ports/unix: Add full uos.dupterm support. - #6080

On CPython I guess you can hook into stdout by redefining sys.stdout... and maybe that's how MicroPython should have done it from the start, instead of adding uos.dupterm() (although it'd still need an efficient multiplexer, eg sys.stdout = micropython.stream_multiplex(sys.stdout, machine.UART(...)), not to mention handling sys.stdin).

Anyway, the way it is now sys.stdout is before the dupterm multiplexer, so writing to sys.stdout goes to all dupterm streams (similar for reading). And what this PR does is adds raw-stdout/stdin objects, ie they do not go via dupterm. On bare-metal the equivalent would be a UART or USB_VCP object.

It just seems like there's a lot of complexity here with two different kinds of stdout/stdin objects (a dupterm one and a raw one). But I don't really see a way around that.

Would it be possible to reuse the extmod/vfs_posix_file.c code to implement the raw stdin/stdout? That code has compile-time dupterm support but maybe that dupterm support can be made optional at runtime (via a flag in the stdio object) so that the same code can be used for both sets of stdio objects?

Thanks for the pointer towards extmod/vfs_posix_file.c - it already has the stdio object instantiated

const mp_obj_vfs_posix_file_t mp_sys_stdin_obj = {{&mp_type_textio}, STDIN_FILENO};
const mp_obj_vfs_posix_file_t mp_sys_stdout_obj = {{&mp_type_textio}, STDOUT_FILENO};
const mp_obj_vfs_posix_file_t mp_sys_stderr_obj = {{&mp_type_textio}, STDERR_FILENO};

I should be able to clean up this patch reusing them.

@dpgeorge there's a catch with using the mp_sys_stdin_obj etc objects above as i/o for dupterm, other things read/write to these objects - and writes to stdout/stderr are redirected to dupterm (hello recursion).

I've resolved this by adding a relatively small amount of extra logic to make a new const mp_obj_vfs_posix_file_t mp_sys_raw_stdio_obj which will automatically read/write to the correct fd and not go via dupterm - this new raw_stdio can be used for unix_mphal.c both with and without dupterm.

It feels like a bit of a tacked-on-the-side addition, but it does make for less duplication than I had previously.

Now I'm also getting a lot of travis fails, for one the minimal port doesn't have the new file handle created, presumably due to a lack of one/both of MICROPY_VFS_POSIX / MICROPY_VFS_POSIX_FILE
Some of the others are failing a number of unit tests, don't know why yet.

Code size report:

Reference:  tools/ci.sh: Increase qemu_arm test run timeout. [5c00edc]
Comparison: ports/unix: Add full uos.dupterm support. [merge of 779e38b]
  mpy-cross:    +0 +0.000% 
   bare-arm:    +0 +0.000% 
minimal x86:    +0 +0.000% 
   unix x64: +2492 +0.291% standard[incl +140(data) +32(bss)]
      stm32:    +8 +0.002% PYBV10
      esp32:    +0 +0.000% ESP32_GENERIC
     mimxrt:    +0 +0.000% TEENSY40
        rp2:    +0 +0.000% RPI_PICO_W
       samd:    +0 +0.000% ADAFRUIT_ITSYBITSY_M4_EXPRESS
  qemu rv32:    +0 +0.000% VIRT_RV32

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.15%. Comparing base (2dc2e30) to head (66a4bc4).

@@            Coverage Diff             @@
##           master    #6080      +/-   ##
==========================================
- Coverage   98.46%   98.15%   -0.32%     
==========================================
  Files         176      176              
  Lines       22784    22890     +106     
==========================================
+ Hits        22435    22467      +32     
- Misses        349      423      +74     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.