CI problem fixed, CONTRIBUTION.md now includes CI prep.

Author: perlindgren

CONTRIBUTING.md

@@ -1,16 +1,53 @@
 # Contributing
+
 ## New features
+
 New features should go through the [RFC process][rfcs] before a Pull Request is made to this repository.
 
 [rfcs](https://github.com/rtic-rs/rfcs)
 
 ## Bugs
+
 Report bugs by creating an issue in this repository.
 
 ## Pull Requests
+
 Please make pull requests against the master branch.
 
 Always use rebase instead of merge when bringing in changes from master to your feature branch.
 
 ## Writing documentation
+
 Documentation improvements are always welcome. The source for the book is in `book/` and API documentation is generated from the source code.
+
+## CI test preparation
+
+To reduce risk of CI failing for your PR, please make sure that tests passes locally before submitting.
+
+```shell
+> cargo xtask --target all
+```
+
+Will execute `run` tests on your local `qemu` install. (You may also pass a single target `--target thumbv6m-none-eabi/thumbv7m-none-eabi` during your development). These test are quite time consuming as they compile and run all `examples`.
+
+If you have added further tests, you need to add the expected output in the `ci/expected` folder.
+
+```shell
+>  cargo run --example <NAME> --target thumbv7m-none-eabi > ci/expected/<NAME>.run
+```
+
+Internal fail tests can be locally run:
+
+```shell
+> cargo test --tests
+```
+
+If you have added fail tests or changed the expected behavior, the expected output needs to be updated (corresponding `.stderr` files). Inspect the error output, when sure that `ACTUAL OUTPUT` is correct you can re-run the test as:
+
+```shell
+> TRYBUILD=overwrite cargo test --tests
+```
+
+This will update the expected output to match the `ACTUAL OUTPUT`. Please check that the updated files are indeed correct as to avoid regressions.
+
+Once all tests pass you are ready to make a PR.

ci/expected/lockall.run

@@ -1,5 +1,3 @@
 foo: a = 1, b = 2
 baz
-still in foo::lock
 bar: a = 3
-still in foo

ci/expected/lockall_destcruct.run renamed to ci/expected/lockall_destruct.run

@@ -1,5 +1,3 @@
 foo: a = 1, b = 2
 baz
-still in foo::lock
 bar: a = 3
-still in foo

examples/cfg-whole-task.rs

@@ -15,7 +15,6 @@ mod app {
 
     #[shared]
     struct Shared {
-        #[cfg(debug_assertions)] // <- `true` when using the `dev` profile
         count: u32,
         #[cfg(never)]
         unused: u32,
@@ -31,7 +30,6 @@ mod app {
 
         (
             Shared {
-                #[cfg(debug_assertions)]
                 count: 0,
                 #[cfg(never)]
                 unused: 1,

examples/lockall.rs

@@ -35,9 +35,7 @@ mod app {
             *s.a += 1;
             bar::spawn().unwrap();
             baz::spawn().unwrap();
-            hprintln!("still in foo::lock").ok();
         });
-        hprintln!("still in foo").ok();
         debug::exit(debug::EXIT_SUCCESS); // Exit QEMU simulator
     }
 

examples/lockall_destruct.rs

@@ -35,19 +35,17 @@ mod app {
             **a += 1;
             bar::spawn().unwrap();
             baz::spawn().unwrap();
-            hprintln!("still in foo::lock").ok();
         });
-        hprintln!("still in foo").ok();
         debug::exit(debug::EXIT_SUCCESS); // Exit QEMU simulator
     }
 
     #[task(priority = 2, shared = [a])]
     fn bar(mut c: bar::Context) {
         // the higher priority task does still need a critical section
-        let a = c.shared.lock(|s| {
-            *s.a += 1;
+        let a = c.shared.lock(|bar::Shared { a }| {
+            **a += 1;
             // *s.b += 1; `b` not accessible
-            *s.a
+            **a
         });
 
         hprintln!("bar: a = {}", a).unwrap();