Enable flow control in Greentea

[marcuschangarm]

Description

Flow control is enabled in Greentea for targets that has
console-uart-flow-control set.

Pull request type

[X] Fix
[ ] Refactor
[ ] New target
[ ] Feature
[ ] Breaking change

[marcuschangarm]

@cmonr @0xc0170

Copy-pasted from mbed_retarget.cpp

[cmonr]

/morph build

[cmonr]

@marcuschangarm Confirming, this was needed because local testing for the #6547 wasn't working, correct?

[mbed-ci]

Build : SUCCESS

Build number : 1780
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/6660/

Triggering tests

/morph test
/morph uvisor-test
/morph export-build
/morph mbed2-build

[marcuschangarm]

It's actually to take full advantage of #6603.

#6547 Depends on both.

[mbed-ci]

Exporter Build : FAILURE

Build number : 1422
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/exporter/6660/

[0xc0170]

/morph export-build

[kjbracey]

This is pretty grim. Trying to use a serial port both through stdin/stdout, and through a separate Serial object is a hack, and this digs deeper

And this defeats buffering if set with platform.stdio-buffered.

Can't you just make greentea use stdout? putc/getc, rather than greentea_serial->putc/getc?

[kjbracey]

Also, we have platforms that don't use serial for the console, right? We've just done the console redirection for that, and this bypasses it. The greentea client code in principle should work on something other than serial ports.

[0xc0170]

@kjbracey-arm This is not related to the patch, however good question! I tracked where this serial console was added, in the original greentea client ARMmbed/greentea-client@ff62371 - main drive to have it in interrupt context. This can be fixed

[kjbracey]

Okay, that looks like there was quite a specific requirement to do this raw IRQ-context output.

That's now possible as of mbed OS 5.8 by cooperating with mbed_retarget.cpp - you can now do

write(STDOUT_FILENO, buffer, len);

That is interrupt safe as long as the underlying console is, and the default non-buffered case (DirectSerial) is. Would currently fail on buffered UARTSerial, but thinking about improving that for fault and error prints.

[mbed-ci]

Exporter Build : FAILURE

Build number : 1423
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/exporter/6660/

[0xc0170]

/morph export-build

[0xc0170]

@studavekar Pipe closed failure , please review

[mbed-ci]

Test : SUCCESS

Build number : 1590
Test logs :http://mbed-os-logs.s3-website-us-west-1.amazonaws.com/?prefix=logs/6660/1590

[mbed-ci]

Exporter Build : SUCCESS

Build number : 1430
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/exporter/6660/

[cmonr]

@marcuschangarm Could you answer @kjbracey-arm's questions?

[marcuschangarm]

I think refactoring Greentea is out of scope of this PR which only adds flow control to the serial connection already present.

[kjbracey]

I'd agree - this patch is safe enough as a fix, as it only activates with a newly-added flag. Messing with it further wouldn't be, so let's just raise an issue for the awkward serial port use.