bug#67930: 29.1; emacs 29.1 follows symlinks when a grep result is selected

[Eli Zaretskii]

> From: "DE BACKER Jurgen (EXT)" <jurgen.de-backer.ext@eurocontrol.int>
> CC: "67930@debbugs.gnu.org" <67930@debbugs.gnu.org>
> Date: Fri, 22 Dec 2023 16:20:42 +0000
> 
> > From: Eli Zaretskii <eliz@gnu.org>
> > Sent: Wednesday, December 20, 2023 7:47 PM
> > To: DE BACKER Jurgen (EXT) <jurgen.de-backer.ext@eurocontrol.int>
> > Cc: 67930@debbugs.gnu.org
> > Subject: Re: bug#67930: 29.1; emacs 29.1 follows symlinks when a grep result
> > is selected
> >
> > Thanks, but I don't think this is a matter of user preferences.  We replaced
> > expand-file-name there by file-truename because in some cases the former
> > doesn't work: it expands to a file that doesn't exist.  See bug#8035, where
> > such cases are presented.  We cannot ask the user to set or reset this 
> > option
> > each time they need to work with these or those file names.
> >
> > So I don't think the fix you propose is the right one.  I think we need to 
> > use
> > expand-file-name where it works, and file-truename where expand-file-
> > name doesn't work.  Or maybe just try expand-file-name first, and if that
> > produces a file name that fails file-exists-p, try file-truename.
> >
> > Would you like to propose and test a patch along these lines?
> 
> Hi Eli,
> 
> Please find attached a patch that does this: first we  try to expand the file 
> name with expand-file-name
> and if this fails, retry with file-truename.

Thanks.  I have a couple of minor comments:

> # In emacs 29.1, when clicking on a file link in the *grep* results buffer,
> # symlinks are resolved to the actual filename(s). In some cases this is not 
> preferable.
> #
> # This patch -1 will first use the previous behaviour (using 
> expand-file-name) to resolve a (relative) filename
> #            -2 if this fails (i.e. the result is not an existing file), we 
> try with the new behaviour 
> #                which uses function file-truename.
> #
> # In emacs 29.1 expand-file-name was replaced by file-truename to solve bug 
> #8035, 
> #   because in specific cases expand-file-name returned a wrong/non existing 
> file.
> # Unlike expand-file-name, file-truename follows symlinks.

Could you please reformat this log message using our style?  The
details can be found in the file CONTRIBUTE in the Emacs Git
repository.

> --- lisp/progmodes/compile.el.orig    2023-02-11 12:06:09.000000000 +0000
> +++ lisp/progmodes/compile.el 2023-12-21 17:22:34.000688000 +0000
> @@ -3108,7 +3108,7 @@
>          (spec-dir (if directory
>                        (expand-file-name directory)
>                      default-directory))
> -        buffer thisdir fmts name)
> +        buffer thisdir fmts expandedname name)
>      (if (and filename
>               (file-name-absolute-p filename))
>          ;; The file name is absolute.  Use its explicit directory as
> @@ -3122,8 +3122,10 @@
>              fmts formats)
>        ;; For each directory, try each format string.
>        (while (and fmts (null buffer))
> -        (setq name (file-truename
> -                    (file-name-concat thisdir (format (car fmts) filename)))
> +        (setq expandedname (expand-file-name (format (car fmts) filename) 
> thisdir)
> +              name (if (file-exists-p expandedname) ;; See bug #8035 
> expand-file-name fails in specific cases
> +                       expandedname
> +                     (file-truename (file-name-concat thisdir (format (car 
> fmts) filename))))
>                buffer (and (file-exists-p name)
>                            (find-file-noselect name))
>                fmts (cdr fmts)))

Your changes repeat the same pattern several times, so I think it
would be better to factor out this code into a separate function, and
then call that function in those places.  Would you like to rewrite
the patch along these lines?

Thanks again for working on this.