HTML parsing segfaults when tags are closed with <?tag> and the document is encoded in ASCII-8BIT

[barelyknown]

On ruby 1.9.3p286 (2012-10-12 revision 37165) [x86_64-darwin12.2.0]

Nokogiri::HTML("<strong>this will segfault<?strong>".force_encoding("ASCII-8BIT"))

Nokogiri::HTML("<strong>this will not segfault<?strong>".force_encoding("UTF-8"))

cc @mgates

[ender672]

When given an ASCII-8BIT string, Nokogiri invokes a SAX push parser to sniff out the document encoding.

You uncovered a segfault in the Nokogiri push parser -- it doesn't properly handle processing instructions with no content.

From ext/nokogiri/xml_sax_parser.c:240

static void processing_instruction(void * ctx, const xmlChar * name, const xmlChar * content)
{
  VALUE self = NOKOGIRI_SAX_SELF(ctx);
  VALUE doc = rb_iv_get(self, "@document");

  rb_funcall( doc,
              id_processing_instruction,
              2,
              NOKOGIRI_STR_NEW2(name),
              NOKOGIRI_STR_NEW2(content)
  );
}

The above will fail if libxml2 calls the function with NULL for content. It does exactly that in HTMLparser.c:3142

ctxt->sax->processingInstruction(ctxt->userData,
      target, NULL);

[barelyknown]

Glad that you found the issue. We spent some time to isolate the bug, but didn't know enough about Nokogiri to dig into the details to figure out why it was segfaulting.

[ender672]

Thanks for taking the time to isolate the bug! That effort made it loads easier to find and fix it.

[barelyknown]

Happy to have helped. Thanks for fixing.

On Wed, Feb 6, 2013 at 1:53 PM, Timothy Elliott [email protected]:

Thanks for taking the time to isolate the bug! That effort made it loads
easier to find and fix it.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/845#issuecomment-13201581.

[Flaburgan]

Looks like there still is a problem with push_parser, I had the exact same error: diaspora/diaspora#4996

[flavorjones]

Can you please open a new issue with code that will allow us to reproduce what you're experiencing?

Thank you.