307625 - OPML export and import seem to have different assumptions about (default) character sets

Status: RESOLVED FIXED

(MailNews Core :: Feed Reader, defect)

Description

[Pontus Freyhult]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; sv-SE; rv:1.8b4) Gecko/20050909 Firefox/1.4
Build Identifier: Tb 1.4

(Verified for swedis characters - å, ä and ö.)

When exporting (from subscription management) feeds with non-ascii characters in
their names, it seems to output a file using iso-8859-1 or iso-8859-15 (or
something like that).

However, if one imports that same file, those characters are removed and the
replacement character (U+FFFD) is used instead. (If the export is repeated after
that, it uses the entity, which works correctly). 

Reproducible: Always

Steps to Reproduce:
1. Subscribe to a feed (such as http://kornet.nu/blindhona/index.xml) with
non-ascii character in the name.

2. Export

3. Remove feeds

4. Import OPML file that was exported

Actual Results:  
Non-ascii (latin-1?) characters in names replaced by U+FFFD.

Expected Results:  
The names should be the same after exporting and importing.

Seems to work correctly when using entities with decimal references, can't we
just  do that for non-ascii chars?

Comment 1

[Robert Sayre]

Yes, the xmlEncoding of the opml DOM is null. We'll tell the serializer to use
utf-8.

Comment 2

[Robert Sayre]

Attachment: Set the XMLSerializer encoding to utf-8

Comment 3

[Scott MacGregor]

fixed branch and trunk.