2012-08-30

MSI + Unicode ≠ Love

I was upgrading my Libre Office installation to version 3.6.1 when I've noticed the question marks in the installer UI:



How can this be? Ș and ț (s and t comma below) characters are present in the Microsoft Sans Serif and Tahoma fonts since Windows 2000.

The only explanation is: ANSI installation program. No way! You would think that Windows Installer would be Unicode, but it's not! Michael Kaplan wrote about this weirdness seven years ago: MSI Databases and Unicode?

WiX also states on their help page: „Top-level elements like Product, Module, Patch, and PatchCreation support a Codepage attribute. You can set this to a valid Windows code page by integer like 1252, or by web name like Windows-1252. UTF-7 and UTF-8 are not officially supported because of user interface issues. Unicode is not supported.”

Programs like 7-zip use MSI for x64 target because NSIS installer did not have support for x64. NSIS officially doesn't support Unicode and x64, but there are forks which do (Unicode fork, x64 fork).

I've filed a bug (#54232) for LibreOffice. I guess they will use the old ş and ţ (s and t cedilla) characters to fix this problem.

Windows 95 End of Life was 1st of January 2003, Windows 98, Windows 98 Second Edition, and Windows ME End of Life was 1st of April 2007.

They haven't fixed this problem even though is has been five years since they do not support any ANSI operating system.
Post a Comment