ap_die() is definitely confused about whether or not it is a recursive error. Any time r->status is set, it thinks it is a recursive error.
On a side note, is it bad for modules to set r->status?
it was because of the former that I thought it was bad to do the latter.
and I was wondering if more people didn't run into this because there were other reasons why modules shouldn't set r->status :)
over in mod_perl-land there is the propensity for people to play with $r->status in handlers (due to Apache::Registry using it as a hack). I always tell them that handlers (Perl or C) communicate their status via their return code, not $r->status - doing so will muck with the error document cycle, and no core module does it, so don't do it.
the confusion about whether the error is recursive can definitely be fixed... it turns out that my 413 example may not be the best one, as there are additional issues when core sets 413 based on LimitRequestBody... see http://nagoya.apache.org/bugzilla/show_bug.cgi?id=21035
it would be nice to hear somebody verify that here :)
definitely!
