On Sat, Nov 5, 2016 at 10:38 AM, Yann Ylavic <[email protected]> wrote:
> On Sat, Nov 5, 2016 at 3:15 AM, William A Rowe Jr <[email protected]> > wrote: > >> > >> On Fri, Nov 4, 2016 at 5:17 PM, Yann Ylavic <[email protected]> > wrote: > >>> > >>> 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) > > > > > > Which would be a brilliant solution if > > > > -static apr_status_t bail_out_on_error(http_ctx_t *ctx, > > - ap_filter_t *f, > > - int http_error) > > > > Wasn't a private, unexported entry point on every sane C compiler. > Sigh... > > Not sure to understand your point, I was talking about the behaviour, > not the API... > bail_out_on_error() is the one that sends an error to the client if > something goes wrong by reading the request body, 2.[24] modules may > rely on it, hence the backport of r1482522 did not include that > change. > Ah, thanks, that makes more sense. I'm going to omit the delta on the http_filter.c from the proposed roll-up fix, and offer the minor delta for chunk line termination as a separate patch for +/-1 consideration. It makes the overall patch easier to review, I'm hoping.
