This is a common problem/limitation already described hundred times on different forums etc. A complete description is in a Ximian Evolution Bugzilla: http://bugzilla.ximian.com/show_bug.cgi?id=14611 It was reported first back 2001, but since then it was NOT FIXED in Evolution. (Even in Evolution 2.0.0). The problem goes as follows: 0. Evolution stores all the data in UTF-8 character set. 1. Palm Evolution Conduits don't allow to select a character encoding of the data on the Palm. 2. Conduits rely on the libpisock library (provided by app-pda/pilot-link) to convert/transfer data to the Palm. 3. libpisock assume (hardcoded in the source) that the Palm character set is CP1251. The result: All the people that speak different languages than English can't use Evolution conduits, because their data gets scrambled in the transfer (I'm talking about language specific special characters, ISO8859-2 characters, etc.). I made a simple PATCH to libpisock in app-pda/pilot-link, that changes CP1251 to ISO-8859-2 and fixes some iconv bugs (shit code). This patch is universal, and should work with any other application that uses libpisock (although - It looks like Evolution Conduits is a single application, that uses libpisock). This way, I (speaking Polish, and using some polish specific accented characters) can use Evolution Conduits without problems. ISO-8859-2 is the default character set for all the Palm Pilots bundled in Central and Eastern Europe. Portage: I think, this small patch should be a part of the app-pda/pilot-link in the Portage. It should ask user at the compile-time, whetever to apply it or not ("What kind of character set you're using on the Palm?"). The patch is temporary (of course). I hope that one day, the pilot-link devs will move their heads, and allow the character set to be selectable (not harcoded in the source). I attach the patched/modified file. It applies to: app-pda/pilot-link File in the source: libpisock/util.c Hope you can add that to the portage. It will help a lot of people. Reproducible: Always Steps to Reproduce: 1. Type any name with a ISO-8859-2 character in the Evolution Address Book 2. Sync that to your pilot. 3. Look at the pilot. Actual Results: I got random characters (screwed character set) on the Palm Pilot, instead of the ISO-8859-2 stuff. Expected Results: Display name excatly same as entered in Evolution Address Book. I checked the patch on: Evolution-2.0.0 pilot-link-0.11.8 gnome-pilot-2.0.12 & 2.0.10
Created attachment 41926 [details, diff] Applies to app-pda/pilot-link. In the source: libpisock/util.c Part of the libpisock library compiled with pilot-link.
One small detail - I made a typo. Should be: CP1252 (not CP1251) as the default characterset assumed by libpisock (doesn't change the sense, but just to make it clear).
please provide the patch as a 'diff -uNr' to the evolution tree. Have you posted this patch upstream ? If not, you should. If you did, where is the bugzilla entry ?
Comment on attachment 41926 [details, diff] Applies to app-pda/pilot-link. In the source: libpisock/util.c --- pilot-link-0.11.8/work/pilot-link-0.11.8/libpisock/util.c 2002-07-30 17:52:24.000000000 +0200 +++ pilot-link-0.11.8.patched/work/pilot-link-0.11.8/libpisock/util.c 2004-10-16 21:46:33.000000000 +0200 @@ -18,7 +18,7 @@ * You should have received a copy of the GNU General Public License along * with this program; if not, write to the Free Software Foundation, Inc., * 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA - */ + THIS FILE WAS PATCHED WITH ISO8859-2 PATCH FOR EVOLUTION CONDUITS */ #ifdef HAVE_CONFIG_H #include <config.h> @@ -26,13 +26,14 @@ #include <stdlib.h> #include <string.h> +#include <stdio.h> #include "pi-util.h" #ifdef HAVE_ICONV #include <iconv.h> #endif -#define PILOT_CHARSET "CP1252" +#define PILOT_CHARSET "ISO8859-2" /*********************************************************************** * @@ -51,11 +52,13 @@ int bytes, char **ptext) { #ifdef HAVE_ICONV - char *ib, + + char *ib, *ob; iconv_t cd; size_t ibl, obl; - + + cd = iconv_open(PILOT_CHARSET, charset); if (!cd) return -1; @@ -65,13 +68,13 @@ ib = (char *)text; *ptext = ob = malloc(obl); - if (iconv(cd, &ib, &ibl, &ob, &obl) == -1) - return -1; + iconv(cd, &ib, &ibl, &ob, &obl); *ob = '\0'; iconv_close(cd); return 0; + #else return -1; #endif @@ -94,11 +97,12 @@ int bytes, char **text) { #ifdef HAVE_ICONV + char *ib, *ob; iconv_t cd; size_t ibl, obl; - + cd = iconv_open(charset, PILOT_CHARSET); if (!cd) return -1; @@ -108,13 +112,13 @@ ib = (char *)ptext; *text = ob = malloc(obl); - if (iconv(cd, &ib, &ibl, &ob, &obl) == -1) - return -1; + iconv(cd, &ib, &ibl, &ob, &obl); *ob = '\0'; iconv_close(cd); return 0; + #else return -1; #endif
Created attachment 41983 [details, diff] app-pda/pilot-link-0.11.8/work/pilot-link-0.11.8/libpisock/util.c
Ups. sorry for pasting the patch as a comment. The new (diff) file is in the attachment. What do you mean by "Have you posted this patch upstream?" ? Should I make a bugzilla entry, and ask some people to test it?
btw, have the pilot-link developers seen this patch. it doesn't look quite right to me by ignoring the status of iconv, given that possibly there maybe a reason why the check is there. need to look into this a bit more.
Is this still a problem?
please get back to us with updated information.
