httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Yann Ylavic <>
Subject Re: svn commit: r1768079 - in /httpd/httpd/branches/2.4.x: ./ CHANGES STATUS docs/manual/ modules/http/http_filters.c
Date Fri, 04 Nov 2016 22:17:10 GMT
On Fri, Nov 4, 2016 at 10:15 PM, William A Rowe Jr <> wrote:
> I'm really not clear how ap_map_http_request_error() arrived without the
> patch patch below, and unclear whether it does what the backport proposal
> intended with this work omitted.
> 1482522 minfrin
> core: Stop the HTTP_IN filter from attempting to write error buckets
> to the output filters, which is bogus in the proxy case. Create a
> clean mapping from APR codes to HTTP status codes, and use it where
> needed.

The above was backported in 2.4.13 ([1]+[2]) and 2.2.30 ([3]), though
not exactly the same way as implemented in r1482522 (yet the intent is
there, and the issue fixed ;)

What differed is that http_filters' bail_out_on_error() was preserved
(so that [third-party-]modules don't notice/have to handle the
response themselves), and "generic" proxy 502 error status was not
turned into 504 (disputed change IIRC).

Otherwise the potential double responses on error were finally handled
by [4] and follow ups (correct raising/handling of AP_FILTER_ERROR),
supposed to be compatible with both module behaviours wrt filters
error handling.

> I'll simply ignore this in the backport/normalization but figured it's worth
> pointing out such a discrepancy.

Not sure there is one...

Regarding the related changes in trunk only (now), we should probably
decide what to do about the 502=>504, but the modules having to
handle/raise the error looks reasonable to me, and even the correct
behaviour for 2.6/3.0.



View raw message