Failing to build the website

Hi Guix!

I tried to update the po files for the website today, but I was unable

to test. Trying to build the website from a clean checkout (without

updating translations) gives this error. I think it's related to the

code that generate the list of packages:

Backtrace:

In ice-9/eval.scm:

177:49 19 (lp (#<procedure 7ffff75d7e20 at ice-9/eval.scm:182:…> …))

177:49 18 (lp (#<procedure 7ffff75d7de0 at ice-9/eval.scm:282:…> …))

177:49 17 (lp (#<procedure 7ffff75d7dc0 at ice-9/eval.scm:282:…> …))

177:49 16 (lp (#<procedure 7ffff75d7da0 at ice-9/eval.scm:282:…> …))

177:49 15 (lp (#<procedure 7ffff75d7d60 at ice-9/eval.scm:182:…> …))

177:49 14 (lp (#<procedure 7ffff75d7d40 at ice-9/eval.scm:282:…> …))

177:49 13 (lp (#<procedure 7ffff75d7d20 at ice-9/eval.scm:182:…> …))

177:49 12 (lp (#<procedure 7ffff75d7ce0 at ice-9/eval.scm:282:…> …))

177:32 11 (lp (#<procedure 7ffff75d7cc0 at ice-9/eval.scm:182:7 …>))

159:9 10 (_ #(#(#(#(#(#<directory (apps packages te…>) …) …) …) …))

174:20 9 (_ #(#(#(#(#(#<directory (apps packages te…>) …) …) …) …))

177:49 8 (lp (#<procedure 7fffed7ab4c0 at ice-9/eval.scm:182:…> …))

177:32 7 (lp (#<procedure 7fffed7ab4a0 at ice-9/eval.scm:182:…> …))

174:20 6 (_ #(#(#(#(#(#<directory (apps packages te…>) …) …) …) …))

177:32 5 (lp (#<procedure 7fffeb50d3c0 at ice-9/eval.scm:182:…> …))

174:20 4 (_ #(#(#(#(#(#<directory (apps packages te…>) …) …) …) …))

177:49 3 (lp (#<procedure 7fffeb0b7f80 at ice-9/eval.scm:282:…> …))

177:32 2 (lp (#<procedure 7fffeb0b7f60 at ice-9/eval.scm:182:7 …>))

159:9 1 (_ #(#(#<directory (apps packages templates compone…>) …))

In srfi/srfi-1.scm:

241:2 0 (map _ _)

srfi/srfi-1.scm:241:2: In procedure map:

In procedure map: Wrong type argument: "https://www.qt.io/"

building pages in '/tmp/gnu.org/software/guix'...

Backtrace:

2 (primitive-load "/gnu/store/qr419pk9hdvg692rkhdlyvr9zh0?")

In ice-9/eval.scm:

619:8 1 (_ #f)

In guix/build/utils.scm:

762:6 0 (invoke "/gnu/store/bgh62naj3w2367h46zv2waq26rs3aspp-h?" ?)

guix/build/utils.scm:762:6: In procedure invoke:

ERROR:

1. &invoke-error:

program:

"/gnu/store/bgh62naj3w2367h46zv2waq26rs3aspp-haunt-0.2.6/bin/haunt"

arguments: ("build") exit-status: 1

term-signal: #f

stop-signal: #f

builder for

`/gnu/store/sh3ihg22yr1s0cdfjqpsv5hqrfmvg646-guix-web-site-de_DE.drv'

failed with exit code 1 la compilation de

/gnu/store/sh3ihg22yr1s0cdfjqpsv5hqrfmvg646-guix-web-site-de_DE.drv a

échoué

Hi Julien

On dim., 04 sept. 2022 at 17:18, Julien Lepiller <julien@lepiller.eu> wrote:

Which revision of Guix do you use? And which commit of guix-artwork?

And how do you build the website? It works for me with this procedure.

Cheers,

simon

I used guix build -f .guix.scm. I'm on the latest guix-artwork but an older guix from maybe a month ago. GUIX_WEB_SITE_LOCAL doesn't build everything, only a subset of packages end up in the generated list. I think my issue is with qt, which probably isn't part of that subset.

