perl-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Geoffrey Young <>
Subject Re: apache/content_length_header todo failures
Date Wed, 04 Aug 2004 12:42:34 GMT

Stas Bekman wrote:
> Geoff,
> so what's going to happen to these todo tests that you've added recently?:
> t/apache/content_length_header..........ok 2/24# Failed test 2 in
> t/apache/content_length_header.t at line 37 *TODO*
> t/apache/content_length_header..........ok 5/24# Failed test 5 in
> t/apache/content_length_header.t at line 47 *TODO*
> This can't stay this way for the release, since everybody and their mom
> are going to complain about this. 

yeah, I was thinking about this.  here is my take on it.

we're testing apache functionality, not our own, so we can safely adjust the
expectations to meet what we know apache is going to do.  basically, I would
view the apache tests as ones that, upon failure, let us know that the
underlying apache core has changed so it gives us something to investigate.

really, I made these TODO because I wasn't sure about the proper behavior.
I think that I am now :)

> If it's not going to be fixed by
> Apache folks (and it's not a mod_perl problem), then may be we should
> turn them into ok tests and instead document the issues?

not to prolong the discussion (perhaps this is a better place for it anyway)
but as I said in my last message I don't see that there is a bug - the tests
show that GET and HEAD present equivalent C-L headers except for the
specific case where you have a content handler (plus filters) that results
in zero content for the request (which is extremely rare).  and this case is
specifically different because it is coded to protect against handlers that
(incorrectly) call r->header_only, which presumably is less rare (since it
is a migration issue) than serving zero content.

but I'll certainly accept that my analysis is off, so if you don't agree
let's start from the begining and flesh out what we really need to happen.
that is, if you don't see it that way, I'm obviously sufficiently dense at
the moment that I need it explained to me again, starting with what the real
issue is and moving to exactly what part of apache core looks to be at fault.


To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message