view target-utils/libprintf/README @ 923:10b4bed10192

gsm-fw/L1: fix for the DSP patch corruption bug The L1 code we got from the LoCosto fw contains a feature for DSP CPU load measurement. This feature is a LoCosto-ism, i.e., not applicable to earlier DBB chips (Calypso) with their respective earlier DSP ROMs. Most of the code dealing with that feature is conditionalized as #if (DSP >= 38), but one spot was missed, and the MCU code was writing into an API word dealing with this feature. In TCS211 this DSP API word happens to be used by the DSP code patch, hence that write was corrupting the patched DSP code.
author Mychaela Falconia <falcon@ivan.Harhan.ORG>
date Mon, 19 Oct 2015 17:13:56 +0000
parents 40f607bb0a2c
children
line wrap: on
line source

The present libprintf is a very light printf implementation that is well-suited
for simple bare-metal programs like loadagent; in the present case it overrides
the much heavier printf implementation in newlib.  Programs like the present
loadagent only need printf in order to scribble on the serial console port,
and the most sensible implementation is to have the "character output" function
from the guts of printf point directly to the physical UART output routine, or
a trivial wrapper that turns \n into \r\n.  In contrast, newlib's version would
pull in the complete FILE table infrastructure and malloc etc - maybe OK for
more complex embedded programs that use those facilities for other things under
a bona fide RTOS, but it would be disgusting to pull that stuff in for a
minimal program like ours.

The present printf implementation has been used earlier by the same author
(Michael Spacefalcon) in the StarMON family of PowerPC bootloaders, and in my
MC68x302-based SDSL CPE devices (Hack-o-Rocket and OSDCU).