https://bz.apache.org/bugzilla/show_bug.cgi?id=61222
--- Comment #18 from Ruediger Pluem <[email protected]> --- (In reply to Joe Orton from comment #17) > Created attachment 35323 [details] > iteration #5 > > I rewrote the loop logic to be more explicit (I hope) in the handling of > different bucket types, at least I can understand it properly now. > > Three specific cases I'm testing: > > 1) a CGI script which writes the header, sleeps, writes the body -- must see > headers hit the network before waiting to read the body > > 2) the bug reported here -- must see no buffering > > 3) for a small fast CGI script with <8K output -- does get C-L on response > > ... plus running the test suite. > > I'm worried there are specific response brigades with FLUSH this is going to > fail to produce C-L where we previously did, e.g. as simple as: > > HEAP FLUSH EOS > > will produced chunked output now. > > I think we could safely handling this by moving the "short-cut" case for EOS > right before the apr_brigade_split_ex call? Agreed. Cool patch. Nitpick: I think we should call apr_brigade_cleanup(ctx->tmpbb) before leaving if the ap_pass_brigade fails in the loop ( (rv != APR_SUCCESS) / (f->c->aborted)). This should free up memory faster. -- 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]
