fixes following core suite v0.11 #11628

trefis · 2018-03-22T15:05:49Z

I left the following packages broken:

camlhighlight (a failed build is logged here)
nocrypto (a failed build is logged here)
yaml

That last one doesn't have any constraint at the moment, looking at the code it opens Sexplib.Conv without depending on sexplib in the opam file: someone else will have to look at it.

For the other two, I'm not sure what's going on. The build fails because some cmi files are not available (because the right -I haven't been passed).
It might be an issue with the content of some of our META files (in which case it's a jbuilder issue I guess), or it might just be that their build system is confused: IDK.

avsm · 2018-03-22T15:08:47Z

i'll fix yaml, thanks :)

camelus · 2018-03-22T15:09:23Z

❗ opam-lint warnings `77f9d92`

frag.0.1.0 has some warnings:
- warning 27: No field 'remove' while a field 'install' is present, uncomplete uninstallation suspected
- warning 36: Missing field 'bug-reports'
- warning 37: Missing field 'dev-repo'
osm_xml.0.0.1 has some warnings:
- warning 36: Missing field 'bug-reports'
These packages passed lint tests: async_inotify.v0.10.0, async_inotify.v0.9.0, bistro.0.3.0, bitcoinml.0.2.4, bitcoinml.0.2, bitcoinml.0.3.0, bitcoinml.0.3.1, boomerang.1.1.0, capnp.3.2.0, cfstream.1.2.3, conduit-async.1.0.3, core-lwt.0.2.0, netml.0.1.0, prettiest.0.0.1, prettiest.0.0.2, session-postgresql-async.0.4.0, trie.0.1.1

✅ Installability check (8602 → 8602)

avsm · 2018-03-22T19:00:45Z

#=== ERROR while installing prettiest.0.0.1 ===================================#
# opam-version         1.2.2 (aa258ecc06d3aea5a67f442a4ffd23f2a457180b)
# os                   linux
# command              jbuilder build -p prettiest -j 2
# path                 /home/opam/.opam/4.05.0/build/prettiest.0.0.1
# compiler             4.05.0
# exit-code            1
# env-file             /home/opam/.opam/4.05.0/build/prettiest.0.0.1/prettiest-7-eecdc3.env
# stdout-file          /home/opam/.opam/4.05.0/build/prettiest.0.0.1/prettiest-7-eecdc3.out
# stderr-file          /home/opam/.opam/4.05.0/build/prettiest.0.0.1/prettiest-7-eecdc3.err
### stderr ###
#       ocamlc lib/.prettiest.objs/prettiest.{cmo,cmt} (exit 2)
# (cd _build/default && /home/opam/.opam/4.05.0/bin/ocamlc.opt -w -40 -g -bin-annot -I lib/.prettiest.objs -I /home/opam/.opam/4.05.0/lib/base -I /home/opam/.opam/4.05.0/lib/base/caml -I /home/opam/.opam/4.05.0/lib/base/shadow_stdlib -I /home/opam/.opam/4.05.0/lib/ppx_compare/runtime-lib -I /home/opam/.opam/4.05.0/lib/sexplib/0 -no-alias-deps -o lib/.prettiest.objs/prettiest.cmo -c -impl lib/prettiest.pp.ml)
# File "lib/prettiest.ml", line 68, characters 43-48:
# Error: Unbound value force
#     ocamlopt lib/.prettiest.objs/prettiest.{cmx,o} (exit 2)
# (cd _build/default && /home/opam/.opam/4.05.0/bin/ocamlopt.opt -w -40 -g -I lib/.prettiest.objs -I /home/opam/.opam/4.05.0/lib/base -I /home/opam/.opam/4.05.0/lib/base/caml -I /home/opam/.opam/4.05.0/lib/base/shadow_stdlib -I /home/opam/.opam/4.05.0/lib/ppx_compare/runtime-lib -I /home/opam/.opam/4.05.0/lib/sexplib/0 -no-alias-deps -o lib/.prettiest.objs/prettiest.cmx -c -impl lib/prettiest.pp.ml)
# File "lib/prettiest.ml", line 68, characters 43-48:
# Error: Unbound value force

trefis · 2018-03-22T19:14:35Z

That failure is happening with base v0.9, indeed force was added in v0.10.
I guess this package needs a lower bound as well...

kit-ty-kate · 2018-03-22T20:50:59Z

Good enough. I'll add the lower-bound in a followup pr

kit-ty-kate · 2018-03-22T21:21:36Z

Ping @pqwy & @hannesm for nocrypto (see the first message)

kit-ty-kate · 2018-03-22T21:23:32Z

Also, mirage has the same problem: https://ci.ocamllabs.io/log/saved/docker-build-79ef4b292a79e33f7957d93853f5ec9f/39d26910164aeef6b82a96d0dc54cbba42d96169

trefis · 2018-03-23T10:23:44Z

Ping @diml as well: could the failures we don't understand be caused by some changes in how jbuilder generates META files?

