Hilton Chain <hako@ultrarare.space> writes:
I appreciate your work and I think it would be valuable if you
can share
what can be improved in [1] based on your experience.
[1]:
https://guix.gnu.org/manual/devel/en/html_node/Bulk-Updates.html
Running `guix refresh` is only the first step. In almost every
upgrade I need to add new packages that are now referenced by
updated packages. So the next step for me is usually to run
`make` to see how big the mess is that I'm now in after running
`guix refresh`. It shows me missing variable references.
Then I try to figure out what all these guixified variable names
correspond to in CRAN or Bioconductor. It would be better if
`guix refresh` included the upstream name.
Since etc/committer.scm cannot deal with package additions I
commit these new packages first. etc/committer.scm also doesn't
deal with changes to `arguments`, so I commit those manually. It
also doesn't like changes to just the order of inputs, so those
packages are next. Only then do I run etc/committer.scm on the
rest.
Is it possible to divide the changes into chunks that can be
reviewed
independently? So that the workload at one time can be more
reasonable,
and this leaves room for others to join in.
I don't think that's feasible. The division would be arbitrary
and instead of doing the best I can with 500+ commits I'd have to
await review for 20+ chunks, rebase each of them and apply them in
the correct order. I'd rather move all those packages out into
another channel.
--
Ricardo
