stmhal: added support for STM32F439 (non DISCO)

[ryannathans]

excluding encryption/hashing hardware acceleration, that's still in the works

[dpgeorge]

This looks very clean. I don't have an F439 so can't test it. @dhylands are you able to test it?

[dpgeorge]

Is this needed?

[ryannathans]

I have no idea what it's for, it must have come from the STM32F439-DISCO port that I used as a base to get started

[dpgeorge]

If it builds without this line, please remove it.

[dhylands]

I don't have a 439 board either.

[ryannathans]

@dpgeorge Removed the define, not sure what its purpose was but it compiled fine without it

[dpgeorge]

@ryannathans what kind of test did you do for this port? I trust that it works, but I can't test it myself so would like to know what has been done from your end.

[ryannathans]

@dbgeorge I've been running my projects on it for a few weeks without any issue (about 15 pcbs running so far). My projects use multiple UARTs, 2 SPIs, some with SD card, take up ~100kB of internal flash in .py files and dump up to a further ~10k of logs on to the flash while running. The code runs identically to when I was using the STM32F405, except it's now faster. I haven't been able to get the tests/ working (failure to connect over the COM port). I'll give them another shot on monday when I'm with the hardware again.

[dpgeorge]

Ok, thanks for that, it sounds great! Do you have any custom C code/modules that you add, or is it just vanilla stmhal built for the F439?

Regarding the test/, they should "just work". Do you have a USB REPL?

[ryannathans]

@dpgeorge Just vanilla at the moment (with that modified storage.c). @david-jimenez is working on a non-public bootloader to run prior to micropython but we're not running it yet. We're also investigating the encryption and hashing acceleration and looking at adding support for that - also not using it yet.

I do have a USB REPL. I was probably just either specifying the wrong COM port or I still had it open...

[dpgeorge]

I'll merge this once the tests are running (and passing).

[ryannathans]

I am unable to get tests to pass on Windows and I haven't worked out how to connect a COM port to my linux VM.

The problem seems to be that in the .exp files all the output has \r\n line endings but the output from the board is all \n endings. I see no other issues otherwise.

Considering there's code for windows support in "run-tests" I expect the tests to work on windows. Is it possible I'm still doing something wrong?

[dhylands]

The .exp files actually end in just \n

The fact that you're seeing \r\n suggests that your git client is doing some type of conversion when checking out the files.

It looks like git has an autocrlf config, and you can also try creating a .gitattribute file with an entry like:

*.exp eol=lf

which according to http://git-scm.com/docs/gitattributes which make line endings on .exp files always have LF.

[ryannathans]

I resolved the problem with the above fix.

[dhylands]

I think that CPython text files created under windows will have \n translated to \r\n.

This looks like it should work under python2 and 3:
http://stackoverflow.com/questions/2536545/how-to-write-unix-end-of-line-characters-in-windows-using-python/23434608#23434608

[dpgeorge]

@ryannathans why the "-1" here? The end of the stack should point to 1 byte above the last byte (the stack is full decending).

[dhylands]

I also noticed that the stm32f411.ld (and the 429 .ld) file has a -1 in it. We should probably fix this.

[dpgeorge]

@dhylands you are right... that's not good! I think the reason it works is because the MCU automagically aligns the pointer (or ignores the lower 2 bits). In that case the -1 loses us 4 bytes of RAM. I've seen this -1 problem in many places, and am pretty sure it's wrong, but would be good to verify it (yet again) and then fix all occurrences in our code.

[dpgeorge]

Merged in ad725a6.

@ryannathans I removed the "-1" from the definition of estack in stm32f439.ld. I also made a few other minor cosmetic changes to this linker script (eg remove whitespace at end of line).