For the record: Ppx_sexp_conv_lib is defined as:

include Base.Exported_for_specific_use.Ppx_sexp_conv_lib

And in there, Sexp.t is = Sexplib0.Sexp.t = Atom of string | List of t list so I don't understand what's going on.

ghost · 2018-03-23T12:58:28Z

I don't think so, there have been no changes in generated META file for a while.

ghost · 2018-03-23T13:10:09Z

I think it's a problem with the hand-written META file for ipaddr.2.7.2 that is the problem:

version = "2.7.2"
description =
"A library for manipulation of IP (and MAC) address representations"
requires = "bytes sexplib"

i.e. it doesn't mention ppx_sexp_conv.runtime-lib, as a result the compiler doesn't see ppx_sexp_conv_lib.cmi when compiling mirage. Ipaddr.2.8.0 is using jbuilder so its META file will be correct, I suggest to simply bump the version constraint on ipaddr.

kit-ty-kate · 2018-03-23T14:28:23Z

@diml according to this log this is not the issue: https://travis-ci.org/ocaml/opam-repository/jobs/355944133

ghost · 2018-03-23T14:33:50Z

It definitely was for mirage, I reproduced the error on my machine, and upgrading ipaddr to 2.8.0 fixed the issue

ghost · 2018-03-23T15:12:05Z

I found the issue for nocrypto: when creating the packed module (with ocamlfind ocamlopt -pack), the build system doesn't pass the -package .. options. If I run the same command with -package ppx_sexp_conv,sexplib the build succeeds.

@pqwy you should try jbuilder for nocrypto :)

ocamlbuild should pass -package(...) flags to ocamlfind when building a -pack-ed file, see ocaml/opam-repository#11628 (comment)

Binaries in <bench/*>, <tests/*> depend on ppx_sexp_conv's runtime library within ppx_sexp_conv. The packed modules <src/nocrypto.cm{x,o}> also depend on the package ppx_sexp_conv: its presence at pack-creation time influences the generated .cmi interface, see ocaml/opam-repository#11628 (comment) Note: the package ppx_sexp_conv.runtime-lib would suffice, but it is only available as such under recent ppx_sexp_conv versions, so its explicit use would make the build description (needlessly) incompatible with older ppx_sexp_conv versions.

ocamlbuild should pass -package(...) flags to ocamlfind when building a -pack-ed file, see ocaml/opam-repository#11628 (comment)

Binaries in <bench/*>, <tests/*> depend on the runtime libraries within ppx_sexp_conv and ocplib-endian. The packed modules <src/nocrypto.cm{x,o}> also depend on the package ppx_sexp_conv: its presence at pack-creation time influences the generated .cmi interface, see ocaml/opam-repository#11628 (comment) Note: the package ppx_sexp_conv.runtime-lib would suffice, but it is only available as such under recent ppx_sexp_conv versions, so its explicit use would make the build description (needlessly) incompatible with older ppx_sexp_conv versions.

Instead of a separate category, "pack" is now another form of linking like "program", "library", "toplevel" and "output_obj". This fixes the issue that package(...) tags where not passed at pack-production time, spotted by Jérémie Dimino in <ocaml/opam-repository#11628 (comment)> <janestreet/ppx_sexp_conv#20> More generally, this extends the meaning of all "link" flags to "pack", which seems to be the correct behavior for all the rules we inspected. I tested that this change fixes the original 'nocrypto' issue (once their _tags is also fixed to provide the right package(..) option at pack-time). This is a somewhat invasive change to the behavior of "pack" and "link" tags, which may affect all user-defined flag declarations. An alternative to this change would be to just add a specific "package" rule when the "pack" tag is present, but my understanding is that linking options quite generally apply to "pack" productions (that really behave mostly like library-archive production already under the ("link";"library") regime), so this more invasive change is more correct: the extra cases where it applies were likely to be small bugs before. (I checked that ocb-stubblr is orthogonal to this change and does not need any update.)

ocamlbuild should pass -package(...) flags to ocamlfind when building a -pack-ed file, see ocaml/opam-repository#11628 (comment)

Binaries in <bench/*>, <tests/*> depend on the runtime libraries within ppx_sexp_conv and ocplib-endian. The packed modules <src/nocrypto.cm{x,o}> also depend on the package ppx_sexp_conv: its presence at pack-creation time influences the generated .cmi interface, see ocaml/opam-repository#11628 (comment) Note: the package ppx_sexp_conv.runtime-lib would suffice, but it is only available as such under recent ppx_sexp_conv versions, so its explicit use would make the build description (needlessly) incompatible with older ppx_sexp_conv versions.

ocamlbuild should pass -package(...) flags to ocamlfind when building a -pack-ed file, see ocaml/opam-repository#11628 (comment)

