GitLab

Using __errno() function, and redirect it to use mp_stream_errno from
stream module. This is pre-requisite for integrating with 3rd-party libs,
like BerkeleyDB.

Requires "embedded" BerkeleyDB BTree implementation.

It allows to access files via a virtual method tables and thus can integrate
with MicroPython's stream objects.

Some boards (like the GHI Electronics G30 Dev Board) don't use
NSS at all and rather just use GPIO chip selects.

When compiling with msys2's gcc there's no need to apply the binary fmode
so adjust the Makefile to reflect that.
When compiling with mingw we need to include malloc.h since there is no
alloca.h, and the 64bit detection in mpconfigport.h needs some adjustment.

To filter out even prototypes of mp_stream_posix_*() functions, which
require POSIX types like ssize_t & off_t, which may be not available in
some ports.

As required for related functions in stream.h.

Helpful when porting existing C libraries to MicroPython. abort()ing in
embedded environment isn't a good idea, so when compiling such library,
-Dabort=abort_ option can be given to redirect standard abort() to this
"safe" version.

Previoussly such read() and write() methods were used by modussl_axtls,
move to py/stream for reuse.

No-op for this object.

The configuration bits for the UART register were wrong and the parity
couldn't be enabled, because the exist_parity member hasn't been updated. I
took this ESP8266 register description (http://esp8266.ru/esp8266-uart-reg/)
as reference.

Verification has been done with a logic analyzer.

Required to pass bytes_compare3.py (opptional warnings) on devices.

This allows to use printf() in a any source file with unix port, for quick
debugging.

Order out-of-bounds check, completion check, and increment in the right way.

There's single str_index_to_ptr() function, called for both bytes and
unicode objects, so should handle each properly.

We have adopted POSIX-compatible error numbers as MicroPython's native.

The idea behind decrease is: bytecode and other static data is also kept on
heap, and can easily become half of heap, then setting threshold to half of
heap will have null effect - GC will happen on complete heap exhaustion like
before. But exactly in such config maintaining heap defragmented is very
important, so lower threshold to accommodate that.