[PD-dev] comport and arduino stalls

[B. Bogart]

Hey all,

I've been using the same version of comport in my installations (to
control a pan/tilt unit) for some time. These have been stable for weeks.

Now I'm working on an arduino installation, and I've been unable to make
comport behave for more than a few hours.

I'm using the original arduino USB w/ an ATMEGA8, with Firmata2.1
"SimpleAnalogFirmata".

And using the latest stable [arduino] object.

The patch starts running normally, I see the TX light on the arduino
stays solid.

after running for a while, from 20min to hours, eventually PD will stall
(the patch is no longer responsive, no console messages),  and the TX
light on the arduino will turn off.

Often this sets off the watchdog, and PD starts using up lots of memory
(more the longer it's running).

If I manually kill the pd process in GDB, here is the backtrace:

Program received signal SIGINT, Interrupt.
[Switching to Thread 0xf7c816b0 (LWP 21173)]
0x081025d0 in alist_list (x=0xa1a7884, s=0x8128fb0, argc=112220,
argv=0xf365f008) at x_list.c:111
111	x_list.c: No such file or directory.
	in x_list.c
(gdb) bt
#0  0x081025d0 in alist_list (x=0xa1a7884, s=0x8128fb0, argc=112220,
argv=0xf365f008) at x_list.c:111
#1  0x080b232f in outlet_list (x=0xa1a7810, s=0x8128fb0, argc=112220,
argv=0xf365f008) at m_obj.c:428
#2  0x08101753 in list_append_list (x=0xa1a77d0, s=0x8128fb0,
argc=112220, argv=0xf3b83008) at x_list.c:200
#3  0x080b232f in outlet_list (x=0xa1a7948, s=0x8128fb0, argc=112220,
argv=0xf3b83008) at m_obj.c:428
#4  0x08101ce9 in list_prepend_list (x=0xa1a7908, s=0x0, argc=1,
argv=0xff8353c4) at x_list.c:260
#5  0x080af290 in pd_defaultfloat (x=0xf373b008, f=0) at m_class.c:73
#6  0x080b2527 in outlet_float (x=0xa1a7610, f=0) at m_obj.c:394
#7  0x080b2527 in outlet_float (x=0xa1a74d8, f=0) at m_obj.c:394
#8  0x080f4f83 in trigger_list (x=0xa1a7480, s=0x0, argc=1,
argv=0xff835464) at x_connective.c:979
#9  0x080f50e7 in trigger_float (x=0xa1a7480, f=0) at x_connective.c:1025
#10 0x080b2527 in outlet_float (x=0xa1a6168, f=0) at m_obj.c:394
#11 0x080b2527 in outlet_float (x=0xa1a6018, f=0) at m_obj.c:394
#12 0x080b2527 in outlet_float (x=0xa1949f8, f=0) at m_obj.c:394
#13 0xf66af389 in comport_tick () from /usr/lib/pd/extra/comport.pd_linux
#14 0x080bd7d4 in m_scheduler () at m_sched.c:357
#15 0x080c0805 in sys_main (argc=3, argv=0xff835774) at s_main.c:318
#16 0x080cc0df in main (argc=1, argv=0x0) at s_entry.c:31

Which looks to me like comport is the culprit. A test of simulating the
comport output (when attached to arduino) replacing the comport in the
arduino.pd patch is rock solid, so it does appear comport is the problem.

I'll keep doing some testing.

I'm running debian lenny (stable) with a 2.6.26-1-amd64 kernel (32bit).

I've tried many different versions of comport from SVN, all seem to
cause the same trouble.

So what revision of comport is known to work stable (on linux) over long
periods of simple analog input from arduino?

I'll keep testing, but wanted to see what people have already tested and
is known to work.

Thanks,
B. Bogart