[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Transition to OCaml 3.12.0...



Hello,

Now that Squeeze is released, we can actively work on getting OCaml 3.12.0 into Debian... I've updated the coordination wiki page [1]. I've not yet set up monitoring pages, but we should be able to directly use Ben's services [2], now. For information, Ben is based on my transition monitoring tool and has been deployed for use by the Release Team by Mehdi Dogguy.

[1] http://wiki.debian.org/Teams/OCamlTaskForce/OCamlTransitions
[2] http://release.debian.org/transitions/

All packages not in the table should be safely binNMU-ed. The table summarizes work that will need to be done manually. Please have a look, check your packages, and put your name in the slots you want to handle yourself.

I based this table on data from the unofficial repository that I made available shortly after the release of OCaml 3.12.0 (on Sat, 24 Jul 2010 14:24:19 +0200). I didn't take into account all new uploads since then (nor to unstable, nor to experimental). Feel free to update whatever needs updating. Before packaging a new upstream version of some library, please be sure that all its reverse-dependencies compile with the new toolchain (use official packages or packages you compiled yourself with them and NOT whatever is available in my unofficial repository as they might be a bit outdated)... or just wait for OCaml 3.12.0 to be in testing before pushing new versions.

New in OCaml 3.12.0 from Debian's point-of-view:

 * All ocaml{objinfo,byteinfo,...} have been merged upstream into a
   single tool, ocamlobjinfo. dh-ocaml (as of 1.0.0) has been updated
   accordingly.

 * The native compiler has been ported to armel. I've activated it in
   my test repository, and almost everything works. Still, there remains
   some issues (see the wiki page) and native dynlink doesn't work. To
   cope with this, I've introduced in dh-ocaml 0.9.4 an additional
   variable OCAML_NATDYNLINK and support for "DYN:"-prefixed lines in
   *.in files. Most of the time, replacing "OPT:"-prefixed lines
   referring to .cmxs by "DYN:" ones suffices. If your application
   relies heavily on natdynlink, it's probably simpler to just not
   compile the native version.

Currently, there are very few packages that are not ready for this transition; basically jocaml and packages FTBFS'ing on armel. If nobody wants to take care of these promptly, I guess we could just drop them from testing so that they do not block the transition... what is your opinion on this?

Please feel free to provide feedback to any point mentioned in this mail.


Cheers,

--
Stéphane


Reply to: