bug#69058: Failure when fallback to Disarchive+SWH

[Simon Tournier]

Hi Ludo,

On lun., 12 févr. 2024 at 12:24, Ludovic Courtès <ludovic.courtes@inria.fr> 
wrote:
> Ludovic Courtès <ludovic.courtes@inria.fr> skribis:
>
>> I can reproduce it with c4372f7ebfac17236299346fe60c6ca125336b1f:
>>
>> scheme@(guile-user)> (swh-download "https://gitlab.com/geontime/redkite.git"; 
>> "v0.8.0" "/tmp/redkite")
>> SWH: found revision 68276b9b88491198c33b03c4188dfe7d6e5dbf2a with directory 
>> at 
>> 'https://archive.softwareheritage.org/api/1/directory/d14c0eb6801bca920d1e1baac7a4b9e2a0589728/'
>> SWH: object swh:1:dir:d14c0eb6801bca920d1e1baac7a4b9e2a0589728 could not be 
>> fetched from the vault
>> $5 = #f
>>
>> The bug is that the Vault now returns 302 (redirect) in this case, which
>> is similar to the bug I fixed in <https://issues.guix.gnu.org/68741>.
>
> In fact it’s the exact same problem.
>
> Fixed in 1610a632d4b3097282d18af27ff3e9e178d7dfcb.

Cool!  Thanks.

However, it means that it stays broken for all commits before this fix,
right?  Somehow, if only SWH is available, then

    guix time-machine -C channels.scm -- shell -m manifest.scm

will fail for all channels older than today.  Right?

For instance, redoing your ReScience paper from 2020 using only SWH as I
did [1] some months ago is now broken (because a change on SWH side).

Well, I am fine to keep this bug closed but for the posterity we need
something like:

    guix swh get -C channels.scm -m manifest.scm

as we discussed elsewhere (off-list IIRC :-)).

1: https://simon.tournier.info/posts/2023-06-23-hackathon-repro.html

Cheers,
simon