ocamlot triage 20130830 mainly for Linux 4.00.1 and some 4.1

[avsm]

Triage results to keep track of issues before fixing them all. Note that the link URLs will break later (we need a permalink mechanism for these -- I'll edit the issue with the version of ocamlot-state required to reconstruct these links later).

• [x] why3 broken because of coq pl2 (transient - coq server was down)
• [ ] fftw fails with system compiler, 4.00.1 system -- (can be fixed in opam 1.1 with new available field, see OCamlPro/opam#789)
• [x] ocaml-lua cannot find topfind on both 4.00.1 and 4.01 - failure is only on system compiler
• [x] websocket.0.6 needs an newer version of cohttp (vbmithr/ocaml-websocket#9)
• [x] xen-block-driver.0.2.2 has a dependency on mirage.0.9.4 (old), which fails due a mismatch on xenstore interface change. (mirage/ocaml-xen-block-driver#2)
• [x] opam2web fails to build with a linking error (OCamlPro/opam2web#45)
• [x] eliom is trying to create a /tmp/log for some reason - https://ocamlot.recoil.org/github/OCamlPro/opam-repository/task/2373 (fixed upstream in ocsigen/eliom#2)
• [x] sexplib.7.0.5 doesnt work with Toploop on 4 -- compiler constraint on 3.12.1 if superseded by Core version? https://ocamlot.recoil.org/github/OCamlPro/opam-repository/task/5637 (fixed in 94d667fd487eca8b54c293a452e48ef87e2b62e1 and ocamlot bug filed in ocamllabs/ocamlot#42)
• [ ] spotinstall - yet another omake error https://ocamlot.recoil.org/github/OCamlPro/opam-repository/task/2901
• [x] batteries.2.1.0 fails to build with O_CLOEXEC failure (opened as #1031)
• [x] libfgs has malformed log file on configure, so needs an oasis setup on ocaml4+
• [ ] planck.2.1.1 fails due to unbound module -- missing dep? https://ocamlot.recoil.org/github/OCamlPro/opam-repository/task/489
• [ ] mirage-www fails due to mirari unix problem

  4.01.00 specific breakage

• [x] zipperposition.0.2 broken with 4.01 due to signature change but works on 4.00.1 (c-cube/zipperposition#1)
• [x] mascot.1.0 fails with 4.1 (maintainer emailed, can add constraint to 4.0)
• [x] treeprint is ocaml 4.00.1 only due to transitive dependency on meta_conv which is restricted like that.
• [x] mlorg doesnt work with 4.01 since it has multiple duplicate method definitions in a single object
• [x] ocamleditor requires 4.00.[01] and fails on newer versions.
• [x] rml.1.08.06 fails with an ocamlbuild plugin error only in 4.01.0 (have mailed devel list)
• [x] ocamlspot has a compiler dependency that includes 4.1, but it only works 4.00.0 and 4.00.1 (mailed maintainer)
• [x] treeprint.1.0.2 is missing a dep on ocaml_conv -- omake error https://ocamlot.recoil.org/github/OCamlPro/opam-repository/task/2985
• [x] pa_do fails with camlp4 type mismatch, need patch for 4.1 version
• [x] pomap fails in 4.01 with incorrect signature for Set (needs new find function also)
• [ ] ocamlgraph fails on 4.01 due to warnings-as-errors from new warnings (backtracking/ocamlgraph#5)
• [ ] spotlib_js fails due to warnings-as-errors from 4.01 specific warnings
• [ ] ocaml-indent.1.1.0 fails in 4.01 due to compiler-libs type error
• [ ] quickcheck has warnings-as-errors failure due to new operator in Pervasives (open statement shadows |>)
• [ ] opa-base has warnings-as-errors (deprecated feature ||)
• [ ] lablgtk-extras has warnings-as-errors (deprecated feature ||)
• [ ] plasma has warnings-as-errors (deprecated feature ||)
• [ ] biocaml.0.2.0 depends on an older Core that doesnt have the CLOEXEC patch
• [ ] mesh.0.7.6 has Camlp4 Delete.Rule_not_found compile type error
• [x] pxp had too tight a dep on ocamlnet=3.6.0 (which doesnt work with 4.01). Possibly due to some change in how Netstring_pcre works.

