fix: Use the right channels when upgrading global packages #1326
Conversation
olivier-lacroix
force-pushed
the
refactorupgrade
branch
from
5c10df1
to
da84431
Compare
olivier-lacroix
changed the title
olivier-lacroix
force-pushed
the
refactorupgrade
branch
from
da84431
to
2c86a3a
Compare
|
I actually think we should store the initial request in the environment - requested platform and channel. In the future we migth store more (like |
|
@wolfv is there a specific method to do that? |
|
Not yet I believe. My idea was to store a little JSON file in the |
|
Ok @wolfv . Let me see about that. |
olivier-lacroix
force-pushed
the
refactorupgrade
branch
from
2c86a3a
to
e7f950a
Compare
olivier-lacroix
force-pushed
the
refactorupgrade
branch
from
e7f950a
to
e9f7c07
Compare
olivier-lacroix
force-pushed
the
refactorupgrade
branch
from
f66421b
to
f9e1824
Compare
|
@wolfv , I think the question of storing data in a file for global packages is broader than this PR. So I prefer to not get into that rabbit hole here. @pavelzw any chance you could check this indeed resolves #1318 ? I made a number of steps parallel. It does not seem to have a meaningful impact in terms of speed, so happy to either leave it as is, or revert to a simpler, non-parallel upgrade function. |
| // Fetch sparse repodata across all channels | ||
| let all_channels = channels.values().chain(channel_cli.iter()).unique(); | ||
| let (client, repodata) = | ||
| get_client_and_sparse_repodata(all_channels, *platform, &config).await?; |
There was a problem hiding this comment.
I think this hasn't changed, but I just noticed if you installed a package from a channel that isn't available anymore, this fails. And it doesn't update any environment.
Might be nice to filter these and just list which where skipped in the end.
I'm not avoiding this to merge because of it but just picking your brain.
The case for me was that I installed a locally build package which I had push to /tmp and installed from there.
Fixes #1318