bug#13006: recognizing compressed files with arbitrary names

bug#13006: recognizing compressed files with arbitrary names

[Karl Berry]

As the documentation says
(http://www.gnu.org/software/emacs/manual/html_node/emacs/Compressed-Files.html)
"Emacs recognizes compressed files by their file names."

But compressed files don't always have the expected names.  For
instance, using the autobackup feature of Emacs or many other programs
will result in names ending in, say, .~1~ ... Example for concreteness:

$ echo foo | gzip >/tmp/foo.gz.~1~
$ emacs --no-site-file --no-init /tmp/foo.gz.~1~

And observe binary junk instead of text.  (This is with 24.2.)

Another implication is problems using M-x grep on such files.  After
changing the program to zgrep from grep, matches in the *grep* buffer
are displayed as usual, but then when one visits the file in Emacs,
it's junk.

It would be nice to at least have an easy way to manually tell Emacs
"this is a compressed file, despite its weird name".  If there's a way
to do that now, I'd like to know (and it should presumably be in the
manual).  I looked at jka-compr.el but could not find it.

Of course, a more robust solution would be to use magic numbers instead
of filenames to recognize file types.  I see there was a long thread
about that several years ago, starting at
http://lists.gnu.org/archive/html/emacs-devel/2009-08/msg01023.html.  I
know it's been discussed at other times, too, but I can't see any way to
accomplish it (short of using crypt++, which doesn't work well with
Emacs 23 or later, which is why I stopped using it ... anyway).

Thanks,
karl

bug#13006: recognizing compressed files with arbitrary names

[Andreas Schwab]

karl@freefriends.org (Karl Berry) writes:

> $ echo foo | gzip >/tmp/foo.gz.~1~
> $ emacs --no-site-file --no-init /tmp/foo.gz.~1~
>
> And observe binary junk instead of text.  (This is with 24.2.)

This is a regression from Emacs 23.  It should disregard the backup
suffix when matching the extension.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#13006: recognizing compressed files with arbitrary names

[Glenn Morris]

Andreas Schwab wrote:

> This is a regression from Emacs 23.  It should disregard the backup
> suffix when matching the extension.

Presumably the 2010-12-08 change to jka-compr-compression-info-list.
(There is a comment above jka-compr-compression-info-list which was
ignored and no longer applies. Also the :version was not bumped.)
Anyway, this seems to fix the immediate problem:

*** lisp/jka-cmpr-hook.el 2012-01-29 17:41:43 +0000
--- lisp/jka-cmpr-hook.el 2012-11-27 18:35:37 +0000
***************
*** 109,114 ****
--- 109,115 ----
    "Return information about the compression scheme of FILENAME.
  The determination as to which compression scheme, if any, to use is
  based on the filename itself and `jka-compr-compression-info-list'."
+   (setq filename (file-name-sans-versions filename))
    (catch 'compression-info
      (let ((case-fold-search nil))
        (dolist (x jka-compr-compression-info-list)


There's a separate question of handling arbitrarily named compressed
files via magic-fallback-mode-alist or somesuch.

bug#13006: recognizing compressed files with arbitrary names

[Karl Berry]

    +   (setq filename (file-name-sans-versions filename))

Thanks.

    There's a separate question of handling arbitrarily named compressed
    files via magic-fallback-mode-alist or somesuch.

Right.  The backup versions was just an example of problematic names.
Though it's certainly good to have that fixed.

There's also the (potentially related) question of making it possible
for users to tell Emacs "this is a compressed file".  Regardless of the
method(s) used to detect compression, this seems like it would be
useful.  I have plenty of times been forced to decompress files outside
of Emacs and then visit the plain file.  Clearly it's not the end of the
world to do that, but ...

If you want me to send separate reports for either or both of those
things, let me know.  

Thanks,
k

bug#13006: recognizing compressed files with arbitrary names

[Glenn Morris]

Glenn Morris wrote:

> +   (setq filename (file-name-sans-versions filename))

Applied for 24.2.91.

Leaving this open as a wishlist for:
a) detecting compressed files by magic bytes
b) explicit command to open a file using a given compression