Le 5 septembre 2022 10:35:43 GMT+02:00, zimoun <zimon.toutoune@gmail.com> a écrit :

zimoun <zimon.toutoune@gmail.com> writes:

Yes but `guix build -f .guix.scm` fails for me too, perhaps because it

uses latest-guix from %default-channels. It’s not fixed by using old

haunt, so I guess a change in guix makes the difference, but I have not

found the commit and will not continue looking for it immediately.

Thank you for looking at it.

Regards,

Florian

On lun., 05 sept. 2022 at 12:27, Julien Lepiller <julien@lepiller.eu> wrote:

It works with GUIX_WEB_SITE_LOCAL=no but indeed not using .guix.scm.

Cheers,

simon

Hi,

On lun., 05 sept. 2022 at 12:59, "pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> wrote:

The file manifest.scm contains:

and the file .guix.scm contains

so it should be the same Haunt. I mean, there is no gap between the

version of Guile of latest-guix and the package guix. But, I am missing

something because the failure comes from:

It requires some investigations. :-)

Cheers,

simon

Quoting my own email I think this is the relevant bit. I don't think it's an issue in haunt, but something with qt:

srfi/srfi-1.scm:241:2: In procedure map:

In procedure map: Wrong type argument: "https://www.qt.io/"

building pages in '/tmp/gnu.org/software/guix'...

Le 5 septembre 2022 14:04:10 GMT+02:00, zimoun <zimon.toutoune@gmail.com> a écrit :

Hi,

On dim., 04 sept. 2022 at 17:18, Julien Lepiller <julien@lepiller.eu> wrote:

The culprit seems the package ’qtserialport’ in ’packages-builder’ from

(apps packages builder). Maybe introduced by commit

1ef04fb2288dade3ad2883026ae286a68ef13a1e. Well, the website probably

requires a tweak in ’package-t’, more or less here:

but I do not know exactly what. Maybe ’shtml’ somewhere?

Cheers,

simon

zimoun <zimon.toutoune@gmail.com> writes:

This was a first failing commit for me too, but the reason the website

does not build appears to be the license field of qtshadertools in

commit 1d65ff8fdeb20cc2db956093f0ecb1f3f72afc0e (Cc to Maxim). I could

make a patch but not today. Maxim, would you fix it?

I don’t know if there is a way to catch such mistakes early. Field

sanitizers?

Regards,

Florian

Hi,

On Wed, 07 Sep 2022 at 18:31, "pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> wrote:

From my opinion, this

is a bad practise; instead ’(inherit qtbase)’ and rewrite all the others

fields is better.

Well, the field reads,

+ (license (package-home-page qtbase))))

so why is the compiler accepting that? Is a warning raised?

Cheers,

simon

Le 7 septembre 2022 19:07:10 GMT+02:00, zimoun <zimon.toutoune@gmail.com> a écrit :

No warning because the compiler has no way to know you don't want to do that. license is a field, and fields don't have a specific type except for convention.

package-home-page will read a field from another object, return a perfectly valid Scgeme object (a string) and Scheme is happy to put it in the license field. Why would it care? (:

It fails much later when something tries to read the license field and gets a string it treats as a license object.

Hi Julien,

Thanks for explaining.

On Wed, 07 Sep 2022 at 19:14, Julien Lepiller <julien@lepiller.eu> wrote:

Arf, it is where OCaml shines. ;-)

Cheers,

simon

severity 57581 important

quit

Hi,

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> skribis:

Fixed in e61c581805b2338ea8feaac86617c5d7bff41c41, thanks for the

investigation!

Yes, we should do that. We can arrange to have most checks done at

compile time, similar to how the ‘synopsis’ and ‘description’ sanitizers

work (well, with extra macro magic).

Ludo’.

Hi Ludo,

On Tue, 13 Sep 2022 at 16:41, Ludovic Courtès <ludo@gnu.org> wrote:

For the record, please see this thread [1] opening a discussion on the

topic.

1: https://yhetil.org/guix/868rmuxpnw.fsf@gmail.com

Cheers,

simon