Created attachment 351643 [details] [review]
libxml2_patch_colons_in_pinames.patch

Hi.

I was asked to report this bug upstream. Since it might have side effects, which I cannot rule out. The following is the text from the original report:


When parsing an XML Document, libxml2 wrongly reports an error
(seemingly without further consequences) about colons not being allowed
in Processing-Instruction names.

This is not true. The XML Specification
(https://www.w3.org/TR/REC-xml/#sec-pi and
https://www.w3.org/TR/REC-xml/#NT-Name) explicitly allows a colon (':')
as part of a PITarget, which is a Name in the XML specifications terms.
The last Version of the XML Specification, where a colon was not allowed
in a processing instruction name is over 20 years old and I think can be
safely ignored by now: https://www.w3.org/TR/WD-xml-lang-970331#NT-Name

Attached (in the next reply to this bug) is a test document (test.xml), which triggers the warning, when
used for example with xmlstarlet like this:

$ xmlstarlet fo test.xml
test.xml:2.8: colons are forbidden from PI names 'pi:pi'
<?pi:pi pi?>
       ^
<?xml version="1.0"?>
<root>
  <?pi:pi pi?>
</root>

Also attached is a patch (libxml2_patch_colons_in_pinames.patch), that
removes the bit of code from libxml2, that triggers the message.

I am not sure about the side effects of the removed function-call, but
from the comments in the code it seems it should have aborted
processing. It specifically states for xmlNsErr: "Handle a fatal parser
error, i.e. violating Well-Formedness constraints". However it does not
really abort, which might warrant another bug report.

Regards
  Andre