Wietse, Thank you very much! I'll let you know as soon as this patch is applied (we're in an extended holiday here).
> Wietse Venema: > > The long-term solution is to implement an asynchronous atrace_flush() > > operation, similar to the asynchronous abounce_flush() and > > adefer_flush() operations (another solution, moving work from the > > qmgr to the bounce/defer/flush daemons, would require a structural > > redesign and that is not an option at this time). > > > > This long-term solution would involve: > > > > - Cloning a few lines of existing code in src/global/abounce.[hc]. > > > > - Replacing the trace_flush() call in src/qmgr/qmgr_active.c by > > atrace_flush() and a call-back function, similar to the way that > > the abounce_flush() and adefer_flush() operations are used. > > > > That is under a day of work, but I have three deadlines early May. > > > > However, if code can't be changed, then configuration must be > > changed: announce DSN support more selectively. > > Actually, it took much less time, because this was indeed a trivial > job of cloning a few lines of already-existing code. There are no > changes to existing protocols or APIs. > > In the attached patch, all work is done by less than half a dozen > new lines of code: three lines in abounce.c, and four lines in each > of the two identical qmgr_active_done.c files. The remainder of the > patch is spent on comments and declarations. > > You may want to reconsider whether qmgr can/can't be changed because > this change is really very localized. Even if this part of qmgr is > customized (which seems unlikely) this should be easy to paste in > by hand. > > Wietse > > Esta mensagem foi verificada pelo E-mail Protegido Terra. > Atualizado em 20/04/2011 -- Eduardo M. Stelmaszczyk <emspo...@terra.com.br>
