Alex Kost writes: > Jan Nieuwenhuizen (2017-09-02 11:11 +0200) wrote: > >> --- >> gnu/packages/emacs.scm | 25 +++++++++++++++++++++++++ >> 1 file changed, 25 insertions(+) > Please use the source directly from the upstream, i.e.: > > https://github.com/szermatt/emacs-bash-completion/archive/v2.0.0.tar.gz Ok! > There is a problem with MELPA: when there will be a new commit in the > repo, the new melpa tarball will be generated and the previous one > ("20170901.1445") will be *removed*, so this package will lose its > source. Ugh...how reproducible! I didn't know (or investigate...), thanks for the explanation. >> + version >> + ".tar")) > > I wouldn't leave ‘version’ and ‘".tar"’ on separate lines, but it's a > matter of taste :-) I like your suggestion. >> + (sha256 >> + (base32 >> + "12ggygcfhny5sqq18zn6qyyqbh9314niywjagv9nclwxfa0i9pvr")))) >> + (native-inputs `(("bash" ,bash))) > > You probably meant to make this package call this "bash" (btw it should > be 'input', not 'native-input') from store, but this is not enough. Indeed. It took me quite a while of debugging until I finally stumbled upon bash-minimal. Google had a very hard time telling me how complete or compgen could not exist. > You also need to patch 'bash-completion-prog' variable using > 'emacs-substitute-variables' procedure (see (gnu packages emacs) > module for the examples). I have added something here and built package works and looks okay... Because adding bash to (native-)inputs already made that 'emacs-substitute-variables' is set to bash instead of bash-minimal, I don't really know how to test it. >> + (build-system emacs-build-system) >> + (home-page "https://github.com/szermatt/emacs-bash-completion") >> + (synopsis "BASH completion for the shell buffer") >> + (description >> + "@code{bash-completion} defines dynamic completion hooks for shell-mode >> +and shell-command prompts that are based on bash completion. >> +") > > Please put this quote and parenthesis to the previous line. Ok. >> + (license license:gpl3+))) > > I see GPL 2 or later in "bash-completion.el". I used guix import elpa --archive=melpa and didn't check. It seems we cannot rely on that! New version attached, still 13 of the 30 lines were unchanged, wow...thanks for all your helpful remarks! Greetings, janneke