Nickez/add debug target

[NickeZ]

This improves the developer experience. You can now run the following shortcuts in separate terminals.

Let the gdb-server and rtt-client run forever:

make gdb-server  # launches JLinkGDBServer

make rtt-client  # connects to gdb server with telnet to get the RTT output

Whenever you built the firmware, rerun this:

make run-debug  # runs gdb and flashes the firmware over the gdb-server before executing it

To compile the debug firmware, run:

make firmware-debug

[benma]

Please add docs for how to use this in BUILD.md.

869864 build-debug/bin/firmware.bin
715956 Build/bin/firmware.bin

This new debug firmware uses 849.48kB of the available 864kB, and adds 150kB on top of the release builds. This is super tight - when we add more code to the firmware, the debug build will be too large to fit. Any ideas what to do about it?

150kB seems generally like way too much for RTT support, any idea what uses so much space? The panic strings, or something else too?

[NickeZ]

Please add docs for how to use this in BUILD.md.

869864 build-debug/bin/firmware.bin 715956 Build/bin/firmware.bin

This new debug firmware uses 849.48kB of the available 864kB, and adds 150kB on top of the release builds. This is super tight - when we add more code to the firmware, the debug build will be too large to fit. Any ideas what to do about it?

150kB seems generally like way too much for RTT support, any idea what uses so much space? The panic strings, or something else too?

Debug builds (by default) check for example overflow in arithmetics, so they introduce a lot of new panic places, with that comes of course lots of static strings and probably some formatting.

Setting debug-assertions = false in [profile.dev] gives a binary size of 774264.

Setting codege-units = 1 further reduces it to 750568.

We could add both, but eventually the release build will be equivalent to the debug build. So then we can just go back to only using release build and ignore the debug build.

edit: I guess there still is a point in having a debug build that enables pretty-printing of panics and enables RTT but keeps everything else the same.

[benma]

Is there any way to make rust_log in C have a sprintf signature like the screen debug equivalent?

void screen_sprintf_debug(int duration, const char* fmt, ...) __attribute__((format(printf, 2, 0)));

then one can much more conveniently log data, like rust_log("result: %d", result);

I guess you could just add a wrapper around rust_log in C that builds the string that is passed to rust_log.

[NickeZ]

Is there any way to make rust_log in C have a sprintf signature like the screen debug equivalent?

void screen_sprintf_debug(int duration, const char* fmt, ...) __attribute__((format(printf, 2, 0)));

then one can much more conveniently log data, like rust_log("result: %d", result);

I guess you could just add a wrapper around rust_log in C that builds the string that is passed to rust_log.

Good idea, I called it util_log because it does some stuff in C-land before calling rust_log.

[NickeZ]

i have some bootloader specific changes that I will create separate PRs of.

[benma]

To run the make run command inside the Docker container, the libncurses5 package must be installed. Would be useful to put it into the Dockerfile.

Yeah, I noticed that as well.

In the end the networking didn't work out for me. So I have to investigate why --network=host doesn't work with docker.

Even if you don't find a solution to the networking thing, it would still help me to just install libncurses5 in the docker image so I can use it. Currently I apt-get update && apt-get install libncurses5 every time.

[NickeZ]

To run the make run command inside the Docker container, the libncurses5 package must be installed. Would be useful to put it into the Dockerfile.

Yeah, I noticed that as well.
In the end the networking didn't work out for me. So I have to investigate why --network=host doesn't work with docker.

Even if you don't find a solution to the networking thing, it would still help me to just install libncurses5 in the docker image so I can use it. Currently I apt-get update && apt-get install libncurses5 every time.

I'll add that in a separate PR.

[benma]

Super mega useful 👏