We need to be able to stop processing the request if a downstream
filter fails, which is why that change was made originally. As long as
we continue to return on errors, cool.
Ryan
On Wednesday 22 August 2001 11:31, Greg Ames wrote:
> Greg Ames wrote:
> > We got a second seg fault running 2.0.24 tonight with mod_include
> > instrumented.
>
> Problem debugged. There was a commit to mod_include.h after 2.0.22 to
> the SPLIT_AND_PASS_PRETAG_BUCKETS macro which exited early if
> ap_pass_brigade failed. It returned before setting a brigade pointer to
> the result of a split, which caused the seg faults when there are
> dropped connections.
>
> Note: this is a return hidden in the middle of a macro. Evil. Try
> stepping thru it with a debugger sometime. (hey! how did I end up in
> the caller all of a sudden?)
>
> Paul Reder said he'd work on getting rid of this macro and the
> CREATE_ERROR_BUCKET macro also, as well as doing the Right Thing in the
> error legs. I've got 2.0.24 with a tweaked SPLIT_AND_PASS... macro
> running on port 8092 daedalus now, and I'm about to do log replay
> testing on it. I'm hoping tonight we can run apache.org production on
> 2.0.24 plus a new improved mod_include.
>
> Greg
--
______________________________________________________________
Ryan Bloom [EMAIL PROTECTED]
Covalent Technologies [EMAIL PROTECTED]
--------------------------------------------------------------
