Dan,
This one occurs even less frequently than the last one, but
unfortunately it still segfaults occasionally when garbage collection
kicks in. It appears to crop up in random places (inside active-record,
inside erb, etc.) instead of just when the XML document access is
occurring like the p
Calvin,
This bug took some time to find, but I got it. XPath was probably the
best way to expose it, since it had to do with random access to nodes
(really freeing them, but what's the difference). It would have had
an impact somewhere else for sure.
It is all committed and an update should
Dan,
The TAP code I pasted in the original email is fine to use for a test
case. As far as the XML document, let me create a smaller version of
the one you currently have with the stuff I don't need ripped out of it
and send that to you in place of the current XML document you have so
the res
Calvin,
The change in stack trace was helpful (never imagined I would say
that, since they are usually worthless). I looked at how xpath.find
was handling its arguments and there was something, well, not quite
right. I reworked it and ran your case with this code:
IO.popen("gzcat rwtest/tap
I have just committed a re-write of XPath/XPointer and *.find (calls
to XPath). It is svn
revision 183. See if it fixes this problem.
Note, there is an incompatibility, the new XPath::Object class does
not support .xpath method, but I could find no reason to carry it
forward.
If it works, I
