Number: 1239
Title: Non-ASCII user name on Windows produces non-wellformed SVG
Submitter: Martin Duerst
Date: Mon Nov 26 06:03:13 2007
Subsys: Output generation
Version: 2.16
System: x86-Windows-XP (probably others, too)
Severity: major
Problem:
In its SVG output, grgaphviz includes a comment with the graphviz version and, after the string "For user:", the name of the user. On Windows systems, the name of the user can contain non-ASCII characters. Graphviz obtains the name via an older interface that uses what's on Windows called the "ANSI" character encoding, which for most language versions of Windows is not Unicode (and if it were Unicode, it would be UTF-16, not UTF-8). Graphviz takes the user name, obtained as bytes, and stuffs these bytes into the SVG output, which is declared to be UTF-8. When a downstream program reads the SVG, it produces a well-formedness error, because the user name is not a correct UTF-8 byte sequence. This error can be fixed rather easily by hand for single files, but is very annoying for automatic reuse (we encountered it in the context of isaViz). That's why I have marked it as Major.
Comments: The problem is independent of .dot input.
Fix:
We fixed the problem locally by creating a new Windows user with an US-ASCII-only name. But that's not really a fix :-( .
Owner: erg
Status: Fixed (27 Nov 2007)