bug#20634: 25.0.50; Conflict between Tramp and EasyPG

bug#20634: 25.0.50; Conflict between Tramp and EasyPG

[Jens Lechtenboerger]

Hi there,

I use EasyPG for transparent access of encrypted files with
extension .gpg.  If I copy an encrypted file, say test.gpg, via ssh
(Tramp) to a remote location, then I’m asked what public key to use
for encryption (none for symmetric encryption), and the encrypted
file is encrypted again, rendering it useless from within Emacs.
(If I open the remote file via Tramp, the outer layer of encryption
is removed, and I see the original encrypted file.)

Recipe:

0. Encrypt a file: test.gpg

1. emacs -Q

2. Open directory of test.gpg (C-x C-f).

3. Copy (C) the file to a remote location
   (e.g., /user@localhost:/tmp).

   The key chooser opens.  Copy is cancelled if I cancel the choice.
   Otherwise, a remote file is created, which is not a copy but an
   encrypted file with two layers of encryption.

4. Open the remote file (C-x C-f /user@localhost:/tmp/test.gpg).
   Observe encrypted data.

In contrast, if I copy an encrypted file to a local directory, it is
copied without extra encryption.

Note also that if I copy a plaintext file (test.txt) locally to one
with gpg extension (test.gpg), it remains a plaintext file.  If I
copy that to a remote location, it gets encrypted.  This may be a
bug or a feature, but is certainly inconsistent behavior.

Best wishes
Jens

bug#20634: 25.0.50; Conflict between Tramp and EasyPG

[Michael Albinus]

Jens Lechtenboerger <jens.lechtenboerger@fsfe.org> writes:

> Hi there,

Hi Jens,

> I use EasyPG for transparent access of encrypted files with
> extension .gpg.  If I copy an encrypted file, say test.gpg, via ssh
> (Tramp) to a remote location, then I’m asked what public key to use
> for encryption (none for symmetric encryption), and the encrypted
> file is encrypted again, rendering it useless from within Emacs.
> (If I open the remote file via Tramp, the outer layer of encryption
> is removed, and I see the original encrypted file.)

Thanks for the report; I could reproduce the problem using your
recipe. Could you, please, check whether the following patch fixes the
problem?

--8<---------------cut here---------------start------------->8---
*** /home/albinus/src/tramp/lisp/tramp-sh.el.~master~	2015-05-24 18:17:53.333961247 +0200
--- /home/albinus/src/tramp/lisp/tramp-sh.el	2015-05-24 18:13:15.299684599 +0200
***************
*** 2122,2136 ****
  KEEP-DATE is non-nil if NEWNAME should have the same timestamp as FILENAME."
    ;; We must disable multibyte, because binary data shall not be
    ;; converted.  We don't want the target file to be compressed, so we
!   ;; let-bind `jka-compr-inhibit' to t.
!   ;; We remove `tramp-file-name-handler' from
    ;; `inhibit-file-name-handlers'; otherwise the file name handler for
    ;; `insert-file-contents' might be deactivated in some corner cases.
    (let ((coding-system-for-read 'binary)
  	(coding-system-for-write 'binary)
  	(jka-compr-inhibit t)
  	(inhibit-file-name-handlers
! 	 (remq 'tramp-file-name-handler inhibit-file-name-handlers)))
      (with-temp-file newname
        (set-buffer-multibyte nil)
        (insert-file-contents-literally filename)))
--- 2122,2137 ----
  KEEP-DATE is non-nil if NEWNAME should have the same timestamp as FILENAME."
    ;; We must disable multibyte, because binary data shall not be
    ;; converted.  We don't want the target file to be compressed, so we
!   ;; let-bind `jka-compr-inhibit' to t.  `epa-file-handler' shall not
!   ;; be called either.  We remove `tramp-file-name-handler' from
    ;; `inhibit-file-name-handlers'; otherwise the file name handler for
    ;; `insert-file-contents' might be deactivated in some corner cases.
    (let ((coding-system-for-read 'binary)
  	(coding-system-for-write 'binary)
  	(jka-compr-inhibit t)
  	(inhibit-file-name-handlers
! 	 (cons epa-file-handler
! 	       . (remq 'tramp-file-name-handler inhibit-file-name-handlers))))
      (with-temp-file newname
        (set-buffer-multibyte nil)
        (insert-file-contents-literally filename)))
***************
*** 3205,3211 ****
  	 (if (fboundp 'find-buffer-file-type)
  	     (symbol-function 'find-buffer-file-type)
  	   nil))
! 	(inhibit-file-name-handlers '(jka-compr-handler image-file-handler))
  	(inhibit-file-name-operation 'insert-file-contents))
      (unwind-protect
  	(progn
--- 3206,3213 ----
  	 (if (fboundp 'find-buffer-file-type)
  	     (symbol-function 'find-buffer-file-type)
  	   nil))
! 	(inhibit-file-name-handlers
! 	 '(epa-file-handler image-file-handler jka-compr-handler))
  	(inhibit-file-name-operation 'insert-file-contents))
      (unwind-protect
  	(progn
--8<---------------cut here---------------end--------------->8---

