I bought a set of 3 laser sensors and it came with a software and its C++ source code for windows to receive the lasers info through the serial port.

Since the entire project is running on Linux, I need to port the windows code to Linux. At the present moment, I already can receive the data from each of the 3 sensors, but the info is coming concatenated!

Tracking down the original windows code, I found that it is probably happening cause I didn't ported these lines to my Linux version:

Code:

COMMTIMEOUTS CommTimeouts;
[...]
CommTimeouts.ReadIntervalTimeout = 2;
CommTimeouts.ReadTotalTimeoutMultiplier = 0;
CommTimeouts.ReadTotalTimeoutConstant = 2;
CommTimeouts.WriteTotalTimeoutMultiplier = 5;
CommTimeouts.WriteTotalTimeoutConstant = 5;

I've searched already about how to specify the serial port timeouts on Linux, but didn't find a solution.

Does anyone have any idea about how to port these lines to Linux or maybe where to find more info about how to set serial port timeouts on Linux?

Frankly, I'm at a loss. Serial I/O is a byte stream; there is no externally imposed grouping of bytes into records, no record boundaries at all. All serial data is, ultimately, "concatenated".

If you're relying on some sort of timeout mechanism to determine where one chunk of data ends and another begins, I would invite you to consider the possibility of keeping track of when you receive each byte. Use gettimeofday() for this purpose. If the gap between the time of day of reception of one byte and the time of day of reception of the next one is over a certain threshold, then you know you have the beginning of a new chunk of data. By implication, any bytes you have been collecting but have not used so far, not including the new byte, are part of an old chunk.

Hope this helps.

Have a look at this document on serial port programming in Windows and Linux.

Although I have no experience in what you are trying to do it would seem these stuct members are relevant - termio.c_cc[VMIN] being the minimum amount of characters to read and termio.c_cc[V_TIME] the wait time in tenths of a second. This structure is defined in /usr/include/asm/termios.h

I hadn't mentioned tcgetattr() and tcsetattr() because I assumed that the only problem that farofeiro was having was with timeout; therefore, he had resolved all the byte size and parity and baud rate and noncanonical I/O issues; therefore, he was already familiar with tcgetattr() and tcsetattr(); therefore, he was already familiar with the timeout capability offered there.

Again, if you wish to set the per-character timeout as well as the overall "record" timeout, as offered by the Windows API, you'll have to do that yourself. If the purpose is to separate records, set the timeout to whatever you want, read one character at a time, and use gettimeofday().

While the document cited by bgeddy is somewhat useful, you'll get a far better discussion of the relationship between VTIME and VMIN in the tcgetattr() man page.

Perhaps farofeiro is simply being confused by Windows vs Linux end-of-line conventions.
As wje_lq has pointed out, there is little else within a serial byte stream to delimit distinct records. Perhaps the original poster could clarify what he means by 'concatenated' and show some code that he has already tried.

farofeiro, does your laser speak in human-readable text, or does it use non-ASCII binary data? If strictly ASCII, I suggest using a tool such as C-Kermit to try to observe the likes of CR/LF usage for end-of-lines. Writing a test/debug stub that dumps all data as human readable text in a format similar to the output of 'od -cx' can also be extremely illuminating.

--- rod.