Re: [misterhouse-users] Large writes to serial object

[Nigel T. <ni...@pa...>]

On Tue, Mar 12, 2002 at 02:11:13PM -0500, Bbi...@ao... wrote:
> Hi Nigel and Bruce,
>=20
> I got 4111 bytes on RH7 as well. I am guessing there is a buffer somewher=
e in=20
> the serial driver (and that changing it might be a real nuisance) or POSI=
X.pm=20
> (likewise, the write method is an Autoloaded XS).

I'm so pleased you've managed to duplicate it. I've been literally dreaming=
=20
about this problem, and things are getting bad when you start debugging in
your sleep.

> I must have tested on Windows since it was easy to duplicate once I set u=
p to=20
> do it. =20

Which kernel version are you running? I'm running 2.2.18pre21.

> I could not find any correlation with flow control. It is certainly a buf=
fer=20
> overrun, however, and there is a workaround - the SerialPort write_drain=
=20
> method will tell you when the buffer is empty. See the example below. I'm=
 not=20
> sure if any of this should happen automatically, though (at 9600 baud, 40=
00=20
> characters is about 4 seconds).

Yes I did some playing with flow control last night. By the way is there any
way in MH to set up rts/cts flow control? I've tried setting serial?_handsh=
ake
to rts in mh.private.ini, but I haven't had a chance to follow the code thr=
ough
to see if it is working or not...... and it still only sends 4111 bytes :-)

If I can't find the bug I'll have to try the workaround, but its a bit ugly.

Nigel