On Wed, Oct 14, 2015 at 11:04:47PM +0200, Julia Lawall wrote: > On Wed, 14 Oct 2015, Luis R. Rodriguez wrote: > > On Wed, Oct 14, 2015 at 10:53:50AM +0200, Sébastien Hinderer wrote: > > > Hi, > > > > > > Luis: here is what I am guessing. > > > > > > At some point in the past you have installed OCaml and then compiled > > > Menhir from sources with that version of OCaml. Meanwhile the OCaml > > > related packages got updated but Menhir has not been recompiled with > > > the updated version of OCaml, hence the inconsistency. > > > > > > My suggestion would thus be to make sure you recompile Menhir with the > > > currently installed version of OCaml. > > > > Sébastien, > > > > your guess was correct. I also tested using a new base OpenSUSE factory > > install for both github and inria git repos for coccinelle and found > > that: > > > > * github: works as expected > > * inria git: required menhir but fails with a compilation failure [0] > > > > Then with only the latest and greatest on inira git: > > > > * menhir: compiled > > * coccinelle: compiles > > > > Once I forced recompilation of menhir things worked well. The > > reason things seem to work on debian is menhir is packaged and > > updated regularly, and I suppose it might be updated as ocaml > > gets updated as otherwise Debian would run into similar issues, > > and if not I suspect they will once ocaml on Debian does get > > upgraded :) > > > > I wonder how we'll avoid these dependency issues creeping up moving > > forward. This is pretty fragile, and if the git tree will represent > > later the latest and greatest what should we do about mehir > > dependencies? > > I guess that it is normal that if you install something with your package > manager, and another thing without your package manager, then they will > get out of synch in strange ways. It doesn't really seem like > Coccinelle's problem, specifically.
The problem here stems from users of the internal inria git tree prior to a release, nothing more. For those users like myself the issue is real, and I am now aware of it, and so should others. Since we know we want to later strive towards one tree for development we should consider this a bit more seriously though. Perhaps once compromise might be to carry the pre-built files (as in github now) and *only* if a specific non-default 'make menhir-update' type of target is used would we force a re-generation of the files? Distributions should not need to run this, ony admins of the git tree. > Now that it has compiled, do you still have the problem with the > Common.union_set reference? Oh yes. And the instructions I gave can be used to reproduce that build I think as well now. To reproduce one can build coccinelle on opensuse (I guess inria git version), and then apply this draft patch onto linux-next (git am the file). http://drvbp1.linux-foundation.org/~mcgrof/coccinelle/common-union-issue/fw-warn-init-probe-draft-v2.patch Then try it as follows: # Note: cancel after 20 seconds otherwise, this patch is still # being worked on, the current iteration use will run for a long # time, don't waste your machine cycles / overheat your box. export COCCI=scripts/coccinelle/api/request_firmware.cocci make coccicheck MODE=report I get: Please check for false positives in the output before submitting a patch. When using "patch" mode, carefully review the patch before submitting it. File "/tmp/ocaml_cocci_27c72e.ml", line 21, characters 23-39: File "/tmp/ocaml_cocci_eb7247.ml", line 21, characters 23-39: File "/tmp/ocaml_cocci_82ec3c.ml", line 21, characters 23-39: File "/tmp/ocaml_cocci_7c147c.ml", line 21, characters 23-39: Error: Unbound value Common.union_setError: Unbound value Common.union_setError: Unbound value Common.union_set Error: Unbound value Common.union_set Fatal error: exception Yes_prepare_ocamlcocci.CompileFailure("/tmp/ocaml_cocci_27c72e.ml") Fatal error: exception Yes_prepare_ocamlcocci.CompileFailure("/tmp/ocaml_cocci_eb7247.ml") Fatal error: exception Yes_prepare_ocamlcocci.CompileFailure("/tmp/ocaml_cocci_7c147c.ml") Fatal error: exception Yes_prepare_ocamlcocci.CompileFailure("/tmp/ocaml_cocci_82ec3c.ml") > Sebastien seemed to think that the ability to > do dynamic linking in ocaml depends on more things than I would have > expected. But I odn't know the details. Interesting, let me know if you have any leads on a possible fix. In the meantime I'll be debugging the above iteration SmPL patch on Debian. Luis _______________________________________________ Cocci mailing list [email protected] https://systeme.lip6.fr/mailman/listinfo/cocci
