GNU bug report logs - #26802
Single source file emacs packages get a ".el.el" extension

Previous Next

Package: guix-patches;

Reported by: Arun Isaac <arunisaac <at> systemreboot.net>

Date: Sat, 6 May 2017 12:53:01 UTC

Severity: normal

Done: Arun Isaac <arunisaac <at> systemreboot.net>

Bug is archived. No further changes may be made.

Full log


View this message in rfc822 format

From: Alex Kost <alezost <at> gmail.com>
To: Arun Isaac <arunisaac <at> systemreboot.net>
Cc: 26802 <at> debbugs.gnu.org
Subject: bug#26802: Single source file emacs packages get a ".el.el" extension
Date: Sun, 14 May 2017 20:15:36 +0300
Arun Isaac (2017-05-13 22:06 +0530) wrote:

> I've pushed the other uncontroversial patches.
>
>> It would be a linter only for *.el files, is that what you mean?  AFAIK
>> we do not have such specific linters, but yeah, why not, I think it's a
>> good idea.
>
> No, it will be a linter for all packages, not just emacs packages.

Hm, but the ".el" source files are special: they must have
"name-version.el" name, while the other sources may be named pretty
arbitrary.

> The following is the current linter for checking source file names. I
> have a couple of issues with it, in addition to the new linter feature I
> am suggesting. Let me explain.
>
>> (define (check-source-file-name package)
>>   "Emit a warning if PACKAGE's origin has no meaningful file name."
>>   (define (origin-file-name-valid? origin)
>>     ;; Return #t if the source file name contains only a version or is #f;
                 ^^    (it's a typo I think, should be ‘#f’ there)
>>     ;; indicates that the origin needs a 'file-name' field.
>
> Isn't this logic somehow backward? Should'nt a predicate called
> `file-name-valid?' return #t if the file name is valid, and #f
> otherwise?  This seems to be doing the opposite of that, returning #f if
> the file name is valid, and #t otherwise.

If I understand correctly, this predicate does the right thing; it's
just a typo in the commentary.

>>     (let ((file-name (origin-actual-file-name origin))
>>           (version (package-version package)))
>>       (and file-name
>>            (not (or (string-prefix? version file-name)
>>                     ;; Common in many projects is for the filename to start
>>                     ;; with a "v" followed by the version,
>>                     ;; e.g. "v3.2.0.tar.gz".
>>                     (string-prefix? (string-append "v" version) file-name))))))
>
> I think this check can be done by matching against a single regexp like
> so:
>
> (string-match (string-append "^v?" version) file-name)

I agree, a single regexp looks better!

> In addition to this logic, we add an extra condition that makes sure the
> version is some substring of the source file name, like so:
>
> (string-match version file-name)
>
> With this new check, single source file packages like
> emacs-transpose-frame, etc. which did not have file-name fields would
> have raised a lint warning.

I'm not sure, I think:

- it's too much for all the sources, as the upstream source may not
contain a version in the file name at all.  Do we really want to raise a
warning in this case?

- and it's not enough for ".el" sources, I mean "something-version.el"
is not enough, as the file name must exactly be "name-version.el" (as it
is with ELPA single-filed sources), so the emacs-build-system will
output "name.el" file which will correspond to 'name' feature provided
by this file.

[...]
>> P.S.  Could you please keep my email in Cc?  It would be easier to
>> follow the discussion for me, thanks.
>
> Sorry, I thought debbugs does some magic to send out mails to everybody
> involved in a specific bug. Didn't realize it was only using Cc. Will
> keep you in Cc for this and future mails.

Thanks!

-- 
Alex




This bug report was last modified 8 years and 87 days ago.

Previous Next


GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.