Trouble with recent USB CDC library

andrewg · Posted: Sun May 10, 2009 3:45 am

Is anyone else using and having trouble with the CCS USB CDC library? The version history in "usb_cdc.h" is 7 April 2009 in the one I have.

My problem is that only the first character of each packet of data I transmit is being sent. Very, very occasionally, the whole packet will be correctly sent.

I've reverted to the next most recent copy of the library I have, which is dated 27 October 2005, and everything is working fine.

If nobody else is having problems, then I guess I should take a closer look at my code, otherwise I'll add it to my rapidly growing list of tech support issues to raise with CCS.

Thanks.
_________________
Andrew

gnaur · Guest

Yes, I have a similar problem, I do get more data sent, it will send for a few seconds ok but eventually will lock up and not send any more. It is still receiving data ok.

andrewg · Posted: Sun May 10, 2009 9:24 pm

Thanks for reminding me that receiving works fine - at least I'm not having any problems there. What is the most recent date in your usb_cdc.h file?
_________________
Andrew

FvM · Joined: 27 Aug 2008 Posts: 2337 Location: Germany

As reported earlier in the forum, I had some problems with interrupt handling in my implementation, of the CDC example.
http://www.ccsinfo.com/forum/viewtopic.php?t=36100

Although the issue couldn't be completely clarified, it was solved for my application by adding a reenable global interrupts after USB processing in the main loop.

I started application development with PCH V4.079, my most recent (production stable) version is based on V4.081, I didn't recompile it with newer PCD versions, so I can't say, if new problems adds since then.

I don't plan an update right now, but if I'm aware of any new issues in the PCH versions since V4.082, I'll report it in the forum.

andrewg · Posted: Mon May 11, 2009 1:06 am

That's interesting because, due to timing issues elsewhere in my project, I've been forced to use USB polling and have followed the instructions in the USB CDC source to do so. It means that my project doesn't have any of those "interrupts disabled" messages anywhere. I'm pretty sure it's a logic problem somewhere in the CDC support code.
_________________
Andrew

FvM · Joined: 27 Aug 2008 Posts: 2337 Location: Germany

It may be also the case, that the poll USB solution doesn't work correctly.
The existence of an option doesn't necessarily mean, that it has been ever used or tested.

I wasn't referring to those "interrupt diasbled" messages, rather than global interrupts actually staying disabled after calling CDC functions to send data in some cases.

andrewg · Posted: Mon May 11, 2009 6:27 am

I thought your problem was the CCS-generated code related to those "interrupt disabled" messages that disables and re-enables global interrupts. That that code wasn't correctly re-enabling global interrupts. Given that my code doesn't have any re-entrant code issues, I was trying to say that the issues you're having with the CDC code probably aren't the same as the issues I'm having.

Now polling could very well be the problem in my case. The fact that the old 2005 code is working fine doesn't mean that the new 2009 code will! Smile

I'm debating whether to spend a bit more time analyzing the CDC code, or just email CCS with a "doesn't work" message. I'd much rather have something a bit more substantial before reporting it.
_________________
Andrew