Just before going on holiday (for one week) I'd like to say that I've solved my problem.
1) I ran some ugly php scripts which update some fields (page.page_title, categorylinks.cl_to & cl_sortkey...), utf8_decoding them. 2) A lot of pages where OK after that script, especially those with latin and not-so-extended-latin letters. 3) Other extended-latin characters + greek & cyrillic were not correct after my script, they were replaced by ? (e.g. K?link? instead of Kalinka <- writted in cyrillic). So I couldn't enter on that page to view/edit/move article... 4) I hacked the URL (GetLocalUrl hook) to show the page_id. For each ugly link, I looked the id, update page set page_title='TMP' where page_id=ID, and then I could move the article TMP to the right title using the rename tab while viewing the article. Quite long but it worked !
All diacritics problem should have disappeared now :) maybe there are some error elsewhere but I didn't find them till now.
Bye
2006/4/21, iubito iubito@gmail.com:
It seems that when I create a new page, the title is correctly encoded, but somewhere during the import/export of DB dump (in phpMyAdmin), diacritics in the page_title field has been utf8_encode()d.
Looking for Macédoine (the country Macedonia), I get : SELECT * FROM `wiki_page` WHERE page_title LIKE 'Mac%' old host : MacÃ(c)doine new host : MacÃÂ(c)doine
I think I could fix the titles with a PHP script that updates wiki_page, with utf8_decode(page_title).
Do you have an idea of where the bug comes from ? phpMyAdmin dump ? MediaWiki update (1.4.14 => 1.5.3) on a special config (my second host)
The line for Macédoine in the sql dump is : INSERT INTO `wiki_page` VALUES (969, 14, 0x4d6163c383c2a9646f696e65, '', 610, 0, 0, 0.908237832938, 0x3230303630333239313134323334, 2593, 2749);
Really strange...
Sylvain
2006/4/18, iubito iubito@gmail.com:
I'm back, still with my problem. Does the debug trace give you some ideas ?
2006/4/15, iubito iubito@gmail.com:
Thanks for your reply. Bad news, it doesn't work :( Clearing the objectcache didn't change anything to the problem.
Setting wgDBmysql5 to true fails with a syntax error.
I debugged and see error here : SQL: SELECT /* Parser::replaceLinkHolders */ page_id, page_namespace, page_title FROM `wiki_page` WHERE (page_namespace=0 AND page_title IN('CatÃÂ(c)gorie:Baglama', 'CatÃÂ(c)gorie:Balalaïka', 'CatÃÂ(c)gorie:Bandoura', ... , 'CatÃÂ(c)gorie:Rythme', 'CatÃÂ(c)gorie:Cuisine', 'CatÃÂ(c)gorie:St SQL ERROR: Illegal mix of collations for operation ' IN ' (<mysql host>)
I checked collation of wiki_page table. Collation = latin1_general_ci except for the 2 fields page_title and page_touched where it is latin_bin.
Just a few queries after, I get :
SQL: SELECT /* MediaWikiBagOStuff::_doquery */ value,exptime FROM `wiki_objectcache` WHERE keyname='tousauxbalkans:messages:Outils' ** no caching ** WARNING: Abrupt exit in /home/.../includes/GlobalFunctions.php at line 579 WARNING: Abrupt exit in /home/.../includes/OutputPage.php at line 798 WARNING: Abrupt exit in /home/.../includes/Database.php at line 476 WARNING: Abrupt exit in /home/.../includes/Database.php at line 419 WARNING: Abrupt exit in /home/.../includes/Parser.php at line 3677 WARNING: Abrupt exit in /home/.../includes/Parser.php at line 238 WARNING: Abrupt exit in /home/.../includes/OutputPage.php at line 314 WARNING: Abrupt exit in /home/.../includes/Article.php at line 905 WARNING: Abrupt exit in /home/.../includes/Wiki.php at line 326 WARNING: Abrupt exit in /home/.../includes/Wiki.php at line 50 WARNING: Abrupt exit in /home/.../index.php at line 127
2006/4/14, Brion Vibber brion@pobox.com:
iubito wrote: [snip]
How do you explain that ? Which table(s) I must look better to compare in what they are
different ?
Does the experimental new support of UTF8 should solve this ? (I
kept the
compatible UTF8)
I'm not 100% sure this will help, but try:
Try:
- Add to LocalSettings.php:
$wgDBmysql5 = true;
- Clear the cache in database:
TRUNCATE TABLE objectcache;
Happy Easter everybody ! :-)
-- Sylvain Machefert
-- Sylvain Machefert http://iubito.free.fr http://tousauxbalkans.free.fr
-- Sylvain Machefert http://iubito.free.fr http://tousauxbalkans.jexiste.fr