Binaries in <bench/*>, <tests/*> depend on the runtime libraries within ppx_sexp_conv and ocplib-endian. The packed modules <src/nocrypto.cm{x,o}> also depend on the package ppx_sexp_conv: its presence at pack-creation time influences the generated .cmi interface, see ocaml/opam-repository#11628 (comment) Note: the package ppx_sexp_conv.runtime-lib would suffice, but it is only available as such under recent ppx_sexp_conv versions, so its explicit use would make the build description (needlessly) incompatible with older ppx_sexp_conv versions.

fix following v0.11

1179604

kit-ty-kate added 2 commits March 22, 2018 16:05

Restrict some packages to < v0.10

d603942

boomerang: Add an OCaml upper-bound

77f9d92

kit-ty-kate merged commit e7296aa into ocaml:master Mar 22, 2018

kit-ty-kate mentioned this pull request Mar 23, 2018

nocrypto: restrict to <v0.11.0 core release #11636

Merged

kit-ty-kate mentioned this pull request Mar 23, 2018

Make ipaddr.2.7.x incompatible with sexplib >= v0.11 #11637

Merged

hannesm mentioned this pull request Mar 26, 2018

Incompatible with sexplib/ppx_sexp_conv v0.11.0 mirleft/ocaml-nocrypto#143

Closed

This was referenced Mar 26, 2018

build failure of nocrypto.0.5.9 probably caused by ppx_sexp_conv janestreet/ppx_sexp_conv#20

Closed

the pack action must also pass package(...) options to ocamlfind ocaml/ocamlbuild#272

Closed

gasche added a commit to gasche/ocaml-nocrypto that referenced this pull request Mar 26, 2018

pack+package: workaround ocamlbuild#272

39a20d4

ocamlbuild should pass -package(...) flags to ocamlfind when building a -pack-ed file, see ocaml/opam-repository#11628 (comment)

gasche added a commit to gasche/ocaml-nocrypto that referenced this pull request Mar 26, 2018

pack+package: workaround ocamlbuild#272

ee7305f

ocamlbuild should pass -package(...) flags to ocamlfind when building a -pack-ed file, see ocaml/opam-repository#11628 (comment)

gasche mentioned this pull request Mar 26, 2018

fix for ocamlbuild's pack+ocamlfind handling mirleft/ocaml-nocrypto#144

Open

gasche mentioned this pull request Mar 26, 2018

add the "link" tag to "pack" actions (fixes #272) ocaml/ocamlbuild#273

Merged

hannesm pushed a commit to hannesm/ocaml-nocrypto that referenced this pull request May 23, 2018

pack+package: workaround ocamlbuild#272

c1bdff8

ocamlbuild should pass -package(...) flags to ocamlfind when building a -pack-ed file, see ocaml/opam-repository#11628 (comment)

hannesm pushed a commit to hannesm/ocaml-nocrypto that referenced this pull request Jan 12, 2019

pack+package: workaround ocamlbuild#272

1913ec4

ocamlbuild should pass -package(...) flags to ocamlfind when building a -pack-ed file, see ocaml/opam-repository#11628 (comment)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fixes following core suite v0.11 #11628

fixes following core suite v0.11 #11628

trefis commented Mar 22, 2018

avsm commented Mar 22, 2018

camelus commented Mar 22, 2018 •

edited

Loading

avsm commented Mar 22, 2018

trefis commented Mar 22, 2018

kit-ty-kate commented Mar 22, 2018

kit-ty-kate commented Mar 22, 2018

kit-ty-kate commented Mar 22, 2018

trefis commented Mar 23, 2018 •

edited

Loading

ghost commented Mar 23, 2018

ghost commented Mar 23, 2018

kit-ty-kate commented Mar 23, 2018

ghost commented Mar 23, 2018 •

edited by ghost

Loading

ghost commented Mar 23, 2018

fixes following core suite v0.11 #11628

fixes following core suite v0.11 #11628

Conversation

trefis commented Mar 22, 2018

avsm commented Mar 22, 2018

camelus commented Mar 22, 2018 • edited Loading

❗ opam-lint warnings 77f9d92

✅ Installability check (8602 → 8602)

avsm commented Mar 22, 2018

trefis commented Mar 22, 2018

kit-ty-kate commented Mar 22, 2018

kit-ty-kate commented Mar 22, 2018

kit-ty-kate commented Mar 22, 2018

trefis commented Mar 23, 2018 • edited Loading

ghost commented Mar 23, 2018

ghost commented Mar 23, 2018

kit-ty-kate commented Mar 23, 2018

ghost commented Mar 23, 2018 • edited by ghost Loading

ghost commented Mar 23, 2018

camelus commented Mar 22, 2018 •

edited

Loading

❗ opam-lint warnings `77f9d92`

trefis commented Mar 23, 2018 •

edited

Loading

ghost commented Mar 23, 2018 •

edited by ghost

Loading