https://forum.pjrc.com/threads/40893-Incorrect-return-for-Serial-write()-on-Teensy-3-6?p=127732#post127732

I'm not sure if this fix has some side-effects - how will other functions react when it does not return zero anymore ?

https://forum.pjrc.com/threads/40893-Incorrect-return-for-Serial-write()-on-Teensy-3-6?p=127764&viewfull=1#post127764

AudioStream.h: Allow inclusion from assembler .S files

Set RTS Watermarks dependend of rx-buffersize

Print.cpp: Add checks for nan, infm and ovr to increase arduino-compatibility

serialX buffersizes

Additional PR for SERIALX_DISABLE follows

Fix fcvtf warning in newer versions of newlib

Fix compilation error with recent newlib/gcc versions because of random()

Fix compilation warnings related to kinetis_hsrun_enable() for older teensies

fixed toolspath

The spaces after TOOLSPATH will append to the variable value. Merged the on Linux comment with previous comment to fix.

EEPROM write on T_3.6 under HSRUN - Drop and Raise

Fix issue #181. Improved pin interrupt dispatch performance.

T3.5/3.6 - SPI has additional Miso/Mosi/SCK pins

Earlier I had pin 20 as the default SCK1, but card shows that pin 32 is
the default pin

nbytes transfer count when minor loop enabled is 10 bits not 5

close hole for race condition

Has to add core_pins.h to get delayMicroseconds()
made empty while()'s on HSRUN state change have a measurable 2us wait
each iteration
added explicit 100us delay leaving HSRUN to allow internal V to
satbilize, seems CPU does this already as 10-50 us made no change, and
100 is not measurable on larger writes.
since eeprom_initialize() blindly restores interrupts amd can run under
HSRUN, moved hsrun_off()