also able to reproduce the bug by running guix gc and rebooting again.

[-- Attachment #2: Type: text/html, Size: 4694 bytes --]

bug#36865: Guix gc breaks grub

[Ricardo Wurmus]

Hi Xavier,

> guix gc thinks that grub (and a lot of other stuff that it shouldn't
> think is dead) is dead, as shown by the following commands (given by
> nckx):
>
>
> $ grep unicode /boot/grub/grub.cfg
> search --file --set /gnu/store/0fijjh7qw8ppsks0ba3q9m873ljhj797-grub-2.04/share/grub/unicode.pf2
> if loadfont /gnu/store/0fijjh7qw8ppsks0ba3q9m873ljhj797-grub-2.04/share/grub/unicode.pf2; then
>
> $ guix gc --list-dead | grep 0fijjh7qw8ppsks0ba3q9m873ljhj797
> finding garbage collector roots...
> determining live/dead paths...
> /gnu/store/0fijjh7qw8ppsks0ba3q9m873ljhj797-grub-2.04
>
> I was also able to reproduce the bug by running guix gc and rebooting again.

Oh, that’s not good.  Can you still reproduce this after running “guix
system reconfigure …”?  It should create a new system generation,
register a GC root, and update GRUB.

It should not say that GRUB itself is dead and thus not free it up for
garbage collection.

Could you perhaps share something about how you installed the system (if
it’s anything out of the ordinary)?

--
Ricardo

bug#36865: Guix gc breaks grub

[Xavier Montillet]

[-- Attachment #1: Type: text/plain, Size: 1640 bytes --]

Hi Rocardo,

I guix pull'ed and grub is no longer marked as dead. "guix gc --list-dead | grep 0fijjh7qw8ppsks0ba3q9m873ljhj797" now returns nothing, while "grep unicode /boot/grub/grub.cfg" till returns the same thing.

So I guess this means it's fixed?

I installed with the 1.0.1 installer I think, nothing out of the ordinary, except maybe everything being encrypted, including the /boot.

Xavier

On Wed, Jul 31, 2019, at 9:21 AM, Ricardo Wurmus wrote:
> 
> Hi Xavier,
> 
> > guix gc thinks that grub (and a lot of other stuff that it shouldn't
> > think is dead) is dead, as shown by the following commands (given by
> > nckx):
> >
> >
> > $ grep unicode /boot/grub/grub.cfg
> > search --file --set /gnu/store/0fijjh7qw8ppsks0ba3q9m873ljhj797-grub-2.04/share/grub/unicode.pf2
> > if loadfont /gnu/store/0fijjh7qw8ppsks0ba3q9m873ljhj797-grub-2.04/share/grub/unicode.pf2; then
> >
> > $ guix gc --list-dead | grep 0fijjh7qw8ppsks0ba3q9m873ljhj797
> > finding garbage collector roots...
> > determining live/dead paths...
> > /gnu/store/0fijjh7qw8ppsks0ba3q9m873ljhj797-grub-2.04
> >
> > I was also able to reproduce the bug by running guix gc and rebooting again.
> 
> Oh, that’s not good. Can you still reproduce this after running “guix
> system reconfigure …”? It should create a new system generation,
> register a GC root, and update GRUB.
> 
> It should not say that GRUB itself is dead and thus not free it up for
> garbage collection.
> 
> Could you perhaps share something about how you installed the system (if
> it’s anything out of the ordinary)?
> 
> --
> Ricardo
> 
> 

[-- Attachment #2: Type: text/html, Size: 2399 bytes --]

bug#36865: Guix gc breaks grub

[Xavier Montillet]

[-- Attachment #1: Type: text/plain, Size: 1852 bytes --]

I just tried running guix gc and restarting and can confirm that the symptoms disappeared.

On Wed, Jul 31, 2019, at 9:05 PM, Xavier Montillet wrote:
> Hi Rocardo,
> 
> I guix pull'ed and grub is no longer marked as dead. "guix gc --list-dead | grep 0fijjh7qw8ppsks0ba3q9m873ljhj797" now returns nothing, while "grep unicode /boot/grub/grub.cfg" till returns the same thing.
> 
> So I guess this means it's fixed?
> 
> I installed with the 1.0.1 installer I think, nothing out of the ordinary, except maybe everything being encrypted, including the /boot.
> 
> Xavier
> 
> On Wed, Jul 31, 2019, at 9:21 AM, Ricardo Wurmus wrote:
>> 
>> Hi Xavier,
>> 
>> > guix gc thinks that grub (and a lot of other stuff that it shouldn't
>> > think is dead) is dead, as shown by the following commands (given by
>> > nckx):
>> >
>> >
>> > $ grep unicode /boot/grub/grub.cfg
>> > search --file --set /gnu/store/0fijjh7qw8ppsks0ba3q9m873ljhj797-grub-2.04/share/grub/unicode.pf2
>> > if loadfont /gnu/store/0fijjh7qw8ppsks0ba3q9m873ljhj797-grub-2.04/share/grub/unicode.pf2; then
>> >
>> > $ guix gc --list-dead | grep 0fijjh7qw8ppsks0ba3q9m873ljhj797
>> > finding garbage collector roots...
>> > determining live/dead paths...
>> > /gnu/store/0fijjh7qw8ppsks0ba3q9m873ljhj797-grub-2.04
>> >
>> > I was also able to reproduce the bug by running guix gc and rebooting again.
>> 
>> Oh, that’s not good. Can you still reproduce this after running “guix
>> system reconfigure …”? It should create a new system generation,
>> register a GC root, and update GRUB.
>> 
>> It should not say that GRUB itself is dead and thus not free it up for
>> garbage collection.
>> 
>> Could you perhaps share something about how you installed the system (if
>> it’s anything out of the ordinary)?
>> 
>> --
>> Ricardo
>> 
>> 
> 

