On Mon, Sep 23 2019, Jeremie Courreges-Anglas <j...@wxcvbn.org> wrote: > On Mon, Sep 23 2019, Christopher Zimmermann <chr...@openbsd.org> wrote: >> Here is an updated diff fixing the reported problems: >> - no more duplicate entries. They sneaked in while I updated the PLISTS >> for bytecode-only ocaml >> - update the distfile (again ?!?), use https. > > I guess you checked the tarball for changes, did you spot anything evil? > > The ocaml-graphics port ought to use the proper .tbz tarballs published > at > > https://github.com/ocaml/graphics/releases > > instead of GH_TAGNAME.
Updated ocaml-graphics tarball, using the tarball from upstream.
lib/ocaml/graphics/META, lib/ocaml/graphics/dune-package and
lib/ocaml/graphics/opam now specify a version ("5.0.0").
ocaml-graphics.tgz
Description: Binary data
> Was ocaml-4.08.1+ tested at all on a non-x86 arch? The update to > ocaml-4.08.1 is broken since its import on at least arm64 and sparc64, > despite my request for a wip diff. :-/ > > http://build-failures.rhaalovely.net//sparc64/2019-09-18/lang/ocaml.log It turns out that this update to 4.09.0 unbreaks ocaml on sparc64 (and probably others): the .cmt* files are now installed. I hit one problem with findlib though (not sparc64 specific). The graphics META file is now shipped by ocaml-graphics. This updated diff (attached) fixes this, removes the hardcoding of lang/ocaml=4.09.0, and strips an outdated comment in graphics/ocaml-cairo.
ocaml49.3.diff.gz
Description: Binary data
I've built a bunch of ocaml ports on amd64 and sparc64, no fallout. If you agree with this diff, ok jca@ -- jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE
