SerialX.read() ring buffer « LeafLabs Garden

LeafLabs Garden » Support » Maple Hardware Support

SerialX.read() ring buffer

(9 posts) (5 voices)

Started 3 years ago by taylorapgar
Latest reply from mbolivar

taylorapgar
Member

Hello,

I have been having issues with the Serial.read function and I keep thinking that I fixed the problem but it keeps coming back. I am expecting 12 bytes back from a device after I write to a register. The program seems to run fine for a while until all of the sudden Serial.available returns 63. After looking through the forum it sounds like the ring buffer is filled. I am pretty sure its not the device that is causing the issue since I had it working fine with arduino.

My code layout basically looks like

Serial2.write()
if Serial2.available
Serial2.read for 12 bytes

Is there a better way of doing this so that the ring buffer doesn't get filled? Once this happens I am not able to tell which byte is supposed to be which.

Thank you,
Taylor

Posted 3 years ago #
crenn
Moderator

You could always read the buffer until it's empty (using maybe a do while loop in the if statement block) or flush the buffer after you've gotten the data you want.

Posted 3 years ago #
asterix
Member

I've noted the same problem. I've also noticed that when the buffer should be empty, it still returns 63 (from what I can gather).

Posted 3 years ago #
taylorapgar
Member

I switched to flushing the buffer just before writing to the device and then only read if there is exactly 12 bytes available. It seems to be working much better now. Thanks for getting back to me.

I was running into a similar problem with i2c and spi, mainly the bytes read were shifted out of order do these have a similar ring buffer?

Posted 3 years ago #
asterix
Member

I found you can also increase the ringbuffer by modifying maple/hardware/leaflabs/cores/maple/usart.h

#define USART_RX_BUF_SIZE 256 // set it to whatever you like, however if you set it too high I would imagine it might chew some resources!

When you compile the ringbuffer will be increased. Set it to the max bytes you'd like to read perhaps.

Posted 3 years ago #
mbolivar
LeafLabs

if Serial2.available
Serial2.read for 12 bytes

That's not safe. You need to check if Serial2.available() >= 12 before reading 12 bytes, or you'll corrupt the ring buffer contents.

This seems like it's been causing confusion. I think we should make HardwareSerial::read() block until there are bytes to read. I've pushed this into master:

https://github.com/leaflabs/libmaple/commit/15b8a03de97078823ed431e472aedb0988d62220/

Posted 3 years ago #
mbolivar
LeafLabs

I was running into a similar problem with i2c and spi, mainly the bytes read were shifted out of order do these have a similar ring buffer?

No; they don't use ring buffers. Can you open another topic with some sample code showing the bad behavior (out-of-order data)? Please try to keep the sample code short.

Posted 3 years ago #
feurig
Member

So (mbolivar) what you are saying above is that because they are reading more data than is actually available the buffers are getting corrupted therefore they are returning the 63 instead of the correct result (no data) ? It took me a while to decipher that.

Posted 3 years ago #
mbolivar
LeafLabs

feurig: That is what I'm saying. The patch fixes the issue by waiting until data are available. This also makes SerialX behave like SerialUSB, which is nice.

Edit: more precisely, I mean this: in v0.0.12, if you call SerialX.read() when SerialX.available() == 0, then you'll get a garbage character back from the read(), and the next call to available() will return 63.

Posted 3 years ago #

RSS feed for this topic

Reply

You must log in to post.