X-Git-Url: http://git.vanrenterghem.biz/git.ikiwiki.info.git/blobdiff_plain/ec47514bdb354fa1d25483e76f7d128d5ab5dc04..8b659301274ab0127b9a8e76b70c788cafe929ac:/doc/bugs/syslog_fails_with_non-ASCII_wikinames.mdwn?ds=sidebyside

diff --git a/doc/bugs/syslog_fails_with_non-ASCII_wikinames.mdwn b/doc/bugs/syslog_fails_with_non-ASCII_wikinames.mdwn
index 5d549b00c..0d40d232a 100644
--- a/doc/bugs/syslog_fails_with_non-ASCII_wikinames.mdwn
+++ b/doc/bugs/syslog_fails_with_non-ASCII_wikinames.mdwn
@@ -1,3 +1,5 @@
+[[!template  id=gitbranch branch=anarcat/dev/syslog_utf8 author="[[anarcat]]"]]
+
 [[this feature|todo/syslog_should_show_wiki_name]] made it so syslog doesn't work anymore if the site being logged has non-ASCII characters it in.
 
 Specifically, my wiki was named "CⒶTS", and nothing was showing up in syslog. When I changed that to "C@TS", it worked again.
@@ -11,3 +13,20 @@ My guess is this sits somewhere here:
 """]]
 
 Yet I am not sure how to fix that kind of problem in Perl... --[[anarcat]]
+
+> If I remove the "eval" above, I get:
+> 
+>     Error: Wide character in syswrite at /usr/lib/perl/5.14/Sys/Syslog.pm line 485.
+> 
+> I have improved a little the error handling in log_message() so that we see *something* when syslog fails, see the branch documented above. I can also confirm that  reverting [[todo/syslog_should_show_wiki_name]] fixes the bug. Finally, I have a unit test that reproduces the problem in git, and a working patch for the bug, again in git.
+> 
+> > One last note: I noticed that this problem also happens elsewhere in ikiwiki. For example, the [[plugins/notifyemail]] plugin will silently fail to send notifications if the pages contain unicode. The [[plugins/notifychanges]] plugin I am working on (in [[todo/option to send only the diff in notifyemail]]) seems to be working around the issue so far, but there's no telling which similar problem are out there.
+
+>> I'd merge it. --[[smcv]]
+
+>>> I've merged it, but I don't feel it fixes this bug. --[[Joey]]
+
+>>>> (I removed the patch tag to take it off the patches list.)
+>>>>
+>>>> What else is needed? Systematic classification of outputs into
+>>>> those that do and don't cope with Unicode? --[[smcv]]