> Note also that if I copy a plaintext file (test.txt) locally to one
> with gpg extension (test.gpg), it remains a plaintext file.  If I
> copy that to a remote location, it gets encrypted.  This may be a
> bug or a feature, but is certainly inconsistent behavior.

It is rather a feature (a copy operation shall not change the file
contents). The above patch shall fix this as well.

> Best wishes
> Jens

Best regards, Michael.

bug#20634: 25.0.50; Conflict between Tramp and EasyPG

[Michael Albinus]

Michael Albinus <michael.albinus@gmx.de> writes:

Hi Jens,

> Thanks for the report; I could reproduce the problem using your
> recipe. Could you, please, check whether the following patch fixes the
> problem?

There was a typo in the patch. I've fixed it in the Emacs repository;
could you please test that version?

>> Best wishes
>> Jens

Best regards, Michael.

bug#20634: 25.0.50; Conflict between Tramp and EasyPG

[Jens Lechtenboerger]

On 2015-05-25, at 12:47, Michael Albinus wrote:

> Michael Albinus <michael.albinus@gmx.de> writes:
>
> Hi Jens,
>
>> Thanks for the report; I could reproduce the problem using your
>> recipe. Could you, please, check whether the following patch fixes the
>> problem?
>
> There was a typo in the patch. I've fixed it in the Emacs repository;
> could you please test that version?

Hi Michael,

many thanks for your quick reply.  Unfortunately, that does not
solve my issue.  The new file still gets encrypted a second time.
I traced what happens:

In tramp-do-copy-or-rename-file-via-buffer, the call to
with-temp-file results in a call to epa-file-handler, where the
argument operation has the value 'write-region.  In
epa-file-handler, op is let-bound to 'epa-file-write-region, which
gets called without looking at inhibit-file-name-handlers.

I also tried to let-bind inhibit-file-name-operation to
'write-region in tramp-do-copy-or-rename-file-via-buffer, but that
did not work either.

Best wishes
Jens

bug#20634: 25.0.50; Conflict between Tramp and EasyPG

[Michael Albinus]

Jens Lechtenboerger <jens.lechtenboerger@fsfe.org> writes:

> Hi Michael,

Hi Jens,

> many thanks for your quick reply.  Unfortunately, that does not
> solve my issue.  The new file still gets encrypted a second time.

Indeed, I haven't tested all use cases :-(

I've pushed an improved version of the patch; could you please check?
That seems to work for me.

> Best wishes
> Jens

Best regards, Michael.

bug#20634: 25.0.50; Conflict between Tramp and EasyPG

[Michael Albinus]

Version: 25.1

Jens Lechtenboerger <jens.lechtenboerger@fsfe.org> writes:

> Hi Michael,

Hi Jens,

> fixed :-)

Thanks for confirmation. I'm closing the ticket.

> Many thanks
> Jens

Best regards, Michael.

bug#20634: 25.0.50; Conflict between Tramp and EasyPG

[Jens Lechtenboerger]

> I've pushed an improved version of the patch; could you please check?
> That seems to work for me.

Hi Michael,

fixed :-)

Many thanks
Jens

P.S. My first reply went to Michael only.  Sorry about that.