The following report was already sent upstream <hack@robf.de> at 22 Nov 2008. Since the dawn of time (actually, at least since 1993 and VM 5.40 ;-) I use Supercite together with VM and never had any problems with it. Now it looks like it ceased to work with VM version 8.0.12. To reproduce the problem (this is with GNU Emacs 22.3 and VM 8.0.12): 1. Save attached file as folder "test1" 2. Eval the following expression: (add-hook 'mail-citation-hook 'sc-cite-original) 3. Type "M-x vm-visit-folder RET test1 RET" 4. Type "M-x load-library RET vm-pine RET" (to work around the missing autoload of vm-decode-postponed-mime-message, as reported earlier) 5a. Move to the first message ("Subject: test 1") and type "R" Now some warning messages appear: ,---- | Mail header "Dies ist e..." doesn't conform to RFC 822. skipping... | No mail headers found! Restoring old information. `---- 6a. When asked for the attribution name, type "RET" The body of the reply buffer will now contain the following: ,---- | Anon> Dies ist ein Test. `---- Expected behaviour: 5b. Move to the second message ("Subject: test 2") and type "R" 6b. When asked for the attribution name, type "RET" The body of the reply buffer will look as follows in this case: ,---- | >>>>> "Ulrich" == Ulrich Mueller <ulm@kph.uni-mainz.de> writes: | | Ulrich> Dies ist ein Test. `---- With VM 8.0.9 it is working as expected for _both_ messages. The difference between the two messages is that the first one contains MIME headers, while the second one doesn't. For completeness, I'm also attaching VM's state information.
Created attachment 180220 [details] "test1" VM folder
Created attachment 180221 [details] VM state information
Created attachment 180226 [details, diff] vm-8.0.12-fix-supercite-yank.patch Seems the problem is in function vm-yank-message, which doesn't work as documented: ,---- | All message headers are yanked along with the text. Point is | left before the inserted text, the mark after. Any hook | functions bound to `mail-citation-hook' are run, after inserting | the text and setting point and mark. `---- (See also sc.info, node "Hints to MUA Authors".) But vm-yank message includes the headers only in the non-MIME case: (if (eq layout 'none) (vm-insert-region-from-buffer (vm-buffer-of message) (vm-headers-of message) (vm-text-end-of message)) (vm-decode-mime-layout layout) (if vm-mime-yank-attachments (vm-decode-postponed-mime-message)))) Inserting another "vm-insert-region-from-buffer" command (but with a third argument "(vm-text-of message)") at the beginning of the "else" part, as in attached patch, seems to fix the Supercite problem. However, I don't understand the code well enough to decide if this is a proper fix, or has other undesired side effects. (For example, what is function "vm-decode-mime-layout" supposed to do? It is rather long and has no documentation.)
> Created an attachment (id=180226) [edit] > vm-8.0.12-fix-supercite-yank.patch > However, I don't understand the code well enough to decide if this is > a proper fix, or has other undesired side effects. Committed as -r2, as I'm using this myself since quite some time and haven't had any problems with it.
VM should also decode the headers before calling mail-citation-hook. See <http://thread.gmane.org/gmane.emacs.devel/116483>. Reopening.
Created attachment 209113 [details, diff] Updated patch with MIME header decoding
Fixed in vm-8.0.12-r6.
