https://bz.apache.org/bugzilla/show_bug.cgi?id=61616
Yann Ylavic <[email protected]> changed: What |Removed |Added ---------------------------------------------------------------------------- Attachment #35783|0 |1 is obsolete| | Attachment #35784|0 |1 is obsolete| | --- Comment #70 from Yann Ylavic <[email protected]> --- Created attachment 35789 --> https://bz.apache.org/bugzilla/attachment.cgi?id=35789&action=edit Handle POLLOUT in proxy_connect (v13) Previous patch (v12) was bogus, indeed. This new one should work regardless of THRESHOLD_MIN_WRITE (or FlushMinThreshold from attachment 35620), i.e. it's really self contained. The does use or play with "data_in_input_filters" because the core filter can't be responsible for clearing it (other filters might retain data too, suppose it's used by third-party modules...). So I kept my "drain" flag, set according to ap_proxy_transfer_between_connections() since a successful return from there implies that there is no pending data. It allows to reduce pressure on buffering as you noted, but mainly to avoid blocking on the write side of ap_proxy_transfer_between_connections() by progressively accumulating data there, should it not "follow" the read side (which defeats the purpose of checking data_in_output_filters anyway). I tries to keep the best of our respective proposals (no buffering nor code duplication...), if it works we should find an agreement right? :) -- You are receiving this mail because: You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
