Add some Rust library building infrastructure by chrysn · Pull Request #16833 · RIOT-OS/RIOT

chrysn · 2021-09-10T09:25:22Z

Contribution description

Building on #16274 this adds a single module that can encapsulate arbitrarily many Rust-based modules.

It's ~~heavily WIP, and will be rebased~~, but the core idea is something that I already want to have discussable:

For every RIOT build, there should be at most one Rust crate built and statically linked. That allows cargo to manage shared dependencies well, and avoid code like formatting being pulled in twice.
The module introduced here is intended to serve as that module if there if crates should come in but no application Rust code is used.
Conceptually, both the RIOT modules and the Rust crates stay separate:
- Pure-Rust RIOT modules are pseudomodules on the RIOT side and pull in the single module.
- The module checks the list of enabled modules, and pulls in the relevant crates.

Note that Rust-only modules have little options to be usable, as otherwise they won't be called into:

They can export XFA sections (like the demo does).
They can export implementations of otherwise weak symbols.
They can do what other modules do, and add extern int module_setup(void); to wherever in the init system they need to be called initially. (Personally I'd discourage that in favor of XFA.)

Testing procedure

It's not a full test, but it's a good one:

$ make BOARD=microbit-v2 -C examples/saul all flash term

TBD

~~It's all a bit rough so far...~~

but these items are still open:

Become less hard-coded -- right now adding a module means changing Cargo.toml, src/lib.rs and the Makefiles
Be available for application Rust code: Application Rust code should easily be able to take over this module's responsibilities.

Issues/PRs references

This is based on #16274, ~~and will be rebased onto that. Only look at the commits not on that other PR, currently it's the last one only.~~

[edit: testing procedure]

chrysn · 2021-11-15T18:44:22Z

Rebased to track #16274

MrKevinWeiss

From the Kconfig perspective, I don't think we need to expose the APPLICATION_RUST_MODULE if we just manually turn off the MODULE_RUST_RIOTMODULES_STANDALONE in the app.test.config.

If the one-liner is too much, we could think about adding a flag to the kconfig.mk similar to the RIOT_CONFIG_DEVELHELP, however, I don't know how scalable this solution would be as we are injecting some make dependency resolution aspects into kconfig.

diff

diff --git a/boards/microbit-v2/Kconfig b/boards/microbit-v2/Kconfig
index 5b6d68fe28..44eacaf949 100644
--- a/boards/microbit-v2/Kconfig
+++ b/boards/microbit-v2/Kconfig
@@ -19,6 +19,7 @@ config BOARD_MICROBIT_V2
     select HAS_VDD_LC_FILTER_REG1
 
     select HAVE_SAUL_GPIO
+    select HAVE_LSM303AGR
 
 source "$(RIOTBOARD)/common/microbit/Kconfig"
 source "$(RIOTBOARD)/common/nrf52/Kconfig"
diff --git a/drivers/Kconfig b/drivers/Kconfig
index e93c467c7a..3efeedec0f 100644
--- a/drivers/Kconfig
+++ b/drivers/Kconfig
@@ -106,6 +106,7 @@ rsource "lpd8808/Kconfig"
 rsource "lpsxxx/Kconfig"
 rsource "lsm6dsl/Kconfig"
 rsource "lsm303dlhc/Kconfig"
+rsource "lsm303agr/Kconfig"
 rsource "ltc4150/Kconfig"
 rsource "mag3110/Kconfig"
 rsource "mhz19/Kconfig"
diff --git a/drivers/lsm303agr/Kconfig b/drivers/lsm303agr/Kconfig
new file mode 100644
index 0000000000..089158ac31
--- /dev/null
+++ b/drivers/lsm303agr/Kconfig
@@ -0,0 +1,20 @@
+# Copyright (c) 2022 HAW Hamburg
+#
+# This file is subject to the terms and conditions of the GNU Lesser
+# General Public License v2.1. See the file LICENSE in the top level
+# directory for more details.
+#
+
+config MODULE_LSM303AGR
+    bool
+    prompt "LSM303AGR 3D accelerometer/magnetometer" if !(MODULE_SAUL_DEFAULT && HAVE_LSM303AGR)
+    default y if (MODULE_SAUL_DEFAULT && HAVE_LSM303AGR)
+    depends on HAS_PERIPH_I2C
+    depends on TEST_KCONFIG
+    select  MODULE_RUST_RIOTMODULES
+    select MODULE_PERIPH_I2C
+
+config HAVE_LSM303AGR
+    bool
+    help
+        Indicates that a lsm303agr sensor is present.
diff --git a/examples/rust-hello-world/app.config.test b/examples/rust-hello-world/app.config.test
new file mode 100644
index 0000000000..35e6e62250
--- /dev/null
+++ b/examples/rust-hello-world/app.config.test
@@ -0,0 +1,2 @@
+CONFIG_MODULE_RUST_RIOTMODULES=y
+CONFIG_MODULE_RUST_RIOTMODULES_STANDALONE=n
diff --git a/sys/Kconfig b/sys/Kconfig
index 37943f299b..2b1e75d678 100644
--- a/sys/Kconfig
+++ b/sys/Kconfig
@@ -79,6 +79,7 @@ rsource "progress_bar/Kconfig"
 rsource "ps/Kconfig"
 rsource "random/Kconfig"
 rsource "rtc_utils/Kconfig"
+rsource "rust_riotmodules/Kconfig"
 rsource "saul_reg/Kconfig"
 rsource "schedstatistics/Kconfig"
 rsource "sema/Kconfig"
diff --git a/sys/rust_riotmodules/Kconfig b/sys/rust_riotmodules/Kconfig
new file mode 100644
index 0000000000..a289a26506
--- /dev/null
+++ b/sys/rust_riotmodules/Kconfig
@@ -0,0 +1,19 @@
+# Copyright (c) 2022 HAW Hamburg
+#
+# This file is subject to the terms and conditions of the GNU Lesser
+# General Public License v2.1. See the file LICENSE in the top level
+# directory for more details.
+#
+
+menuconfig MODULE_RUST_RIOTMODULES
+    bool "RUST RIOT modules"
+    depends on TEST_KCONFIG
+    depends on HAS_RUST_TARGET
+    help
+        This module is used when some module asking to have its code built
+        through rust_riotmodules
+
+config MODULE_RUST_RIOTMODULES_STANDALONE
+    bool "RUST RIOT modules are standalone"
+    default y
+    depends on MODULE_RUST_RIOTMODULES

