On Wed, Mar 23, 2016 at 2:11 AM, Graham Leggett <minf...@sharp.fm> wrote: > On 22 Mar 2016, at 7:08 PM, yla...@apache.org wrote: > >> URL: http://svn.apache.org/viewvc?rev=1736216&view=rev >> Log: >> Follow up to r1734656: restore c->data_in_input_filters usage to >> see if it helps unblocking test framework. > > I created an almost identical patch before doing an update, the only > difference was this: > > Index: modules/http/http_request.c > =================================================================== > --- modules/http/http_request.c (revision 1736263) > +++ modules/http/http_request.c (working copy) > @@ -453,7 +453,7 @@ > > ap_process_async_request(r); > > - if (!c->data_in_input_filters || ap_run_input_pending(c) != OK) { > + if (!c->data_in_input_filters && ap_run_input_pending(c) != OK) { > bb = apr_brigade_create(c->pool, c->bucket_alloc); > b = apr_bucket_flush_create(c->bucket_alloc); > APR_BRIGADE_INSERT_HEAD(bb, b); > > I get hangs without the above patch on the test suite, but it runs clean > with. Can you give it a try?
Yes this is the right logic. I wonder how the hook can return anything but DECLINED though, I see no registering in our code. Unfortunately I can't reproduce the issue on my system(s), so testing is good, but it was already... Regards, Yann.