Mod_include IS WORKING.

[rb...@covalent.net]

We are rolling the alpha today, in about two hours.  Speak up now if this
is no good for you.

The problem with the mod_include as posted by Paul this morning is
actually in our handling of sub_requests and filters.  What we used to do,
is that we would set the first filter in the new request to the top of the
original request's filters, and then add the sub_request output
filter.  This is wrong, and most of us agree that it is wrong.

Mod_include was setting the first filter for the sub_request to the next
filter in the chain.  Which, avoids the sub_request output filter, so we
send multiple EOS buckets to the core_output_filter, which messes things
up subtly.

If you remove the lines in mod_include that set the output_filters to the
next filter, then we are really screwed, because the sub_request's data is
sent through the exact same filter as the current request, with the same
ctx structure.  This causes more non-subtle bugs.  (Seg fault)

The solution is simple enough.  When you create a sub-request, we need to
be able to specify what the first filter in the sub-request should
be.  This is done with an argument to the functions that create the
sub-request.  If this argument is NULL (most handlers will use this
value), then we just automatically use the first filter in the original
request.

I will have a clean patch in a few minutes.

Ryan


_______________________________________________________________________________
Ryan Bloom                        	rbb@apache.org
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------