[SVN-4064] UTF-8 files containing only a byte order mark are added as binary

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: all
• Fix Version/s: 1.7.2
• Component/s: cmdline client
• Assignee: Stefan Sperling
• Reporter: Subversion Importer
• Created: 18/Nov/11 14:55
• Updated: 27/Oct/18 16:28
• Resolved: 21/Nov/11 20:06

Description

When adding a file that is empty except for a UTF-8 byte order mark to an SVN 
repository, that file gets assigned an svn:mime-type property of 
application/octet-stream even though it is clearly meant to be a text file. 
This is a problem because some text editors use the UTF-8 byte order mark to 
identify the encoding of a file and thus automatically add that mark even to 
empty files.

An exemplary tool combination causing this problem to appear is that of 
Microsoft Visual Studio with the VisualSVN add-in because when Visual Studio 
creates an empty code file, VisualSVN will automatically add this file to SVN 
and the code file will thus be identified as a binary file.

The reason for this issue is that the svn_io_detect_mimetype2 (and 
svn_io_is_binary_data) functions have no specific support for detecting the UTF-
8 byte order mark.

The solution discussed on the mailing list would be to patch one of these 
functions so that the byte order mark is not interpreted as binary data.

http://svn.haxx.se/dev/archive-2011-10/0331.shtml

Original issue reported by fabianschmied