1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
|
BUGS
====
This file describes several bugs known as of today. There do exist probably
some more, but I'm not aware of them right now. Do not hesitate to send me
<freddy@snarl.nl> your bug-reports though. I will gladly add them to this file
and maybe even solve some of them. :-)
Order of bytes broke when reading several tty devices at once
=============================================================
The display of incoming bytes can be broke if you use multiple tty devices to
read from. At the moment I do not have a solution for this problem. Since I use
select() to watch the several tty devices and after the select() I have to
read() them one by one, I can not completely 100% display which bytes came after
which on different tty devices. Take the example below:
$ jpnevulator --ascii --timing-print --tty /dev/ttyS0 --tty /dev/ttyUSB0 --read
2006-05-30 13:23:49.461075: /dev/ttyS0
00 00 05 3B 0D 00 00 05 ...;....
2006-05-30 13:23:49.461113: /dev/ttyUSB0
00 05 3B 0D 00 00 05 3B 0D ..;....;.
2006-05-30 13:23:49.473074: /dev/ttyS0
3B 0D 00 00 05 3B 0D ;....;.
2006-05-30 13:23:49.473105: /dev/ttyUSB0
00 12 05 06 39 00 12 05 06 39 1F 00 22 80 00 0E ....9....9.."...
$
And now see the order in which things really got sent on the line:
/dev/ttyS0:
00 00 05 3B 0D
/dev/ttyUSB0:
00 00 05 3B 0D
/dev/ttyS0:
00 00 05 3B 0D
/dev/ttyUSB0:
00 00 05 3B 0D
/dev/ttyS0:
00 00 05 3B 0D
/dev/ttyUSB0:
00 00 05 3B 0D 00 12 05 06 39 00 12 05 06 39 ...
As you can see /dev/ttyUSB0 receives the echo of all things sent by /dev/ttyS0.
This is exactly what happens. But since there does exist a small time between
the select() who is happy expressing something is available and the read() who
does get the available data, some extra data will be available. I have no idea
on how I can use high level system call like select() and read() and be still
able to put the bytes in the correct order. Anyone an idea?
During the course of 2014 I've received a suggestion from both Colin Foster and
Rolf Freitag to use threads to solve this problem. Just create a thread for every
reading process and a thread for the writing process. I myself am a little bit
sceptic about the solution, as I expect we are only moving the problem away into
the kernel. But it might be worth trying to implement this idea. Since it's quiet
a big change to the core of the software I do not expect myself to dive into it.
But maybe my future self thinks differently. ;)
|
