On Mon, Mar 15 2010, Tatsuya Kinoshita wrote: >> So, though I don't yet fully understand the nature of this workaround >> and why it's necessary, I'm pretty sure that it _is_ needed for the >> current emacs snapshot 24.x, as well as for 23.1. > > Hmm, I coudn't reproduce the quoted-printable problem on wl-beta > 2.15.9+0.20100303-1 with Romain's emacs-snapshot 1:20100311-1. > > Could you please tell the procedure to reproduce your problem?
Hello, I don't know what's the debian BTS tradition about attachments, so I put some auxilary materials on my server (as ~10K zip archive). http://www.siftsoft.com/inprogress/flim-emacs-snapshot-woes.zip This archive contains: 1. original-russian-utf8.txt: A mail message source (cyrillic characters, UTF-8 encoding) 2. qp-encoded-bad.txt: raw message body that was sent over the wire from a clean (chrooted) system with wl-beta (debian), flim (debian), emacs-snapshot 1:20100311-1 (Romain). 3. qp-encoded-good: raw message body that was sent over the wire from a system similar to described above, but having flim package modified, as I propose. 4. modified.mel-q-ccl.el - a version of mel-q-ccl.el that produced qp-encoded-good.txt. (qp-encoded-good and qp-encoded-bad are not from _exactly_ the same text, but their sources consisted of the same pattern repeated multiple times, so the texts should identical till the end of the shortest one. Unfortunately, emacs segfaulted when I was going to save the source for "bad" message). If you open each of the qp-encoded*.txt messages, e.g. using gnus or wl, you can see easily that they differ, and the "bad" variant decodes to something really weird at certain points. I understand that it's far from a minimal self-contained test case, but the following things are visible: * No other packages, unusual system settings or configuration influence the bug: it's reproduced in the clean, freshly-debootstrapped system. * Modifying mel-q-ccl makes the bug disappear. * The corruption of the text starts after a certain number of characters and occurs periodically (or almost periodically), * The corruption includes repetition of already-encoded text fragments (maybe a sign of "byte/character length confusing"...) I don't know how to reproduced the same thing without Wanderlust (sorry, I'm not familiar with the entiry flim/semi/apel chain); I don't know either if message source charset really matters. However, I hope that these materials will help you to reproduce the problem on your own system. There are just two unusual things on my test system, that have invisible chances to be relevant, but who knows: 1. It's an amd64 (including userspace and emacs-snapshot) 2. Its locale is ru_RU.UTF-8. .emacs in the chrooted environment was almost empty during both of the tests (had to set debug-on-error to learn I had to set smtp-local-domain, to send those test messages). -- Regards, Anton Kovalenko +7(916)345-34-02 | Elektrostal' MO, Russia -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
