common: remove rdimon when debug is enabled

[tlyu]

Save about 1.2KiB (depending on GCC version) by redirecting _write directly in libnosys, instead of pulling in librdimon.

Detailed description

• Improves firmware footprint on native BMP when debug output is enabled
• There doesn't appear to be a current need to use librdimon semihosting capabilities
• libnosys I/O functions such as _write can be retargeted directly without using semihosting
• Initializing librdimon semihosting brings in a lot of unused code

I haven't tested on other probe platforms, so I might have unintentionally broken something.

Your checklist for this pull request

• I've read the Code of Conduct
• I've read the guidelines for contributing to this repository
• It builds for hardware native (see Building the firmware)
• It builds as BMDA (see Building the BMDA)
• I've tested it to the best of my ability
• My commit messages provide a useful short description of what the commits do

Closing issues

N/A

[ALTracer]

Please rebase on main -- half of usb_serial.c moved to syscalls.c. I just tried rebasing your commit on main and got a -1344 byte size-diff for meson setup build-stlink --cross-file=cross-file/stlink.ini -Ddebug_output=true -Dbmd_bootloader=false -Dtargets="cortexm,at32f4,rp,stm", which is very nice. While writes definitely don't have to go through DebugMon handler, I expect another bonus: RDP1-locked stlinkv2 (and RDP2-locked stlinkv3) may stop hanging on their first DEBUG_INFO() call, because it no longer uses breakpoints or any (disabled) debug facilities. I can't confirm whether ST disables DebugMon, but it might be the case. Needs testing on ST_BOOTLOADER=1 stlinkv2 and stlinkv3 to confirm the benefit (of allowing users to do all the DEBUG_INFO prints on-probe without BMDA). Or any stm32f103 with normal BMD bootloader but read protection activated after confirming DEBUG_INFO works, to observe that it starts crashing.
If you don't want to handle all the Makefile.inc per-platform, I might handle it myself, checking the result on a few bluepills. f072 can't use debugmon anyways (goes into hardfault), so it's a win-win for it.

 src/platforms/common/stm32/meson.build |  2 +-
 src/platforms/common/syscalls.c        | 65 -----------------------------------------------------------------
 src/platforms/common/usb_serial.c      | 12 ------------
 3 files changed, 1 insertion(+), 78 deletions(-)

[tlyu]

Thanks for taking a look! I will try to make time to work on this a bit next week.

[dragonmux]

Please rebase this on main. We'll need to have a chat with Esden about debug methodologies, but principally this change seems fine to us. If we want to properly re-introduce semihosting based debug output, we can do so anyway without pulling in rdimon.spec using the lessons learned in https://github.com/blackmagic-debug/blackmagic-test-fw-archive/tree/main/semihosting/stm32f411 while preventing pulling in large amounts of unused code and the complicated spaghetti that the newlib implementation makes.

[tlyu]

Rebased. Builds for me, haven't tested yet.

[tlyu]

With -Dtargets=cortexm,nrf,rp,sam,stm,ti and ArmGNUToolchain 13.2.Rel1:

Before:

Memory region         Used Size  Region Size  %age Used
             rom:      123560 B       128 KB     94.27%
             ram:        8228 B        20 KB     40.18%

After:

Memory region         Used Size  Region Size  %age Used
             rom:      122208 B       128 KB     93.24%
             ram:        8048 B        20 KB     39.30%

[dragonmux]

That's a pretty nice saving! We'll get this reviewed in the next 24h we hope, with the aim to merge it once you can confirm it still works as intended.

[tlyu]

Did a quick check scanning and attaching a nRF target with BMP. Debug output to aux CDC as expected. (At least as far as I can recall; I don't remember all the details of how that debug output should look.)

[dragonmux]

This all LGTM, so merging. Thank you for the contribution! Sorry that it took quite so long to get to merging this.