Unicode site title breaks CKAN

[tobes]

We are using CKAN v1.8 and we noticed that if we use a site title that contains greek characters, an exception is thrown:

UnicodeDecodeError: 'ascii' codec can't decode byte 0xce in position 0: ordinal not in range(128)

In our local installation we managed to overcome this by 'overriding' every template file that used the site title. You might use the unicode() function to output the string correctly

[tobes]

This also affects 2.0. The following patch worked for me I'm not sure why the config is latin-1 not utf-8 does anyone know why this is?

diff --git a/ckan/lib/app_globals.py b/ckan/lib/app_globals.py
index f93a11e..d478f5e 100644
--- a/ckan/lib/app_globals.py
+++ b/ckan/lib/app_globals.py
@@ -110,6 +110,8 @@ def reset():
         # we want to store the config the first time we get here so we can
         # reset them if needed
         config_value = config.get(key)
+        if config_value:
+            config_value = config_value.decode('latin-1')
         if key not in _CONFIG_CACHE:
             _CONFIG_CACHE[key] = config_value
         if value is not None:

[davidread]

paste/deploy/loadwsgi.py just opens the file and calls ConfigParser.readfp() when it could be decoded as UTF8 first: http://stackoverflow.com/questions/1648517/configparser-with-unicode-items so you might solve this with a monkey patch to paste. Or probably easier to just do what you've done here...

[tobes]

@kindly
I think we should fix this as in my above patch and backport - does that seem the correct approach if so the I'll do a pull request etc and get adria to backport

[tobes]

@kindly @amercader This looks like we should fix and get it into 2.0 are there objections to my patch above? should I make it a pull request

[amercader]

@tobes go for it. Will this support all unicode characters? (eg ckan.site_title = हाय बिल्लियों)

[tobes]

@amercader I have changed this to also work with unicode as far as I can tell it is all good please consider for 2.0 thanks.
code here now