USER_THREADS specifiable in platform options

[siljesu]

In reference to this PR: lf-lang/reactor-c#194

USER_THREADS can now be specified under platform options in target properties:

platform: {
name: Zephyr,
user_threads: 3
}

[erlingrj]

I think this looks okay. For now it is only Zephyr which needs this so its okay to only expose it (i.e. add the NUM_USER_THREADS compile def) when Zephyr is the target platform.

[lhstrh]

Thanks, @siljesu! This looks like a good start. I've requested some changes that are mostly to enforce conventions. Two things that I think we should add:

• a test
• a validator check to point out that this property is ignored if the platform is not Zephyr

[erlingrj]

Did you get a chance to look at Martens suggestion, @siljesu ?

[siljesu]

Yes, take a look at the newest commits.

[erlingrj]

Try running bin/lff test/C from the root of the project to run the Lingua Franca Formatter on all the tests. This will format your newly added tests

[lhstrh]

LGTM!

[erlingrj]

The remaining problem is that, in the test system, I explicitly set the threading=false on all tests executed by Zephyr.

In the file Configurators.java you can look at makeZephyrCompatible, This is used in CZephyrTests.java. I propose to make two separate. One:
makeZephyrCompatibleUnthreaded and makeZephyrCompatible. Use the former for runGenericTests and use the altter for runZephyrTests. This way your test will run, but runGenericTests will still be with unthreaded.

[siljesu]

OK, take a look at the newest commit.

In addition, threaded Zephyr will generate an error "Threaded support on Zephyr is not supported" as of now (from lf_zephyr_support.c). This should probably be removed soon, or else threaded Zephyr tests will always fail regardless. I propose replacing it with a warning "Threaded support on Zephyr is still experimental". If you agree, I can make that change in lf-lang/reactor-c#194.

[erlingrj]

Yes, change it into a warning about it being experimental. These two PRs should be merged together. I will then get the threaded tests running on Zephyr, at which point we can remove that warning

[erlingrj]

OK. I have merged the changes into reactor-c. Now you only have to update your reactor-c pointer to the most recent commit on main and we can actually run the test in CI

[erlingrj]

Silje, there is something not working with the the joining of zephyr threads. The UserThreads program does not exit as expected, rather the main thread is stuct at lf_thread_join even after the worker has completed. I wonder if this could be a QEMU problem. Could you test the UserThreads.lf program on your NXP and verify that it indeed terminates. It should print exit at the end of the execution

[siljesu]

Yes, the program exits as expected when running on an NXP.

[erlingrj]

@lhstrh Lets merge this, finally...

[lhstrh]

Thanks you for your persistence, @erlingrj and @siljesu, and getting this to pass all the checks.

[cmnrd]

I noticed that this PR checked in the binary file test/C/lib/libapp.a but it does not seem to be referenced anywhere. Was this on purpose? If so, why is it needed? I think we should avoid checking in binary files.

[erlingrj]

Ops. That is a mistake!

[erlingrj]

I removed it again in #1803