chrysn · 2022-06-13T11:31:26Z

Thanks, distributed over fixups to give it a try.

Ad the one-liner, would it help to express this as rust_riotmodules depending on either rust_application (preferred) or rust_riotmodules_standalone, but the former is only provided by the applications that have a main module?

This two-line addition is currently necessary for all Kconfig builds of Rust applications that use Rust modules; alternatives are being explored[1] [1]: RIOT-OS#16833 (comment)

gschorcht · 2022-06-13T12:30:16Z

The test run of tesst/periph_timer must fail on esp32-wroom-32 because TIMER_DEV(0) is occupied by ztimer_usec which is enabled by the module libc_gettimeofday which is enabled by default 😟

Some of this *should* be committed (like rust-gcoap following the best pracitce of using the rust_lib_catchall module), but I'd rather have a useful Rust module. This two-line addition is currently necessary for all Kconfig builds of Rust applications that use Rust modules; alternatives are being explored[1] [1]: RIOT-OS#16833 (comment)

maribu

looks good to me :)

doc/doxygen/src/using-rust.md

drivers/lsm303agr/doc.txt

chrysn · 2022-06-30T15:48:17Z

All test failures are unrelated, and have now promising PRs to resolve them (#18210, #18255, #18236).

@maribu wrote:

looks good to me :)

Is that something you'd turn into an ACK? Does it extend to the lsm303agr driver (which is a candidate for a separate PR and primarily in here to give a good example), or would you rather have that removed when squashing / removing the REMOVEME commits?

drivers/lsm303agr/src/lib.rs

sys/shell/democommands/src/lib.rs

chrysn · 2022-07-08T11:17:14Z

@kaspar030, are you happy with the adjustment?

Rerunning tests given the failed one looks like a glitch (being unrelated and not having failed in other recent runs)

kaspar030

ACK.

kaspar030 · 2022-07-09T18:13:40Z

Please squash!

chrysn · 2022-07-09T19:09:25Z

Thanks! Squashed. In removing the "REMOVE ME" commits, I split one of them (the one that made the rust-gcoap example use other Rust modules) off and kept half of it as the now most recent commit: With how the rust_riotmodules developed since those lines were written, it's now best to just use the riot_rustmodules crate all the time (the usual no-op has no deployment impact), as to tolerate any Rust-written RIOT modules that might be activated as they come in (eg. if saul_default gains support for a board's sensor). While murdock is working this off, I'll run another round of self-review whether everything that's in the commits is still in the righ place, and still makes sense.

chrysn · 2022-07-09T19:26:24Z

Updated again to fix some commit wordings, and documentation typos or misreferences.

Co-authored-by: Marian Buschsieweke <[email protected]>

It is enabled by saul_default on microbit-v2. Co-authored-by: Marian Buschsieweke <[email protected]>

Taken and simplified from shell test.

