UnbufferedSerial does not expose enable_* functions from SerialBase; behavior not predictable when exposed

[petertorelli]

Description of defect

The documentation claims UnbufferedSerial has two functions, enable_input and enable_output. However, they are not defined in the base class and SerialBase is not accessible as it is a private base class. There is also an issue where using the enable_* functions on an STM board has no problem, but running on a Cypress board corrupts the port.

Target(s) affected by this defect ?

I tested on a CY8CPROTO_062_4343W and NUCLEO_F429ZI, but I assume all since it is a header issue.

Toolchain(s) (name and version) displaying this defect ?

GCC_ARM

What version of Mbed-os are you using (tag or sha) ?

6.5.0

What version(s) of tools are you using. List all that apply (E.g. mbed-cli)

mbed: 1.10.4

How is this defect reproduced ?

#include "mbed.h"

UnbufferedSerial pc(USBTX, USBRX);

int
main()
{
	printf("Serial enabled, disabling.\n");
	pc.sync();
	pc.enable_input(false);
	pc.enable_output(false);
	pc.enable_input(true);
	pc.enable_output(true);
	printf("Serial enabled, this should be visible.\n");

}

This will fail to compile with this error

./main.cpp:13:5: error: request for member 'enable_output' is ambiguous
  pc.enable_output(true);
     ^~~~~~~~~~~~~
In file included from ./mbed-os/connectivity/netsocket/include/netsocket/nsapi_ppp.h:22,
                 from ./mbed-os/mbed.h:28,
                 from ./main.cpp:1:
./mbed-os/platform/include/platform/FileHandle.h:210:17: note: candidates are: 'virtual int mbed::FileHandle::enable_output(bool)'
     virtual int enable_output(bool enabled)
                 ^~~~~~~~~~~~~
In file included from ./mbed-os/drivers/include/drivers/UnbufferedSerial.h:26,
                 from ./mbed-os/mbed.h:71,
                 from ./main.cpp:1:
./mbed-os/drivers/include/drivers/SerialBase.h:144:10: note:                 'void mbed::SerialBase::enable_output(bool)'
     void enable_output(bool enable = true);
          ^~~~~~~~~~~~~

It can be fixed by adding these two lines to UnbufferedSerial.h:

    using SerialBase::enable_input;
    using SerialBase::enable_output;

Furthermore, this code works with the fixes on the NUCLEO, but it fails on the CYPRESS board (screen fills with garbage characters). I suspect the deinit/reinit of the serial object is the culprit in the Cypress case.

[ciarmcom]

@petertorelli thank you for raising this issue.Please take a look at the following comments:

Could you add some more detail to the description? A good description should be at least 25 words.

NOTE: If there are fields which are not applicable then please just add 'n/a' or 'None'. This indicates to us that at least all the fields have been considered.
Please update the issue header with the missing information, the issue will not be mirrored to our internal defect tracking system or investigated until this has been fully resolved.

[ciarmcom]

Thank you for raising this detailed GitHub issue. I am now notifying our internal issue triagers.
Internal Jira reference: https://jira.arm.com/browse/IOTOSM-3028

[0xc0170]

@ARMmbed/mbed-os-hal

[ithinuel]

This is higher level abstraction than the HAL.

@kjbracey-arm for FileHandle::enable_input and @hugueskamba for UnbufferedSerial.

[kjbracey]

running on a Cypress board corrupts the port.

It looks like you're running two drivers on the port - your pc and the C library stdout.

That's likely to cause problems, not directly connected to the enable calls. You need to have only one driver.

Either arrange for the C library to use your pc rather than its own driver by defining mbed_override_console, or get access to the C library's driver by calling mbed_file_handle(STDOUT_FILENO). Both are in mbed_retarget.h.

eg

#include "mbed.h"

int
main()
{
        // both are probably the same, but let's be picky...
        FileHandle *console_in = mbed_file_handle(STDIN_FILENO);
        FileHandle *console_out = mbed_file_handle(STDOUT_FILENO);
	printf("Serial enabled, disabling.\n");
	console_out->sync();
	console_in->enable_input(false);
	console_out->enable_output(false);
	console_in->enable_input(true);
	console_out->enable_output(true);
	printf("Serial enabled, this should be visible.\n");

}

That would sidestep the UnbufferedSerial problem, which still needs fixing.

[kjbracey]

Actually my code above quite possibly doesn't work, if the console is unbuffered, because it seems likely that the same thing causing an ambiguity error would mean UnbufferedSerial hasn't overridden the default "just return error" implementation of the virtual method FileHandle::enable_input.

I think the using SerialBase::enable_input suggested above would sort that out too, but I'm not totally sure.

[ithinuel]

Is that the gist of the issue we're seeing : https://godbolt.org/z/aoE78x ?

[kjbracey]

Interestingly, it looks like the using isn't sufficient to perform the virtual override there. Most curious. I've failed to find anything clear on the topic.