bug#50568: Missing source code

[pelzflorian (Florian Pelz)]

On Mon, Sep 20, 2021 at 09:25:27AM +0200, Ludovic Courtès wrote:
> The thing is that this code is used by the internal ‘guix
> perform-download’ command, invoked by ‘guix-daemon’ when performing a
> source download.

Thank you for enlightening me; I had not expected guix-daemon.

Success!  After

./bootstrap
./configure --localstatedir=/var
make
GUIX_ALLOW_ME_TO_USE_PRIVATE_COMMIT=y make update-guix-package

and git commit -a, guix pull, reconfigure and reboot, your
web.archive.org patch downloads from web.archive.org with the steps
from before:

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> skribis:
> I guix pull with the patch applied, add the IP from
>
> guix environment --ad-hoc bind:utils -- dig web.archive.org
>
> to /etc/hosts, set my DNS server to a non-existing 10.0.0.50, disable
> IPv6, then “sudo herd stop nscd”, re-connect Network Manager (ping
> gnu.org fails now) and LC_ALL=en_US.utf8 guix install hello.

/var/log/guix/drvs/wy/24jyd0mm2jlgpzsxxlip03df065g35-hello-2.10.tar.gz.drv.bz2
says after various failures in name resolution it got downloaded from
web.archive.org.  I enjoy GNU Hello now.  Thank you for making
reproducibility easier.

> To test it, you can first start the new daemon, along these lines:
> 
>   sudo herd stop guix-daemon
>   sudo -E ./pre-inst-env guix-daemon --build-users-group=guixbuild

No, going back one generation and guix gc -D’ing the
/gnu/store/*hello-2.10* after adding web.archive.org to /etc/hosts
again, I for some reason get a

florian@florianacer ~$ bzcat 
/var/log/guix/drvs/wy/24jyd0mm2jlgpzsxxlip03df065g35-hello-2.10.tar.gz.drv.bz2
error: failed to run download program '/home/florian/src/guix/scripts/guix': 
Keine Berechtigung

(Keine Berechtigung is German for Permission denied) but it does not
matter, since the reconfigure and reboot works when and only when
adding web.archive.org to /etc/hosts.

Actually I will later take another look at this, because the manual
says ./pre-inst-env guix-daemon should work.

> Another option is to call ‘url-fetch’ manually at the REPL.

This I didn’t try.

Regards,
Florian