[-- Attachment #2: Type: text/html, Size: 2656 bytes --]

bug#36865: Guix gc breaks grub

[Ludovic Courtès]

Hi Xavier,

"Xavier Montillet" <xavierm02@xavierm02.fr> skribis:

> I just tried running guix gc and restarting and can confirm that the symptoms disappeared.

So, is this bug fixed?  If so, in what commit range was the fix
introduced?  That sounds like a serious issue so I’d rather make sure we
understand what happened.

Jakob, does that ring a bell?

Thanks,
Ludo’.

bug#36865: Guix gc breaks grub

[Xavier Montillet]

Hi Ludo,

All I can say for sure is that the bug was in the guix pull'ed version at some point in the week leading to July 31 (and most likely on July 30), and that the bug is no longer in the version of July 31 evening.

Xavier

On Fri, Aug 23, 2019, at 12:31 PM, Ludovic Courtès wrote:
> Hi Xavier,
> 
> "Xavier Montillet" <xavierm02@xavierm02.fr> skribis:
> 
> > I just tried running guix gc and restarting and can confirm that the symptoms disappeared.
> 
> So, is this bug fixed?  If so, in what commit range was the fix
> introduced?  That sounds like a serious issue so I’d rather make sure we
> understand what happened.
> 
> Jakob, does that ring a bell?
> 
> Thanks,
> Ludo’.
>

bug#36865: Guix gc breaks grub

[Ludovic Courtès]

Hello,

"Xavier Montillet" <xavierm02@guix.xavierm02.fr> skribis:

> All I can say for sure is that the bug was in the guix pull'ed version at some point in the week leading to July 31 (and most likely on July 30), and that the bug is no longer in the version of July 31 evening.

Could you run ‘guix pull -l 1m’ (for this specific ‘guix’ command) to
see exactly what the commits were for these two generations?

Thanks in advance,
Ludo’.

bug#36865: Guix gc breaks grub

[Xavier Montillet]

Hi,

Jul 26 2019 23:55:23   d23a00b599be56694065bd274184b9289fb8b85c
Jul 29 2019 11:32:04   ab20b3ed9152c7c95d0d2c6b2d65e29983ab57ce
Jul 29 2019 23:52:32   18c4b0a27705773e423fb17310394204b7295d4a
Jul 30 2019 before 22:03      nckx saves me (last messages of http://logs.guix.gnu.org/guix/2019-07-30.log)
Jul 30 2019 22:03      I send the bug report (first message of http://logs.guix.gnu.org/guix/2019-07-31.log)
Jul 31 2019 9:21      Ricardo responds to my bug report telling me to guix reconfigure
Jul 31 2019 09:30:31   bab94ffa0e27e39c02d5ce3add5605b676b76bee
Jul 31 2019 21:05      I send the email saying that Grub is no longer marked as dead after guix pulling
Jul 31 2019 13:31:40   716908411b4d393ec82d5b7e40c9817e81c8fa95
 Aug 01 2019 09:12:52   e7dfbae8a99995abc9f088452ca35371d38eb343

So I would say that the bug was in18c4b0a27705773e423fb17310394204b7295d4a and was fixed before (or in) bab94ffa0e27e39c02d5ce3add5605b676b76bee.

Xavier

On Sat, Aug 24, 2019, at 2:04 PM, Ludovic Courtès wrote:
> Hello,
> 
> "Xavier Montillet" <xavierm02@guix.xavierm02.fr> skribis:
> 
> > All I can say for sure is that the bug was in the guix pull'ed version at some point in the week leading to July 31 (and most likely on July 30), and that the bug is no longer in the version of July 31 evening.
> 
> Could you run ‘guix pull -l 1m’ (for this specific ‘guix’ command) to
> see exactly what the commits were for these two generations?
> 
> Thanks in advance,
> Ludo’.
>

bug#36865: Guix gc breaks grub

[Jakob L. Kreuze]

[-- Attachment #1: Type: text/plain, Size: 210 bytes --]

Hi Ludovic,

Apologies for not participating in this thread until just now.

Ludovic Courtès <ludo@gnu.org> writes:

> Jakob, does that ring a bell?

Yes, this was fixed by #36880.

Regards,
Jakob

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]

bug#36865: Guix gc breaks grub

[Ludovic Courtès]

Hi Jakob,

zerodaysfordays@sdf.lonestar.org (Jakob L. Kreuze) skribis:

> Ludovic Courtès <ludo@gnu.org> writes:
>
>> Jakob, does that ring a bell?
>
> Yes, this was fixed by #36880.

Looking more closely, it seems to be
<https://issues.guix.gnu.org/issue/36942>.

Anyway, closing.  Thanks for your feedback!

Ludo’.