  Missing deps

• [x] Benchmark 1.2 needs a depend on pcre https://ocamlot.recoil.org/github/OCamlPro/opam-repository/task/2889
• [x] cfstream is missing a dependency on core https://ocamlot.recoil.org/github/OCamlPro/opam-repository/task/2769
• [x] odisco is missing a dependency on zip
• [x] ibx missing a dependency on textutils
• [x] vhdformat missing a dep on ounit
• [x] flow.0.2 was picking up an old version of Core without CLOEXEC, so added firmer constraint (not really this packages fault)
• [x] xenctrl is missing dep on cmdliner
• [x] extprot is missing dep on omake
• [x] xenctrl is missing dep on cstruct

  Portability

• [x] facile uses make directly instead of $(MAKE) so fails on BSD
• [ ] numerix uses make directly too in OCamlPro/opam-repository#797
• [ ] crypt.1.2 fails on freebsd due to missing -lcrypt - missing reference to crypt symbol
• [ ] odisco tests exceed 32-bit integer literal limit (discoproject/odisco#3)
• [x] qmp tests exceed 32-bit integer literal limit (xapi-project/ocaml-qmp#3)
• [x] lilis.0.1.3 missing dep on menhir
• [ ] jenga dies on i386 due to Buffer size limit (janestreet/jenga#3) -- ocamlplugin is causing this
• [ ] core_extended uses x86-specific rtdsc timer so doesnt work on arm/macppc (janestreet/core_extended#3)
• [ ] posix_resource fails on OpenBSD/macppc with 'RLIMIT_AS' undeclared
• [ ] yajl fails on OpenBSD with missing cmake and some grep unportability
• [ ] Core.109.40.00 IOBuf stubs fail to build on OpenBSD (janestreet/core#26)
• [ ] json-wheel fails to link on OpenBSD due to missing pthread/sem library symbols
• [ ] numerix doesnt have a macppc/powerpc.S and so fails on that arch
• [ ] Core_extended fails to build on FreeBSD 9.1 due to missing quota.h

  Ext deps

• [ ] libvhd has an extdep
• [ ] opus has an extdep on libopus >= 1.0.0
• [ ] ocsfml needs extdep on boost https://ocamlot.recoil.org/github/OCamlPro/opam-repository/task/525
• [ ] sodium has an extdep on libsodium
• [ ] ocaml-usb has an extep

[pdonadeo]

Anil, can you elaborate more about ocaml-lua failure? I just installed the package with opam using OCaml 3.12.1, 4.00.1 and 4.01.0 on my Linux box and everything seems to run smoothly. It's an Ubuntu 13.04 and the three versions of OCaml have been installed via opam, no Debian packages.

[avsm]

I believe this happens only on the system compiler:

stdout

Cannot find file topfind.

stderr

Unknown directive `require'. E: Failure("Command './post_conf.ml' terminated with error code 2")

It happens across the board on 4.00.1 and 4.01.0 in the test runs. See: https://ocamlot.recoil.org/github/OCamlPro/opam-repository (temporary url, will disappear later today as I restart tests with RC1)

On 2 Sep 2013, at 21:37, Paolo Donadeo notifications@github.com wrote:

Anil, can you elaborate more about ocaml-lua failure? I just installed the package with opam using OCaml 3.12.1, 4.00.1 and 4.01.0 on my Linux box and everything seems to run smoothly. It's an Ubuntu 13.04 and the three versions of OCaml have been installed via opam, no Debian packages.

— Reply to this email directly or view it on GitHub.

[pdonadeo]

The script post_conf.ml assumes the presence of ocamlfind, because it's a dependancy of the package ocaml-lua. And topfind comes with ocamlfind, I believe.

If all this story is wrong, I have to find a workaround.

[avsm]

I've just confirmed that it doesn't work if it's using the 'system' compiler switch (with the earlier error), but does work fine if it's using a custom switch.

On 3 Sep 2013, at 09:20, Paolo Donadeo notifications@github.com wrote:

The script post_conf.ml assumes the presence of ocamlfind, because it's a dependancy of the package ocaml-lua. And topfind comes with ocamlfind, I believe.

If all this story is wrong, I have to find a workaround.

— Reply to this email directly or view it on GitHub.

[samoht]

btw, for the system compiler switch, "topfind" is stored into ~/.opam/system/lib/toplevel/topfind.

[avsm]

Right, and the environment variable OCAML_TOPLEVEL_PATH points to it, so that needs to be searched if it's set.

On 3 Sep 2013, at 09:32, Thomas Gazagnaire notifications@github.com wrote:

btw, for the system compiler switch, "topfind" is stored into ~/.opam/system/toplevel/topfind.

— Reply to this email directly or view it on GitHub.

[pdonadeo]

Ok, very easy to fix. I'll submit the patch ASAP.

[avsm]

thanks for the fix @pdonadeo; I'll let you know if it reappears; marked as fixed here.

[samoht]

Could we close this triaging run ?

[avsm]

Still a lot of failures, but I think it's easier to re-triage due to several version upgrades.