As jer suggested in bug #179895 it is a good idea to have them optional rather than having them always built and having arch specific code in ocaml ebuild for arches that do not support it.
However this raises several issues:
- ebuilds/packages that assume ocamlopt is present have to be fixed
- some of those packages can still build without native code support using ocamlc, or ocamlc -custom. The latter triggers some other issues: it embeds ocamlrun (the interpreter) and the bytecode, if you strip them, you lose the bytecode.
- For some weird reasons I dont fully understand (yet?) running debugedit on such files also remove the bytecode.
okey I've made ocamlopt an useflag on >=dev-lang/ocaml-3.10
as you can see I've not yet finished to check every pacakge, but most of them should be sane.
@ppc64: as previously ocamlopt was disabled in ocaml ebuild by some arch specific code, please use.mask ocamlopt useflag on 64bits userland profiles to keep the same behavior.
Moreover, from my checks all dev-ml/findlib versions failed on ppc64, I've fixed 1.2.1 to have the ocamlopt useflag; as stable version should be broken, if you want to have 1.2.1 straight stable that's your call; I'll file a stabilisation bug in some time anyway.
ping ppc64: on my g5 I cant build neither dev-ml/findlib nor dev-ml/facile...
I dunno how you managed to build native code support for dev-ml/facile, and kdeedu/kalzium need native code support for facile to build the solver, so if you plan to disable ocamlopt support, you'll also have to mask the solver useflag for those kde packages.
You seem to have tested it since facile was keyworded recently, but as far as I know, ocaml does *not* support powerpc64 elf target...
who the hell marked those packages stable? (it was me... seems like I fscked it up.)
gimme some time to look into it. I wouldn't mind if you want to fix it yourself, as I'm not using ocaml personaly.
dev-libs/gmetadom has IUSE=ocaml and fails when ocamlopt is missing, even though configure specially checks for it.
(In reply to comment #4)
> dev-libs/gmetadom has IUSE=ocaml and fails when ocamlopt is missing, even
> though configure specially checks for it.
hmmm seems like I had checked reverse rdepends but not reverse build depends...
@markus: it *should* be fixed now, it just needs ocamlopt use masked + the kde solver also use masked (for the reverse deps of dev-ml/facile) and dev-ml/findlib 1.2.1 going stable (because its the one having ocamlopt optional)
(In reply to comment #5)
> @markus: it *should* be fixed now, it just needs ocamlopt use masked + the kde
> solver also use masked (for the reverse deps of dev-ml/facile) and
> dev-ml/findlib 1.2.1 going stable (because its the one having ocamlopt
thanks a lot!
so for ppc64 we've used masked ocamlopt for now. It'll take some effort to support ocaml... perhaps someday .. but not quite yet.
Going to drop us from this bug .. and we'll reopen something related to ocaml when I step up and fix or upstream addresses. Hopefully this sounds acceptable.
we can close this one now I think