This adds one more layer to the output dict, to match critpath
and allow for the possibility that we'll provide info on compose-
critical things that *aren't* packages in future (which would
make the script name a misnomer, but never mind).
Signed-off-by: Adam Williamson <awilliam@redhat.com>
The last commit inadvertently broke the non-'all' mode by not
putting its output filenames into lists as the output code now
expects.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
Introduce build_composeinfo_wrapper.py, which uses compose-utils
legacy conversion when Pungi metadata is available and adds a
release_type field under [product]. Falls back to the existing treeinfo
walker when composeinfo.json is not present.
Signed-off-by: Samyak Jain <samyak.jn11@gmail.com>
Skip packages that already have a checkout, so interrupted runs
Resume past cloned trees instead of deleting and recloning.
Line-buffer output for tee, we used to do it in the last script.
Signed-off-by: jnsamyak <samyak.jn11@gmail.com>
This is probably what we want for Bodhi purposes. It only operates
on Branched and Rawhide...we could work on the container and cloud
composes for stable releases and get a 'compose critical' set
that way, but we can't get the source packages, and it's of
questionable value anyway.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
Whoops, left an empty function from an earlier design lying around
and the spacing between functions wasn't consistent.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
This adds ccp.py - a script for generating the "compose-critical
package list" for a given compose. It's a bit of a WIP but it
does already work and give you a fairly useful list (as a flat
text file for now).
Signed-off-by: Adam Williamson <awilliam@redhat.com>
compose-tracker-issues lives on forge.fedoraproject.org now, so this
script lists and closes open tickets there instead of Pagure.
Signed-off-by: Samyak Jain <samyak.jn11@gmail.com>
- Clone each package repository into a separate temporary directory and delete
it before moving on to the next package to avoid filling up TMPDIR
when retiring many packages at once.
- Ignore comments in package list file
- Flush log file after each package write
- Set Git author to Orphaned Packages Process
- Add naive retry logic for git operations to workaround distgit
instability — thanks AI scrapers :(
- Clean up package listing code
The $STAGE argument has been unused since @kevin split the script
up and fixed some problems in it. Just removing it could cause
problems if people use old invocations fron history or something,
though, so let's allow either passing it or not passing it and
print a warning if it's passed.
Also, update the comment on how the script should be run to
account for the changes to the repository layout and the script's
name.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
Currently we write the data for Rawhide to rawhide.json and
rawhide.txt. This is how Bodhi expects to find it...so long as
the release in question *still is* Rawhide. But there's a timing
issue at branching: the release that was Rawhide suddenly no
longer is, and its branch name changes to fXX, so now Bodhi will
look for the critpath data as fXX.json or fXX.txt. From the time
the Bodhi branch name is changed until the first time critpath
generation runs after a Branched compose has completed and been
synced, Bodhi will no longer have critpath data for the release
and will consider all updates non-critpath, which is bad.
This should completely avoid the problem by just *always* writing
the Rawhide critpath data under both names, so at branching, until
we get 'correct' fresh critpath data for the newly-branched release
we will at least just apply the latest Rawhide critpath to it.
Having the numbered files when we don't need them shouldn't cause
any problems. And given the way this script is written it also
should never be possible for us to write Rawhide numbered files
over top of Branched numbered files. So I'm pretty sure it should
be safe to always do this.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
...like right now, after branching, we're blowing up because alt
arches aren't synced for some reason. We'd rather have some
critpath than no critpath at all, thanks.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
We no longer have Modular updates
We no longer ship i686 updates
We no longer ship armhfp (32bit arm) updates.
Signed-off-by: Kevin Fenzi <kevin@scrye.com>
I've extended the initial comment to better explain how to use the script and removed the `fedscm_admin` code since we're no longer maintain that tool.
Signed-off-by: mattia <mattia.verga@proton.me>
