https://bz.apache.org/bugzilla/show_bug.cgi?id=61222
--- Comment #12 from Ruediger Pluem <[email protected]> --- (In reply to Joe Orton from comment #6) > Created attachment 35321 [details] > Modified rpluem's patch with more FLUSH > > Thanks a lot for looking at this Ruediger! You're correct of course on > r->bytes_sent, which I completely missed... > > So I agree it makes sense to keep the loop, but I think we should keep the > FLUSH bucket insertion for the EAGAIN case - or is there a good reason to > remove it? Hence I'd modify your patch slightly to the attached. My initial thought on the flush was to remove it in order to give the downstream filters some freedom whether they need to flush or can buffer (at least parts) for whatever optimization, but now I agree that it makes sense to flush when we probably block next. But looking at my own an your patch additional points come up: 1. Don't we need to handle flush buckets and do flushing nevertheless what happens else? 2. The if (e->length == (apr_size_t)-1) seems stupid as it would also handle META_BUCKETS trying to read them. Shouldn't we just skip them in the loop (with the exception of the flush buckets above)? -- 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]
