Open sdarwin opened 10 months ago
Sure. I just have to merge the code from cppalliance/boost and review it, as we did for the antora functionality.
That's good news. I didn't know this change was also that important. I actually thought there would be a lot of debate over it.
The defaults you suggested are already similar to the names I used in cppalliance/boost so that should be fine.
@alandefreitas you had mentioned at some point the idea of creating docs-only and source-only archives for boost.
could you add that in release-tools?
@glenfe just informed us the jfrog downloads are taking 200TB of bandwidth per month.
Switching to source-only downloads would save $1000's per month.
These are the hosted bundles https://boostorg.jfrog.io/artifactory/main/
There are two places in release-tools where docs-only and source-only archives might be generated.
There are pros and cons to having the logic in one or the other. Most of the complexity has usually been contained in
ci_boost_release.py
so that is the likely place. Create subdirectores in jfrog under master/ and develop/.Arguably, the name of the snapshots should always stay the same, such as boost_1_83_0-snapshot.tar.gz , and the determining factor is the directory where it is hosted. This makes it easier in other scripts which consume the archives, to simply switch the location (from full/ to source/) .
Possibly, the docs-only archive could have a
-docs
added to the name.ci_boost_release.py
generates and uploads the various snapshot to jfrog.publish_release.py
at the time of a release, downloads snapshots, renames them to release/1.83.0/source/boost_1_83_0.tar.gz, and uploads again.Unfortunately, the current archives are using the name "source" already.
I was hoping the first iteration of this feature would publish to multiple folders, but somehow keep the current behavior for a short time, since the old website depends on /source/ actually being /full/.
Perhaps temporarily then: