The "bug" in my code was the fact that I forgot that a request might also be a subrequest, so now I check if the request is_initial_req and short circuit if not. (And a tiny bug in my trace code caused my debug log to report is_initial_req as true all the time. Stupid operator precedence bugs!) That solves my immediate problem and gives me something to watch for in future.
The thing that triggered it was the fact that the url was an "index.pl" file. When called as "example.com/index.pl", everything worked as expected, but when called as "example.com/", apache (or something) generated a subrequest for index.pl.
NOW my question is, why does apache generate the subrequest *after* the HeaderParser phase, causing it to execute twice? This seems like a pure URI translation issue, and I would expect it to happen in the URI translation phase, *before* the HeaderParser executes at all. I'm wondering if this happens in all requests, or if it is related to Apache::Registry in the content handler. I didn't take the time to track down exactly where the subrequest originates, so I'm not sure if it is a mod_perl issue or just an apache issue. I might one day build some test cases and figure it out, but not soon. Anyone have some insight to add?
Thanks to everyone who sent help and suggestions! -Vince Veselosky http://ice.control-escape.com
Vince Veselosky wrote:
Now here is the thing. The PerlHeaderParserHandler executes TWICE for every request. The second time through is apparently a different Apache object from the first, as when I write to $r->notes, only the values from the second execution are visible to the content handler.
Now, if I comment out the PerlHeaderParserHandler line in the conf file, the handler never executes at all (which is expected), so it isn't getting pushed onto handlers from somewhere else.
Why is this thing running twice, and how can I make it stop??? I hope someone can hit me over the head with a clue-stick, because this thing is driving me completely bananas!
All help is greatly appreciated, Vince Veselosky http://ice.control-escape.com
