Comments page #1 of 3 pages

  • As noted in the spec comments, the Obsoletes and Provides should be removed after Fedora 31. (Actually, I made a mistake, they should be there for 2 releases after the last possible existence of the old name, so they should be removed after Fedora 32.)

    karma: -1 critpath: -1
  • OK, I don't know how many R packages there are in EPEL8 as I don't build anything for it. If this seems to be working there, feel free to push.

  • This works now.

    Also note that autopush got disabled from my earlier karma.

    karma: +1 #1727281: +1 #1744723: +1
  • This works now, and also fixes a few packages that don't load because they were built while 3.6.0 was a build override.

    Also note that autopush got disabled from my earlier karma.

    karma: +1 #1703495: +1 #1727281: +1 #1744723: +1
  • Pulling in R-rpm-macros means that all subsequent builds will gain automatic Requires, but none of the packages have been rebuilt with the corresponding Provides, like they were in F31 with the Mass Rebuild.

    if you want to backport this, feel free to do so, but that means you'll need to rebuild everything. I suggest not doing so, either removing R-rpm-macros from R-devel for <31, or we can disable automatic dependencies in R-rpm-macros on older releases.

    karma: -1
  • Pulling in R-rpm-macros means that all subsequent builds will gain automatic Requires, but none of the packages have been rebuilt with the corresponding Provides, like they were in F31 with the Mass Rebuild.

    if you want to backport this, feel free to do so, but that means you'll need to rebuild everything. I suggest not doing so, either removing R-rpm-macros from R-devel for <31, or we can disable automatic dependencies in R-rpm-macros on older releases.

    karma: -1
  • Pulling in R-rpm-macros means that all subsequent builds will gain automatic Requires, but none of the packages have been rebuilt with the corresponding Provides, like they were in F31 with the Mass Rebuild.

    if you want to backport this, feel free to do so, but that means you'll need to rebuild everything. I suggest not doing so, either removing R-rpm-macros from R-devel for <31, or we can disable automatic dependencies in R-rpm-macros on older releases.

    karma: -1
  • Pulling in R-rpm-macros means that all subsequent builds will gain automatic Requires, but none of the packages have been rebuilt with the corresponding Provides, like they were in F31 with the Mass Rebuild.

    if you want to backport this, feel free to do so, but that means you'll need to rebuild everything. I suggest not doing so, either removing R-rpm-macros from R-devel for <31, or we can disable automatic dependencies in R-rpm-macros on older releases.

    karma: -1
  • Well bugger, I didn't mean to do that; was trying to see what was going on with the search button not working.

  • This update has been unpushed.

  • Works and doesn't break existing packages; actually stuff built while it was in override is broken since it's not available yet, so definitely would like this in.

    karma: +1 #1703495: +1
  • See also koschei builds of R-openssl and R-jose from Rawhide when this package was built there.

  • This triggers a double-free in R-openssl. There's a hotfix available (which is technically leaky), but it'd be nice to submit that fix in the same update. Or see if upstream's https://github.com/openssl/openssl/issues/8375 can be patched.

    karma: -1
  • OK, that's because of a typo that created a directory instead of installing the file. You can delete the directory (it should be empty) to proceed, and I'll have to add a scriptlet to fix the upgrade process.

  • I don't use it directly, but ocrmypdf builds pass and test don't say they've skipped anything due to pdfminer, so it's good enough for me.

    karma: +1 #1657556: +1
  • The taskotron errors are actually real bugs:

    python2-seaborn.noarch: E: non-readable /usr/lib/python2.7/site-packages/seaborn/__pycache__/conftest.cpython-27-PYTEST.pyc 600
    

    These files should not be packaged, and I should have caught that in the PR I made. It can be fixed by setting PYTHONDONTWRITEBYTECODE=1 when running the tests.

  • Those 4 rebuilds have spent their week in testing; is this update going to get pushed to stable?

  • I ran:

    $ mock -r fedora-27-x86_64 --dnf-cmd -- install 'R-*' --skip-broken
    $ mock -r fedora-27-x86_64 --dnf-cmd list 'R-*' | awk '{print $1}' | sed -e 's/\.[^.]\+$//g' -e 's/^R-//' -e 's/-devel$//' | sort -u > pkgs
    $ mock -r fedora-27-x86_64 --copyin pkgs /
    $ mock -r fedora-27-x86_64 --shell
    # for p in `cat pkgs`; do echo "library($p); quit()" | R --quiet --vanilla; done &> results
    $ mock -r fedora-27-x86_64 --copyout /results results-f27
    

    then did it again with --enablerepo=updates-testing, and diffed the results.

    With this update, we are now able to also install:

    • R-blob, and
    • R-prettyunits.

    Also, now able to correctly load:

    • R-future,
    • R-listenv,
    • R-promises,
    • R-R.cache, and
    • R-R.rsp.

    But now these packages are broken:

    • R-chron,
    • R-COPASI,
    • R-GeneR,
    • R-quadprog,
    • R-reshape,
    • R-roxygen2, and
    • R-shogun.

    I have rebuilt R-chron, R-quadprog, R-reshape, and R-roxygen2, but I don't have permissions to build the others. If you don't want to wait another week, I can submit these as a separate update.

  • Ugh, I added +1s on a bunch of the FTBFS bugs, but bodhi didn't save it.