"William A. Rowe, Jr." wrote:
> John Sterling noted this as well, and has submitted a patch to correct it.
> Expect some action early this week.
>
> If you are right, it would be interesting to discover 'what changed' in
> light of the fact that it worked so recently. No doubt another optmization
> or bug fix clobbered this previously correct behavior.
The bug was introduced when the inserting of the HTTP_HEADER filter was
moved to the insert_filter phase. This change made it possible for
AP_FTYPE_HTTP_HEADER filters to run before HTTP_HEADER (and have an
effect), but at the same time left the filter stack empty when errors
occurred before the insert_filter phase.
The patch adds a check to ap_die() to make sure the basic filters are
present, and adds them if they are needed. ap_die() previously assumed
the filter stack would be ok during an error condition, which is not
necessarily the case.
Regards,
Graham
--
-----------------------------------------
[EMAIL PROTECTED] "There's a moon
over Bourbon Street
tonight..."
S/MIME Cryptographic Signature
