From: Tomi Ollila <tomi.ollila@iki.fi>
To: Aaron Ecay <aaronecay@gmail.com>, Robert Horn <rjhorn@alum.mit.edu>,
	notmuch@notmuchmail.org
Subject: Re: bug related to ical
In-Reply-To: <87d319b899.fsf@gmail.com>
References: <m31ul3qo4w.fsf@quad.robs.office> <87d319b899.fsf@gmail.com>
User-Agent: Notmuch/0.14+51~g62cd13b (http://notmuchmail.org) Emacs/24.2.1
	(x86_64-unknown-linux-gnu)
Date: Wed, 03 Oct 2012 16:30:12 +0300
Message-ID: <m2obkjlm3f.fsf@guru.guru-group.fi>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Precedence: list

On Wed, Sep 26 2012, Aaron Ecay wrote:

> The problem is in the =E2=80=98notmuch-show-insert-part-text/calendar=E2=
=80=99
> function.  The call to =E2=80=98icalendar--convert-ical-to-diary=E2=80=99=
 does not
> create a buffer visiting the temp file, so the call to =E2=80=98set-buffe=
r=E2=80=99
> fails.  The following patch fixes the problem.
>
> The ical->diary conversion also doesn=E2=80=99t seem to work =E2=80=93 th=
e calendar
> attachment shows up as an empty part =E2=80=93 but I guess that=E2=80=99s=
 a separate
> issue (and not addressed by the patch).
>
> I guess that part insertion handlers should be called inside a
> =E2=80=98condition-case=E2=80=99, so that an error inside of one can be r=
ecovered from,
> and doesn=E2=80=99t entirely derail the insertion of the messages in the =
buffer.
> (I actually made this patch because I was so annoyed that Olivier=E2=80=
=99s
> buggy test attachment made it impossible for me to read Tomi=E2=80=99s re=
ply.)
>
> ----- cut here -----
>
> diff --git i/emacs/notmuch-show.el w/emacs/notmuch-show.el
> index ce5ea6f..4c89d7e 100644
> --- i/emacs/notmuch-show.el
> +++ w/emacs/notmuch-show.el
> @@ -746,7 +746,7 @@ message at DEPTH in the current thread."
>  	      (icalendar--convert-ical-to-diary
>  	       (icalendar--read-element nil nil)
>  	       file t)
> -	      (set-buffer (get-file-buffer file))
> +	      (set-buffer (find-file-noselect file))
>  	      (setq result (buffer-substring (point-min) (point-max)))
>  	      (set-buffer-modified-p nil)
>  	      (kill-buffer (current-buffer))
>
> ----- cut here -----

The problem seems to be carriage returns in the attachment -- which goes
unnotified when converting from octet-stream content.

I've got the example in question to "work" with the following patch_

--8<----8<----8<-- cut here  --8<----8<----8<--

diff --git a/emacs/notmuch-show.el b/emacs/notmuch-show.el
index 86130ce..f7c08ee 100644
--- a/emacs/notmuch-show.el
+++ b/emacs/notmuch-show.el
@@ -747,17 +747,21 @@ message at DEPTH in the current thread."
   (notmuch-show-insert-part-header nth declared-type content-type (plist-g=
et part :filename))
   (insert (with-temp-buffer
 	    (insert (notmuch-get-bodypart-content msg part nth notmuch-show-proce=
ss-crypto))
-	    (goto-char (point-min))
 	    (let ((file (make-temp-file "notmuch-ical"))
 		  result)
+	      (set-buffer (icalendar--get-unfolded-buffer (current-buffer)))
+	      (beginning-of-buffer)
+	      (replace-string "" "")
+	      (beginning-of-buffer)
 	      (icalendar--convert-ical-to-diary
 	       (icalendar--read-element nil nil)
 	       file t)
+	      (kill-buffer (current-buffer))
 	      (set-buffer (get-file-buffer file))
 	      (setq result (buffer-substring (point-min) (point-max)))
 	      (set-buffer-modified-p nil)
 	      (kill-buffer (current-buffer))
-	      (delete-file file)
+	      ;;;(delete-file file)
 	      result)))
   t)
=20
--8<----8<----8<-- cut here  --8<----8<----8<--

In our case the icalendar--get-unfolded-buffer doesn't seem
to do any conversions (the docstring mentions some LF/CR/BLANK
replacements) -- I just copied that line from icalendar-import-buffer
which I tried first.

The interesting thing is that the notmuch-icalXXXXXX file is
still empty ... hmm, the buffer is just not saved (would /dev/null work ;))

To do:
Check why icalendar--get-unfolded-buffer did not do what it advertised;
the code seems to support the replacements mentioned -- or maybe I failed ;/

The problem with failing to icalendar--convert-ical-to-diary is that
the buffer is not created. The find-file-noselect shields about that
problem here but not elsewhere (and why that failure is so "fatal" ???)
Some shielding in higher level should be implemented (i.e. the
"condition-case" suggestion Aaron mentioned or something :)


self to remember:  M-x debug-on-entry and keys 'd', 'u' and 'c' in debugger=
 :)

Tomi

>
> --=20
> Aaron Ecay
> _______________________________________________
> notmuch mailing list
> notmuch@notmuchmail.org
> http://notmuchmail.org/mailman/listinfo/notmuch