Re: Regression with large XML data input

On 2025-07-24 05:12 +0200, Michael Paquier wrote:
> Switching back to the previous code, where we rely on
> xmlParseBalancedChunkMemory() fixes the issue.  A quick POC is
> attached.  It fails one case in check-world with SERIALIZE because I
> am not sure it is possible to pass down some options through
> xmlParseBalancedChunkMemory(), still the regression is gone, and I am
> wondering if there is not a better solution to be able to dodge the
> original problem and still accept this case.

The whitespace can be preserved by setting xmlKeepBlanksDefault before
parsing.  See attached v2.  That function is deprecated, though.  But
libxml2 uses thread-local globals, so it should be safe.  Other than
that, I see no other way to set XML_PARSE_NOBLANKS with
xmlParseBalancedChunkMemory.

[1] https://gitlab.gnome.org/GNOME/libxml2/-/blob/408bd0e18e6ddba5d18e51d52da0f7b3ca1b4421/parserInternals.c#L2833

-- 
Erik Wienhold