Re: Unresolable collision between ocaml packages

[Brett Gilio]

Julien Lepiller writes:

> Le 6 juin 2019 08:06:42 GMT+02:00, Julien Lepiller <address@hidden> a écrit :
>>Le 5 juin 2019 22:21:37 GMT+02:00, Brett Gilio <address@hidden> a
>>écrit :
>>>Hi all,
>>>
>>>I like to make meta-packages that I can pass to `guix environment
>>>--ad-hoc` for development. I have the following meta package for ocaml
>>>
>>>(define-public meta-ocaml
>>>  (package
>>>   (inherit %meta-base)
>>>   (name "meta-ocaml")
>>>   (propagated-inputs
>>>    `(("meta-emacs" ,meta-emacs)
>>>      ("darcs" ,darcs)
>>>      ("dune" ,dune)
>>>      ("emacs-tuareg" ,emacs-tuareg)
>>>      ("gcc-toolchain" ,gcc-toolchain)
>>>      ("m4" ,m4)
>>>      ("make" ,gnu-make)
>>>      ("mercurial" ,mercurial)
>>>      ("ocaml" ,ocaml)
>>>      ("ocaml-base" ,ocaml-base)
>>>      ("ocaml-core" ,ocaml-core)
>>>      ("ocaml-ctypes" ,ocaml-ctypes)
>>>      ("ocaml-findlib" ,ocaml-findlib)
>>>      ("ocaml-merlin" ,ocaml-merlin)
>>>      ("ocaml-ounit" ,ocaml-ounit)
>>>      ("oocaml-parsexp"
>>>       ,ocaml-parsexp)
>>>      ("ocaml-utop" ,ocaml-utop)
>>>      ("opam" ,opam)
>>>      ("rsync" ,rsync)))))
>>>
>>>
>>>When trying to use that in an environment, I am getting the following
>>>backtrace claiming there is a collision that I assume is not getting
>>>resolved in its usual fashion.
>>>
>>>Backtrace:
>>>           6 (primitive-load
>>"/gnu/store/k45l4ph58xakp3f52kmridr9wyc…")
>>>In guix/build/profiles.scm:
>>>   157:2  5 (build-profile "/gnu/store/c2khaalrpy7cdyr8k4lawgqnbzs…"
>>…)
>>>In unknown file:
>>>          4 (hash-for-each #<procedure 105c180 at guix/build/union…>
>>…)
>>>          3 (hash-for-each #<procedure 10b3900 at guix/build/union…>
>>…)
>>>          2 (hash-for-each #<procedure 110b180 at guix/build/union…>
>>…)
>>>          1 (hash-for-each #<procedure 14368c0 at guix/build/union…>
>>…)
>>>          0 (scm-error misc-error #f "~A ~S" ("union-build: col…" …)
>>…)
>>>
>>>ERROR: In procedure scm-error:
>>>union-build: collision between file and directories ((files
>>>("/gnu/store/h1w5b2l1wny94cnfqm46w12nm3w8wz2k-ocaml-num-1.1/lib/ocaml/site-lib/stublibs"))
>>>(dirs
>>>("/gnu/store/93lp6k3d58dfc8gvx9zd70szapsb5kv6-ocaml-base-0.11.1/lib/ocaml/site-lib/stublibs"
>>>"/gnu/store/7vrg8apsnl6hg9vsszpgcq4ymvz9whap-ocaml-core-0.11.3/lib/ocaml/site-lib/stublibs"
>>>"/gnu/store/iz6wzz9sm4qnal2wlqi66blm8mb35zxv-ocaml-core-kernel-0.11.1/lib/ocaml/site-lib/stublibs"
>>>"/gnu/store/zrpcrfrmyanhqpvhxkclpgiq2sp7k1ip-ocaml-bin-prot-0.11.0/lib/ocaml/site-lib/stublibs"
>>>"/gnu/store/n70q0ypcv74vpwdd0354rjxi7mv9y1rf-ocaml-ppx-inline-test-0.12.0/lib/ocaml/site-lib/stublibs"
>>>"/gnu/store/q42ayh04wy70pl2xdm2yp2i0s8yh5g7b-ocaml-ppx-expect-0.12.0/lib/ocaml/site-lib/stublibs"
>>>"/gnu/store/cwijnaplps7zbby3xq21spjk4fgxpd35-ocaml-spawn-0.13.0/lib/ocaml/site-lib/stublibs"
>>>"/gnu/store/09b6pfbw3n1j2qqni822wjjixh48qwfp-ocaml-lambda-term-1.13/lib/ocaml/site-lib/stublibs"
>>>"/gnu/store/zlibrm9b6pa8hgwmjigzwm2cq3kb4y1r-ocaml-lwt-4.1.0/lib/ocaml/site-lib/stublibs"
>>>"/gnu/store/qzkydyr8rrbpfk2lfjynnhgdbavf4kh0-ocaml-ssl-0.5.7/lib/ocaml/site-lib/stublibs"
>>>"/gnu/store/sqkcaap13wgbjdqnjia8hi12rf2p5qf1-ocaml-dose3-5.0.1/lib/ocaml/site-lib/stublibs"
>>>"/gnu/store/wnl6maf9s6ksvszyy9q2bz0psiy1h15j-ocaml-mccs-1.1+9/lib/ocaml/site-lib/stublibs")))
>>>
>>>
>>>If formatting is bad in this email, here are plaintext pastes.
>>>
>>>meta-packages.scm : http://ix.io/1L1A
>>>
>>>backtrace : http://ix.io/1L1B
>>
>>Looks like ocaml-num's stubslib is a file, but it should be a
>>directory. I'll investigate, thanks for reporting this!
>
> This is now fixed in master. I also found a reproducibility issue in cmdliner 
> while making sure my change didn't break anything and it's also fixed in 
> Guix, and a patch was sent upstream :)
>
> So you can run guix pull and enjoy your metapackage!

Thanks Julien!