![]() |
Tim Rice wrote: On Tue, 2 Jun 1998, Matthias Apitz wrote: > Tim Rice wrote: > > On Mon, 1 Jun 1998, Jim Whitby wrote: > > > David Woolley wrote: > > > > > > > Anyone have any suggestions? > > > > More info required, please, let me know. > > > > > > Leave it at its current priority - the actual fax protocol is very > > > time critical. > > > > Normaly I would, but *anything* that requires 100% cpu time is too much. > > For a maichne that is idling at less that 5% to jump to 100% and not be > > doing anything is too much priority. > > > > It's not a priority issue. It's correctly set to run at "Fixed Class" > priority. > > Are you on UnixWare 2.1.3? > > I just upgraded to 2.1.3 and now my faxgetty takes 50% of the available > CPU time as reported by "ps -A -ouser,pid,ppid,pcpu,vsz,time,tty,args". > > Truss(1) shows pages of > read(1, 0x080473DC, 2047) = 0 > poll(0x08045BC4, 2, -1) = 1 > read(1, 0x080473DC, 2047) = 0 > poll(0x08045BC4, 2, -1) = 1 > read(1, 0x080473DC, 2047) = 0 > poll(0x08045BC4, 2, -1) = 1 > > looks like the common CONFIG_OPENFIFO problem (but never saw this > in faxgetty(1M) before); what is fd 1 for a file? I couldn't find an open() on fid 1 in the truss output so I guess it's stdout. I recompiled with ./configure --with-CONFIG_OPENFIFO=O_RDWR and it seems to work again. There are other sys-calls too returning a fd: dup(2), socket(2). stdout is pre-opened if the faxgetty(1M) was launched from a sh(1) but I think faxgetty(1M) is disconnecting, e.g. closing fd 1, from the tty. There is no reason for faxgetty(1M) to poll and read from "stdout". matthias