GNU bug report logs - #55166
28.1; tar-mode doesn't re-compress when saving remote files

Previous Next

Full log

Message #11 received at 55166 <at> debbugs.gnu.org (full text, mbox):

From: Michael Albinus <michael.albinus <at> gmx.de>
To: Lars Ingebrigtsen <larsi <at> gnus.org>
Cc: 55166 <at> debbugs.gnu.org, Dan McCarthy <daniel.c.mccarthy <at> gmail.com>
Subject: Re: bug#55166: 28.1; tar-mode doesn't re-compress when saving
 remote files
Date: Tue, 03 May 2022 14:36:53 +0200

Lars Ingebrigtsen <larsi <at> gnus.org> writes:

Hi,

>> dan <at> solstice:/tmp$ touch a b c
>> dan <at> solstice:/tmp$ tar cf test.tar a b c
>> dan <at> solstice:/tmp$ bzip2 -z test.tar
>> dan <at> solstice:/tmp$ file test.tar.bz2
>> test.tar.bz2: bzip2 compressed data, block size = 900k
>>
>> If you visit that remote file and save it, the result will have the .bz2
>> extension but won't actually be compressed:
>>
>> dan <at> solstice:/tmp$ file test.tar.bz2
>> test.tar.bz2: POSIX tar archive (GNU)
>>
>> This doesn't happen with compressed remote files which aren't tar archives.
>
> I can reproduce this problem with Emacs 29, too.  (And if I edit the
> file locally instead of via tramp, the tar file is saved compressed, so
> it seems to be tramp-related.)

Well, the problem seems to exist since Emacs 27, likely due to some
subtle changes in basic-save-buffer-1 or basic-save-buffer-2. I haven't
poked further there, because it looks like saving the buffer with this
constellation happened twice in Emacs 26, which looks not optimal.

The problem is an exotic feature of write-region, which first looks for
a file name handler in FILENAME, and if there isn't one, it checks
VISIT.

write-region of the compressed remote tar file in the scenario above is
called like

--8<---------------cut here---------------start------------->8---
  ...
  tramp-file-name-handler(write-region nil nil "/ssh:detlef:/tmp/tmpfSxlA9" nil "/ssh:detlef:/tmp/test.tar.bz2" "/ssh:detlef:/tmp/test.tar.bz2" nil)
  write-region(nil nil "/ssh:detlef:/tmp/tmpfSxlA9" nil "/ssh:detlef:/tmp/test.tar.bz2" "/ssh:detlef:/tmp/test.tar.bz2")
  basic-save-buffer-2()
  basic-save-buffer-1()
  basic-save-buffer(t)
  save-buffer(1)
  funcall-interactively(save-buffer 1)
  call-interactively(save-buffer nil nil)
  command-execute(save-buffer)
--8<---------------cut here---------------end--------------->8---

That is, a remote temporary file ("/ssh:detlef:/tmp/tmpfSxlA9") is
written, and the magic extension of the remote VISIT argument
"/ssh:detlef:/tmp/test.tar.bz2" is ignored.

When FILENAME is a local temporary file, the handler for VISIT
(jka-compr-handler) is triggered. That's the second test Lars has applied.

I've pushed a fix to master, which shall fix this. Dan, could you pls
check?

In parallel I'll try to write a test case for this problem. But that's
unrelated to closing this bug.

Best regards, Michael.

Previous Next

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