See-Also: RIOT-OS#16833 (comment)

Of all the library crates, this is the only root crate, and responsible for pinning good versions.

chrysn · 2022-07-10T19:35:45Z

Builds and tests passed [edit: I hope the link is right -- I lost the original link, reconstructed the URL, and that was the commit that was tested]. I missed squashing after the latest fixup I mentioned in #16833 (comment) (e4279bd). Squashing that in and skipping builds/tests due to the pre- and post-squash having identical tree hashes.

kaspar030 · 2022-07-10T20:46:02Z

Nice!

chrysn added State: WIP State: The PR is still work-in-progress and its code is not in its final presentable form yet Area: build system Area: Build system Area: pkg Area: External package ports labels Sep 10, 2021

chrysn force-pushed the rust-lib branch from 59179ef to 73f7489 Compare November 15, 2021 18:43

chrysn force-pushed the rust-lib branch from 73f7489 to ff426b4 Compare November 16, 2021 10:03

github-actions bot added Area: Kconfig Area: Kconfig integration Area: tools Area: Supplementary tools labels Nov 16, 2021

chrysn force-pushed the rust-lib branch from ff426b4 to c513288 Compare November 18, 2021 09:26

github-actions bot added the Area: CI Area: Continuous Integration of RIOT components label Nov 18, 2021

chrysn mentioned this pull request Dec 16, 2021

Rust support for RIOT #9799

Closed

11 tasks

chrysn force-pushed the rust-lib branch from c513288 to 196b5bb Compare December 18, 2021 05:53

chrysn force-pushed the rust-lib branch from 196b5bb to d1338f8 Compare January 12, 2022 11:33

chrysn mentioned this pull request Jun 13, 2022

Rust error handling: "consider downloading ..." gets lost #18200

Closed

MrKevinWeiss reviewed Jun 13, 2022

View reviewed changes

github-actions bot added the Area: Kconfig Area: Kconfig integration label Jun 13, 2022

gschorcht mentioned this pull request Jun 13, 2022

cpu/esp32: Remove dependency on module libc_gettimeofday #18202

Merged

maribu reviewed Jun 22, 2022

View reviewed changes

doc/doxygen/src/using-rust.md Outdated Show resolved Hide resolved

drivers/lsm303agr/doc.txt Outdated Show resolved Hide resolved

chrysn commented Jun 22, 2022

View reviewed changes

drivers/lsm303agr/doc.txt Outdated Show resolved Hide resolved

kaspar030 reviewed Jul 1, 2022

View reviewed changes

drivers/lsm303agr/src/lib.rs Show resolved Hide resolved

kaspar030 reviewed Jul 1, 2022

View reviewed changes

sys/shell/democommands/src/lib.rs Outdated Show resolved Hide resolved

kaspar030 approved these changes Jul 9, 2022

View reviewed changes

makefiles/cargo: Drop duplicate workaround remnants

7ee66ad

Rust: Add crates-to-module adapter

ba76c6e

chrysn and others added 7 commits July 10, 2022 21:27

doc: Update for Rust libraries

cd6393a

Co-authored-by: Marian Buschsieweke <[email protected]>

lsm303agr: New sensor driver using external Rust crate

841e4de

It is enabled by saul_default on microbit-v2. Co-authored-by: Marian Buschsieweke <[email protected]>

shell/democommands: Add example Rust module

ebfa6f6

tests: Add rust-as-a-module test

08939cd

Taken and simplified from shell test.

rust_riotmodules: pub use instead of extern crate

68985aa

See-Also: RIOT-OS#16833 (comment)

rust_riotmodules_standalone: This needs its Cargo.lock checked in

5e75f4b

Of all the library crates, this is the only root crate, and responsible for pinning good versions.

rust: Make examples / tests use any Rust RIOT modules

bc8ec6d

chrysn mentioned this pull request Oct 3, 2022

rust: Make dependencies visible #18684

Merged

Conversation

chrysn commented Sep 10, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Contribution description

Testing procedure

TBD

Issues/PRs references

Uh oh!

chrysn commented Nov 15, 2021

Uh oh!

MrKevinWeiss left a comment

Choose a reason for hiding this comment

Uh oh!

chrysn commented Jun 13, 2022

Uh oh!

gschorcht commented Jun 13, 2022

Uh oh!

maribu left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

chrysn commented Jun 30, 2022

Uh oh!

Uh oh!

Uh oh!

chrysn commented Jul 8, 2022

Uh oh!

kaspar030 left a comment

Choose a reason for hiding this comment

Uh oh!

kaspar030 commented Jul 9, 2022

Uh oh!

chrysn commented Jul 9, 2022 via email

Uh oh!

chrysn commented Jul 9, 2022

Uh oh!

chrysn commented Jul 10, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

kaspar030 commented Jul 10, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

chrysn commented Sep 10, 2021 •

edited

Loading

chrysn commented Jul 10, 2022 •

edited

Loading