GenericSetup: Buggy TAR-Export for Workflows with Non-Ascii-Characters

[Roman Jansen-Winkeln]

We stumbled in problems with broken TAR-Exports from GenericSetup exports of DCWorkflows.

The reason: our interactively defined workflows contained non-ascii unicode characters.

For the enclosed example file "definitions.xml", the method "writeDataFile" in class "TarballExportContext" in file "context.py" of Products.GenerictSetup-1.6.1 reports in "info.size = len( text )" a length of 8300 characters. The real length of the file is however 8324 bytes, which is len(str(text)).
The relevant workflow in the attached (fixed) export is

We have developped a simple workaround and replaced the original:

        stream = StringIO( text )
        info = TarInfo( filename )
        info.size = len( text )

with

        text = str(text) ### Patch to convert unicode to bytes [RJW 2011-02-09]
        stream = StringIO( text )
        info = TarInfo( filename )
        info.size = len( text ) # len is wrong for utf-8 encoding

It works fine for our purposes. There may surely be a more elegant solution.

The bug appears in Plone 3.3.x and in Plone 4.0.x . Python operates in "utf-8" mode.

Best Regards
roman

[rafaelb (rafaelbco)]

I confirm this bug.

[Tres Seaver (tseaver)]

'str(text)' won't work if 'text' is non-ASCII unicode, unless you have
set 'sys.defaultencoding' (which is frowned upon as a general practice).

I will investigate: valid XML files must be encoded to UTF-8 unless they
specify a different encoding inline, which makes that the right encoding.

[Tres Seaver (tseaver)]

r123435 explicitly prohibits passing Unicode to the 'writeDataFile'
method of an export context: it is now the responsibility of the
handler generating the XML to encode it before passing it to the
context.