Ocamlbuild.0.10.1

[gasche]

Followup on #8155... The joy of software development :-)

[camelus]

✅ All lint checks passed 71f92ad

• These packages passed lint tests: ocamlbuild.0.10.1

---

✅ Installability check (5893 → 5893)

• new installable packages (1): ocamlbuild.0.10.1
• removed installable packages (1): ocamlbuild.0.10.0

[samoht]

Does it fix:

#=== ERROR while installing mtime.0.8.3 =======================================#
# opam-version 1.2.2
# os           linux
# command      ocaml pkg/build.ml native=true native-dynlink=true jsoo=false
# path         /home/travis/.opam/4.03.0/build/mtime.0.8.3
# compiler     4.03.0
# exit-code    10
# env-file     /home/travis/.opam/4.03.0/build/mtime.0.8.3/mtime-24235-865888.env
# stdout-file  /home/travis/.opam/4.03.0/build/mtime.0.8.3/mtime-24235-865888.out
# stderr-file  /home/travis/.opam/4.03.0/build/mtime.0.8.3/mtime-24235-865888.err
### stdout ###
# [...]
# ocamlfind ocamlopt -a -cclib -lmtime_stubs -cclib -lrt src-os/mtime_base.cmx src-os/mtime.cmx -o src-os/mtime.cmxa
# ocamlfind ocamlc -c -g -bin-annot -safe-string -I src-os -o src-os/mtime.cmo src-os/mtime.ml
# ocamlfind ocamlc -a -dllib -lmtime_stubs -cclib -lrt src-os/mtime_base.cmo src-os/mtime.cmo -o src-os/mtime.cma
# ocamlfind ocamlopt -shared -linkall src-os/libmtime_stubs.a src-os/mtime.cmxa -o src-os/mtime.cmxs
# + ocamlfind ocamlopt -shared -linkall src-os/libmtime_stubs.a src-os/mtime.cmxa -o src-os/mtime.cmxs
# /usr/bin/ld: cannot find -lmtime_stubs
# collect2: ld returned 1 exit status
# File "caml_startup", line 1:
# Error: Error during linking
# Command exited with code 2.

[samoht]

I have reverted 0.10.0 as it seems to break some existing packages. Was it expected?

[gasche]

Nope, breaking builds is never expected. Although it is not completely surprising either as 0.10 indeed changes the way .cmxs are built.

Thanks for reverting. I'll investigate locally with the repro-case and see what we should do.

"Jamais deux sans trois"...

[gasche]

I'll close the PR for now while I don't understand the regression better.

[whitequark]

@gasche Can we please get a version with the -toolchain option? This holds up many downstream cross-compilation improvements.

[gasche]

I'll try to do this before the end of the month, yes. This release is painful because of the .cmxs regression (we have to revert the changes before going through), but in general, for future releases, I would warmly welcome help / be glad let someone else do the release. (It has experimentally been determined that I am bad at doing releases.)

[whitequark]

This release is painful because of the .cmxs regression (we have to revert the changes before going through)

Actually, I don't think so. I believe the root cause of the regression is ocamlbuild not passing the necessary -I flags when building a cmxs, most likely caused by ocaml/ocamlbuild@89888f2c40.

[gasche]

This commit is part of the .cmxs changes I had in mind and plan to revert before doing a new release -- I should have released before merging them in the first place.

[whitequark]

@gasche Actually, it's weirder. I bisected the change to ocaml/ocamlbuild@f865497a8. It looks like -linkall wasn't passed before but it is passed now, which causes the immediate failure.

Also, you've added the tag "foo" in ocaml_specific.ml...

[whitequark]

@gasche I think I know what happens. Specifically, you're now linking a cmxa and a into a cmxs, instead of cmx and o. However, cmxa already has an autolink directive, so instead of the a, what's really needed is a -I flag.

[gasche]

I think that the discussion in ocaml/ocamlbuild#131 would give more context.

[whitequark]

@gasche At a glance that was just a mixture of more confusion and everything I already knew. Anyway, I already implemented a fix, PR incoming.

[whitequark]

@gasche Please review ocaml/ocamlbuild#159.