new GuixSD version?

new GuixSD version?

[Ethan Stefan Day]

[-- Attachment #1: Type: text/plain, Size: 357 bytes --]

Hello,

Will be a new version of GuixSD installation image once the dust settles
from the Hydra outage and the Sourceforge URL issue?  I ask because as I
understand it, it will be impossible to install from that image without
running 'guix pull' inside the installer, and I think I recall someone
on guix-devel saying to not do that.

-Ethan Day


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: new GuixSD version?

[Leo Famulari]

On Tue, Jul 19, 2016 at 05:36:50PM -0400, Ethan Stefan Day wrote:
> Will be a new version of GuixSD installation image once the dust settles
> from the Hydra outage and the Sourceforge URL issue?  I ask because as I
> understand it, it will be impossible to install from that image without
> running 'guix pull' inside the installer, and I think I recall someone
> on guix-devel saying to not do that.

Can you clarify why you think it would be impossible to install from the
0.10.0 GuixSD image? Hopefully that's not the case...

Re: new GuixSD version?

[Ethan Stefan Day]

[-- Attachment #1: Type: text/plain, Size: 1573 bytes --]



On 07/19/2016 11:55 PM, Leo Famulari wrote:
> On Tue, Jul 19, 2016 at 05:36:50PM -0400, Ethan Stefan Day wrote:
>> Will be a new version of GuixSD installation image once the dust settles
>> from the Hydra outage and the Sourceforge URL issue?  I ask because as I
>> understand it, it will be impossible to install from that image without
>> running 'guix pull' inside the installer, and I think I recall someone
>> on guix-devel saying to not do that.
> 
> Can you clarify why you think it would be impossible to install from the
> 0.10.0 GuixSD image? Hopefully that's not the case...
> 

The installer requires additional software to be downloaded.  I have not
checked to see if this is actually the case, but it seems likely that
there is an intersection between the set of packages required to be
downloaded at install time, the set of packages whose binaries were lost
in the Hydra failure, the set of packages affected by the Sourceforge
URL issue, and the set of packages whose hashes have changed since
GuixSD 0.10.0 was released.  If that is the case, then there exists a
package P such that:
-The installer requires P to proceed.
-Hydra does not have the binary for P (it is silently ignoring bad URLs).
-The '--fallback' option for 'guix system init' will not work because of
the bad URLs.  (although I think the intent was for users to not have to
use that anyway)

If my concern is based on accurate understanding, I would be willing to
test it.  I was about to reinstall GuixSD on metal for unrelated reasons.

-Ethan Day


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: new GuixSD version?

[Leo Famulari]

On Wed, Jul 20, 2016 at 12:45:49AM -0400, Ethan Stefan Day wrote:
> -The installer requires P to proceed.
> -Hydra does not have the binary for P (it is silently ignoring bad URLs).
> -The '--fallback' option for 'guix system init' will not work because of
> the bad URLs.  (although I think the intent was for users to not have to
> use that anyway)

We should at least be able to re-add the source code to Hydra, and
hopefully the binaries. Since the source code is content-addressed, it
can be fetched from many places, such as the local /gnu/store, an URL,
the substitutes server, and also content-addressed mirrors [0] (but not
in 0.10.0).

In the worst case, users can find new source URLs and use `guix
download` to add the source code of missing binaries whose source URLs
have changed to the store.

We should think about how to make the release tags more resilient.

> If my concern is based on accurate understanding, I would be willing to
> test it.  I was about to reinstall GuixSD on metal for unrelated reasons.

Yes, please report anything that's missing.

[0]
http://git.savannah.gnu.org/cgit/guix.git/commit/?id=cd436bf05a8344acf4462f3602e7d360821a902a

Re: new GuixSD version?

[Ludovic Courtès]

Leo Famulari <leo@famulari.name> skribis:

> On Wed, Jul 20, 2016 at 12:45:49AM -0400, Ethan Stefan Day wrote:
>> -The installer requires P to proceed.
>> -Hydra does not have the binary for P (it is silently ignoring bad URLs).
>> -The '--fallback' option for 'guix system init' will not work because of
>> the bad URLs.  (although I think the intent was for users to not have to
>> use that anyway)
>
> We should at least be able to re-add the source code to Hydra, and
> hopefully the binaries. Since the source code is content-addressed, it
> can be fetched from many places, such as the local /gnu/store, an URL,
> the substitutes server, and also content-addressed mirrors [0] (but not
> in 0.10.0).

I think the main problem is that some of the 0.10.0 binaries have
vanished from {mirror.,}hydra.gnu.org.  Thus, people installing GuixSD
for the first time today have a very bad experience.

We can rebuild the ‘release-0-10’ branch on Hydra, but of course we’re
limited by the available resources.

At any rate, Ethan: I think we’ll make a 0.10.1 release once we’ve merge
‘core-updates’, which could be a week or two from now.

Thanks,
Ludo’.

Re: new GuixSD version?

[Vincent Legoll]

> I think the main problem is that some of the 0.10.0 binaries have
> vanished from {mirror.,}hydra.gnu.org.  Thus, people installing GuixSD
> for the first time today have a very bad experience.

I concurr, today's nightmarish.

Long story short:

My first VM in which I installed guixsd successfully, didn't have enough
disk space. So I decided to make a new one from scratch : Bad idea...

And as I'm a complete scheme newbie, I cannot make sense of the
error messages I get, I cannot even know if PEBCAK...

-- 
Vincent Legoll

Re: new GuixSD version?

[Ludovic Courtès]

Vincent Legoll <vincent.legoll@gmail.com> skribis:

> Long story short:
>
> My first VM in which I installed guixsd successfully, didn't have enough
> disk space. So I decided to make a new one from scratch : Bad idea...

In the meantime, you could also generate your own up-to-date
installation image, which wouldn’t have this missing-substitute issue:

  https://www.gnu.org/software/guix/manual/html_node/Building-the-Installation-Image.html

> And as I'm a complete scheme newbie, I cannot make sense of the
> error messages I get, I cannot even know if PEBCAK...

Please report the confusing error messages.  We need to improve them,
but we need to know exactly what you see and what’s confusing about it!
Old farts like are too much brain-damaged to even realize that error
messages are confusing.  ;-)

Ludo’.

Re: new GuixSD version?

[Vincent Legoll]

> Please report the confusing error messages.  We need to improve them,
> but we need to know exactly what you see and what’s confusing about it!
> Old farts like are too much brain-damaged to even realize that error
> messages are confusing.  ;-)

OK I'll retry and report the error message here.

Thanks for the help, it's much appreciated !

-- 
Vincent Legoll

Re: new GuixSD version?

[Ludovic Courtès]

Vincent Legoll <vincent.legoll@gmail.com> skribis:

>> Please report the confusing error messages.  We need to improve them,
>> but we need to know exactly what you see and what’s confusing about it!
>> Old farts like are too much brain-damaged to even realize that error
                ^^
Like me!

>> messages are confusing.  ;-)
>
> OK I'll retry and report the error message here.

Thanks!

Ludo’.

Re: new GuixSD version?

[Ethan Stefan Day]

[-- Attachment #1: Type: text/plain, Size: 1942 bytes --]



On 07/20/2016 01:59 AM, Leo Famulari wrote:
> On Wed, Jul 20, 2016 at 12:45:49AM -0400, Ethan Stefan Day wrote:
>> If my concern is based on accurate understanding, I would be willing to
>> test it.  I was about to reinstall GuixSD on metal for unrelated reasons.
> 
> Yes, please report anything that's missing.
> 

So I guess the installer was more resilient than I expected;  it worked
fine.  (using default desktop configuration on x86_64)  It took a long
time, but from what Ludovic was saying, that was expected.  Optional
details follow:

My initial plan was to run 'guix system init', look at what failed, and
try to resolve missing dependencies.  So I ran:

time guix system init --fallback --keep-going /mnt/etc/config.scm \
/mnt &>> /mnt/unique-logging-dir-9855/run-one.log

It took about 6 hours (on my fairly powerful modern laptop).
run-one.log is 96.4 MB, so going through it by hand was not an option.
Luckily, the last line was:

Installation Finished.  No error reported.

I have not done anything yet, but it boots and lets me log in as root.
If the log file is of interest, I am willing to share it if someone
provides a way to transfer it or gives me a hint about how to extract
the relevant information programmatically.

There was some other output from the system that resulted from that
command.  I have no idea what it means.  Also, it was copied by hand, so
there may be errors.  here it is:

spurious SIGPOLL
spurious SIGPOLL
spurious SIGPOLL
spurious SIGPOLL
spurious SIGPOLL
[ 6791.830086] traps: test-libnm-link[4412] trap int3 \
    ip:7fffffff9fe0 error:0
[ 6791.831123] NOHZ: local_sofirq_pending 282
[ 6791.832111] NOHZ: local_sofirq_pending 282
[ 6791.832991] NOHZ: local_sofirq_pending 282
[ 6791.833747] NOHZ: local_sofirq_pending 282
[ 6791.830059] perf interrupt took too long (2503 > 2500), \
   lowering kernel.perf_event_max_sample_rate to 50000


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]