Hello, Could someone review and merge https://reviews.freebsd.org/D48228 ? I realize that it is a lot of changes at once. However, I tried very hard to make it as clean as possible (lots of rebase and rewrites on my side...sigh). It should be merged as a whole, that is all ocaml libraries should be rebuilt, due to the way ocaml records library dependencies and to account for the armv7 floating-point (architecture) change. I'm not sure if phabricator allows to export the serie as individual commits, if desired I can send the output of git format-patch. Unfortunately, there are changes which, as of now, have not yet reached the maintainer timeout: - ocaml-pcre (PR 283718) - ocaml-zip (PR 283713) - ocaml-lablgtk2 (PR 283714) - ocaml-ocurl (PR 283716) - ocaml-mysql (PR 283717) - ocaml-annexlib (PR 283718) - ocaml-images (PR 283720) - ocaml-parmap (PR 283814). I'm expecting the first 6 to go into timeout next week-end, because their maintainers have been consistenly not responsive for the past 6 months. The later two can be dropped from the serie (but their PORTREVISION must still be bumped), depending on when this request will be addressed and/or on how the committer feels. If I'm not mistaken, all the other changes are either approved, in (extensive) maintainer timeout or ports@ owned. Tested by rebuilding all ocaml ports on 14.2-RELEASE aarch64, 13.3-RELEASE armv7 and 15-CURRENT amd64. Thanks, Benjamin
Would you be interested in taking up maintainership of the ports with consecutive timeouts?
(In reply to Robert Clausecker from comment #1) Certainly, yes.
Had tried to apply the latest diff from the Phabricator D48228, https://reviews.freebsd.org/file/data/hdaxva3myulcrlmgoyb6/PHID-FILE-cofby45heskbnwcfh3va/D48228.diff Have rejections for these files: {{{ ./devel/ocaml-calendar/Makefile.rej ./devel/ocaml-calendar/distinfo.rej ./devel/ocaml-extlib/Makefile.rej ./devel/ocaml-lacaml/Makefile.rej ./devel/ocaml-lwt/Makefile.rej ./devel/ocaml-parmap/Makefile.rej ./devel/ocaml-pcre/Makefile.rej ./devel/ocaml-react/Makefile.rej ./math/ocamlgsl/Makefile.rej ./textproc/ocaml-text/Makefile.rej }}} devel/ocaml-calendar was updated to 2.04 recently: https://cgit.freebsd.org/ports/commit?id=edd0e13e73667f04b36b0e9a658f0095b5686da5 Other rejections are due to mismatch of diff hunks from Phabricator and the current ports tree. So, a diff in a 'git format-patch' format, rebased to the current tip of the ports tree will be very helpful. And my intention (I'll take this PR) would be to try to split the large code drop into more manageable pieces, unless there are some serious technical obstacles to this. For example, I see that WASH was removed from Uses/ocaml.mk and two ports -- x11-toolkits/ocaml-lablgtk2 and archivers/ocaml-zip (the above-mentioned patch from Phabricator is missing hunks for archivers/ocaml-zip/Makefile that remove wash, by the way): it would be a perfect patch that can be committed alone (if I am not missing anything here). Changes like USE_GITHUB -> MASTER_SITES can also be grouped together, but split from other modifications. Otherwise, I fear, the whole large patch makes too many changes at once and (as the example of update for devel/ocaml-calendar) aims at the moving target: not a very manageable thing, in my view. Anyway, a git-formatted patch will be rather helpful for me: I'll try to split the changes into the pieces myself (reviewing the changes along the way) and see what comes up. But if you see something that will prevent such a split -- please, say so. Thank you for your work and patience ;)
(In reply to Eygene Ryabinkin from comment #3) Hi Eygene, Thanks for looking into this issue. I'll rebase the changes this week-end. Regarding separating in smaller patches, the serie is already made a individual patches that can be applied invidually. As said, all patches of "maintained" packages have already been submitted to their maintainer. I'm not sure how I can make the work more likely to be accepted without introducing a lot of overhead for me and for the committer by tracking each patch separately. > large patch makes too many changes at once and (as the example of update for devel/ocaml-calendar) aims at the moving target: ocaml-calendar is not a moving target, it hadn't been updated since Tue Feb 19 16:27:34 2013, almost 12 years ago... This is pure luck that someone decided to update it just after I submitted an update (and not even to the latest version, go figure...). > OCAML_WASH Thanks for noticing an issue, i'll try to figure what is missing from the DR. FWIW, if support is removed from Mk/ocaml.mk, then nothing can go wrong by having a coupe of ports still using it (== it is dead code). The things was dubious from the start in my opinion. But agreed, this can be removed from the patch set.
Sorry I had trouble to find some time. So now I've done the rebase, and I'm busy testing new tree... I'll update ASAP. I now see what you mean by ocaml-zip, indeed it seems to a hunk got lost (and since the keyword conversion, a bad keyword actually prevents to build it contrary to what i wrote in the previous comment).
I've pushed the refreshed patched onto https://reviews.freebsd.org/D48228 - Fixed merge conflicts - removal of ocaml:wash on ocaml-zip - updated devel/ocaml-domain-name Tested on 15 amd64 by rebuilding all ocaml ports. Test on 15 armv7 is still in progress, looks good so far.
Note: I'll have a look at publishing this patchset as a git tree somewhere, if that can help getting stuff in a more piece-meal fashion.