[new release] dune (1.4.0)

[rgrinberg]

CHANGES:

• Do not fail if the output of ocamlc -config doesn't include
  standard_runtime (Do not fail if standard_runtime is missing in the output of ocamlc -config dune#1326, @diml)

• Let Configurator.V1.C_define.import handle negative integers
  (Configurator: handle negative numbers dune#1334, @Chris00)

• Re-execute actions when a target is modified by the user inside
  _build (Re-execute actions when a target is modified dune#1343, fix Dune doesn't recognize manual revert of executables dune#1342, @diml)

• Pass --set-switch to opam (Pass --set-switch to opam when opam 2.0.0 is detected dune#1341, fix dune triggers opam 2 to report spurious switch warnings dune#1337, @diml)

• Fix bad interaction between multi-directory libraries the menhir
  stanza (Fix #1372 dune#1373, fix include_subdirs + ocamllex + menhir dune#1372, @diml)

• Integration with automatic formatters (Integration with auto formatters dune#1252, fix ocamlformat integration dune#1201, @emillon)

• Better error message when using (self_build_stubs_archive ...) and
  (c_names ...) or (cxx_names ...) simultaneously.
  (Better error message when (self_build_stubs_archive ...) and (c_names ...) or (cxx_names ...) used simultaneously dune#1375, fix regression with 1.2.*: rule conflict dune#1306, @nojb)

• Improve name detection for packages when the prefix isn't an actual package
  (Better name detection dune#1361, fix jbuilder subst -p name leads to an error dune#1360, @rgrinberg)

• Support for new menhir rules (New interpretation of (menhir ...) stanzas. dune#863, fix jbuilder doesn't support menhir with --infer dune#305, @fpottier, @rgrinberg)

• Do not remove flags when compiling compatibility modules for wrapped mode
  (Do not reset flags for compatibility modules dune#1382, fix (wrapped (transition ...)) doesn't set flags correctly dune#1364, @rgrinberg)

• Fix reason support when using staged_pps (Fix staged_pps to work with reason dune#1384, @charlesetc)

• Add support for enabled_if in rule, menhir, ocamllex,
  ocamlyacc (Support enabled_if in more places dune#1387, @diml)

• Exit gracefully when a signal is received (Exit gracefully when a signal is received dune#1366, @diml)

• Load all defined libraries recursively into utop (Fix staged_pps to work with reason dune#1384, fix Load all project's libraries into utop dune#1344,
  @rgrinberg)

• Allow to use libraries bytes, result and uchar without findlib
  installed (meta builtins: add support for bytes/result packages dune#1391, @nojb)

• Take argument to self_build_stubs_archive into account. (self_build_stubs_archive: do not ignore argument dune#1395, @nojb)

• Fix bad interaction between env customization and vendored
  projects: when a vendored project didn't have its own env stanza,
  the env stanza from the enclosing project was in effect (Fix bad interaction between env stanza and vendored projects dune#1408,
  @diml)

• Fix stop early bug when scanning for watermarks (fix stop early bug when scanning for watermarks dune#1423, @diml)

[camelus]

☀️ All lint checks passed 9e08c98

• These packages passed lint tests: dune.1.4.0

---

☀️ Installability check (9775 → 9776)

• new installable packages (1): dune.1.4.0

[kit-ty-kate]

poke @mseri for the eigen/owl related problems: https://ci.ocamllabs.io/log/saved/docker-run-f22b1fa17d182dc0684d8dabe5c30828/11cc8cb6dc8255040ebf775a45da98d9a3bb40ef

Otherwise there seem to be a related problem that wasn't there before:

#=== ERROR while compiling jupyter-archimedes.2.3.2 ===========================#
# context              2.0.0 | linux/x86_64 | ocaml-base-compiler.4.07.0 | git+file:///home/opam/opam-repository
# path                 ~/.opam/4.07/.opam-switch/build/jupyter-archimedes.2.3.2
# command              ~/.opam/4.07/bin/dune build -p jupyter-archimedes -j 127
# exit-code            1
# env-file             ~/.opam/log/jupyter-archimedes-18-58c0d3.env
# output-file          ~/.opam/log/jupyter-archimedes-18-58c0d3.out
### output ###
# File "jupyter-archimedes/src/dune", line 10, characters 14-30:
# 10 |               archimedes.cairo))
#                    ^^^^^^^^^^^^^^^^
# Error: Library "archimedes.cairo" in /home/opam/.opam/4.07/lib/archimedes is hidden (unsatisfied 'exist_if').
# Hint: try: dune external-lib-deps --missing -p jupyter-archimedes @@default

[mseri]

Thanks @kit-ty-kate the issue with eigen is due to a too old gcc, unfortunately the install does not fail properly and makes the owl package unusable. We need to find a better way to deal with the external c++ libraries there, but unfortunately there seems to be no easy or right solution.

It can be safely ignored though as it is not related to dune

[mseri]

The archimedes failure seems to be there also with dune 1.3.1: https://ci.ocamllabs.io/log/saved/docker-build-92a97f1479c3e3d14a715749ec2aa264/dd2bcc2c1ef76359e2430188c28507b78a217251

[kit-ty-kate]

But the weird thing is that both didn't fail when tested for dune 1.3.0: https://ci.ocamllabs.io/ocaml/opam-repository/pr/12693?history=b093b1d937b5d9299bf3bb9ae9bdc200886016e5

[rgrinberg]

If I had to guess, it's probably a race condition. Is it possible to restart the build?

[pmetzger]

If I had to guess, it's probably a race condition. Is it possible to restart the build?

Yes, but you can do it on your own just by force pushing the branch even though it hasn't actually changed. That will trigger a new build.

[Chris00]

The archimedes failure seems to be there also with dune 1.3.1

Archimedes does not use Dune to compile so this is odd.

[mseri]

We have been imprecise. It is jupyter-Archimedes that fails because Archimedes.cairo appears to be an hidden library

[Chris00]

On 11 October 2018 at 16:36 GMT, Marcello Seri wrote: We have been imprecise. It is jupyter-Archimedes that fails because Archimedes.cairo appears to be an hidden library

I guess, this is because Archimedes predates the trend to have several packages for a given library so that dependencies are more explicit. jupyter-Archimedes should depend on cairo — but that may not completely fix the fact that Archimedes will compile archimedes.cairo when cairo is present, that will depend on the order of installation... ☹ The next version of Archimedes will be released as separate packages — but do not hold your breath, I intend to rewrite it from the ground up and split the functionality into several libraries.

[kit-ty-kate]

Yes, but you can do it on your own just by force pushing the branch even though it hasn't actually changed. That will trigger a new build.

as it's just from datakit ci, everyone can just relaunch the build by clicking "rebuild". It would avoid unnecessary rebuilds

[mseri]

Anyways it doesn’t really matter now as @Chris00 clarified how jupyter-archimedes will build correctly only if cairo happens to be compiled before archimedes. So this failure is not dune’s fault