Re: [xml] python bindings are not IN_LIBXML

Stéphane Bidoul said:

I noticed that python/ now generates
#define IN_LIBXML. This causes mysterious
crashes in the windows version of the bindings,
because the linker does not bind to the correct
entry point for xmlFree...

The python binding cannot be considered as
beeing "IN_LIBXML", because, strictly speaking,
it's a client of libxml (a good client,
but a client nevertheless).

The CVS comment mentions the change was made
to silence a warning. I suppose it's about
xmlErrMemory that's undefined. IMHO, the correct
way to suppress that warning is
to adapt in order
to skip xmlErrMemory.

The proposed patch below fixes this.

On a related note, since xmlErrMemory
is an internal libxml function, maybe the
XMLPUBFUN modifier could be removed
so the function would not get exported?


RCS file: /cvs/gnome/gnome-xml/python/,v
retrieving revision 1.47
diff -c -r1.47
***        4 Dec 2003 12:31:48 -0000       1.47
---        30 Dec 2003 20:52:08 -0000
*** 325,330 ****
--- 326,333 ----
          return 1
      if name == "xmlOutputBufferFlush": # handled by by the
          return 1
+     if name == "xmlErrMemory":
+         return 1
      return 0

  def print_function_wrapper(name, output, export, include):
*** 574,580 ****
      wrapper.write("/* Generated */\n\n")
      wrapper.write("#include <Python.h>\n")
  #    wrapper.write("#include \"config.h\"\n")
-     wrapper.write("#define IN_LIBXML\n")
      wrapper.write("#include <libxml/xmlversion.h>\n")
      wrapper.write("#include <libxml/tree.h>\n")
      wrapper.write("#include <libxml/xmlschemastypes.h>\n")
--- 577,582 ----

I'm afraid that was my fault for not looking more carefully into why
this warning occured - my apologies!  I've applied your patch and
will have it committed later in the day.  Thanks for the correction.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]