Feature: Add defmt `println!` macro #569

justahero · 2021-09-06T12:15:23Z

This PR addresses #541, adding a defmt::println! macro to always log content irrespective of a log level.

This changes the level property in Frame struct to be of Option type, which requires a number of refactorings to handle both ways to log, either using the log macros (e.g. defmt::info!) or using the defmt::println! macro.
The addition of the defmt::println macro should not break existing functionality.

parser/src/lib.rs

macros/src/function_like/log.rs

jonas-schievink

Looks generally good, but I'm still not sure what our strategy with log should be. Perhaps the best way forward is to not funnel defmt logs through log and remove that code?

decoder/src/lib.rs

decoder/src/frame.rs

jonas-schievink · 2021-09-10T12:49:20Z

macros/src/lib.rs

+#[proc_macro]
+#[proc_macro_error]
+pub fn println(args: TokenStream) -> TokenStream {
+    function_like::println::expand(args)


Is there a reason you couldn't reuse log::expand like the other logging macros? (if so, please add a comment)

I based it on function write::expand which is simlar, println::expand does not check feature flags and sets a different interned string. But I think both versions could be made work in log::expand.

Ah, that makes sense. Yeah, would be good to merge the logic together.

src/lib.rs

jonas-schievink · 2021-09-10T12:51:27Z

decoder/src/log.rs

+                .module_path(module_path)
+                .file(file)
+                .line(line)
+                .build(),


Since you don't call .level(...) here, which level will end up in the Record?

Ah, looks like info: https://docs.rs/log/0.4.14/src/log/lib.rs.html#1157

The alternative is to be more transparent, for example:

log::logger().log( &Record::builder() .args(format_args!("{}", display)) .level(level.unwrap_or(Level::Trace)) .target(&target) .module_path(module_path) .file(file) .line(line) .build(), );

and remove the if else block completely.

It's better to be consistent with the underlying default level in log, therefore the call is now .level(level.unwrap_or(Level::Info)

justahero · 2021-09-10T14:57:39Z

This feature still requires a bit of writing in the defmt book so that users know about how & when to use this.

jonas-schievink

Code LGTM

Still not sure what the best way to proceed with the log interop code is. I'm thinking that deleting it (and handling defmt frames manually) might be the best way forward, otherwise downstream consumers will have println! converted to INFO-level logs, which is kinda wrong.

jonas-schievink · 2021-09-22T17:07:49Z

decoder/src/log.rs

+        crate::Level::Info => Some(Level::Info),
+        crate::Level::Warn => Some(Level::Warn),
+        crate::Level::Error => Some(Level::Error),
+        crate::Level::Println => None,


If we keep the fallback to Info, this branch could just be => Level::Info and we'd remove the Option

japaric · 2021-09-23T16:15:57Z

I have not looked at the PR but wanted to comment on these:

otherwise downstream consumers will have println! converted to INFO-level logs, which is kinda wrong.

I agree with this. defmt::println! should not go through log's machinery (including its filter mechanisms). Those messages should be directly printed to stdout.

Still not sure what the best way to proceed with the log interop code is.

The log adapter is not part of stable decoder library API (there's no stable decoder API) so I'd be OK with removing it.

I think a "stable" log adapter would be useful in the future when there's more of a proper defmt-decoder library.
In my mind, that adapter should only expose defmt's log statements and not defmt::println. The reason for that is that we kind of (already) have semihosting functionality in defmt so defmt::println should use the host's stdout, defmt::eprintln (does not exist) should use the host's stderr, defmt::exit (this is not in defmt proper) should exit the "host process", defmt::info should use the host's logging infrastructure (log crate), and so on and so forth.
(whether we should more clearly separate semihosting functionality from deferred formatting is a separate discussion)

* fix a few lint issues * export `println` macro * add `Tag` variant `Println` * add `expand` function for `println` macro

* rename log level `None` to `Println` * refactor `println` call to use `log::expand` logic * refactor `if else` block to log text

japaric · 2021-09-28T08:47:42Z

decoder/src/log.rs

@@ -75,8 +75,7 @@ impl<'a> DefmtRecord<'a> {
        }

        let timestamp = &target[DEFMT_TARGET_MARKER.len()..];
-
-        Some(Self {
+        (Self{


does this compile? the return type is an option but this no longer has the Some constructor

japaric · 2021-09-28T08:48:03Z

firmware/defmt-test/macros/src/lib.rs

@@ -210,7 +210,7 @@ fn tests_impl(args: TokenStream, input: TokenStream) -> parse::Result<TokenStrea
                test.func.sig.ident
            );
            quote_spanned! {
-                test.func.sig.ident.span() => defmt::info!(#message, __defmt_test_number, __DEFMT_TEST_COUNT);
+                test.func.sig.ident.span() => defmt::println!(#message, __defmt_test_number, __DEFMT_TEST_COUNT);


japaric · 2021-09-28T08:50:31Z

macros/src/function_like/println.rs

+    quote!({
+        match (#(&(#formatting_exprs)),*) {
+            (#(#patterns),*) => {
+                defmt::export::istr(&#format_tag);


this should follow the acquire - release pattern used in the log macros, including the header call. the main difference between the two will be that tag used to constructor the header (there's a make_istr call in the expansion I believe)

This is similar to the `function_like::log::expand` function now.

The `println` macro uses similar logic as the logging macros. When creating a new log frame the log level is optional now, adjusting for the case when a `Frame` is generated from a `println` call.

japaric

Changes look good to me. It's not clear to me, though, what's the output when timestamps are enabled. Could you extend the timestamp snapshot test to include a defmt::println! call?

The other thing that's not obvious from here is whether probe-run will also show the source code location for defmt::println! statements. Including the location would not be standard but certainly useful. I think either way is fine for now personally; same thing for the timestamps.

* refactor `print_location` to take separate arguments instead of a log record

japaric

looks great! feel free to send it to bors with the inline nit addressed 🚀

decoder/src/log.rs

justahero · 2021-09-30T12:47:20Z

bors r+

bors · 2021-09-30T12:52:42Z

Build succeeded:

ci

Urhengulas reviewed Sep 6, 2021

View reviewed changes

parser/src/lib.rs Outdated Show resolved Hide resolved

justahero commented Sep 6, 2021

View reviewed changes

parser/src/lib.rs Outdated Show resolved Hide resolved

justahero commented Sep 6, 2021

View reviewed changes

macros/src/function_like/log.rs Outdated Show resolved Hide resolved

justahero force-pushed the feature/add-defmt-println-macro branch 4 times, most recently from 0e4ac44 to ba979b5 Compare September 9, 2021 14:51

justahero marked this pull request as ready for review September 10, 2021 12:08

jonas-schievink reviewed Sep 10, 2021

View reviewed changes

japaric mentioned this pull request Sep 13, 2021

target-side env_logger-like env filter #519

Merged

7 tasks

jonas-schievink approved these changes Sep 22, 2021

View reviewed changes

justahero force-pushed the feature/add-defmt-println-macro branch 2 times, most recently from fb80fec to 2e04bab Compare September 27, 2021 14:08

justahero added 5 commits September 27, 2021 16:09

Add println macro

038da9a

* fix a few lint issues * export `println` macro * add `Tag` variant `Println` * add `expand` function for `println` macro

Add entry on println! macro to book

e103fd4

Replace defmt::info calls with defmt::println! in defmt-test

17d1d2a

Refactor code, incorporate review eeedback

a1dc9a8

* rename log level `None` to `Println` * refactor `println` call to use `log::expand` logic * refactor `if else` block to log text

Revert log output of sample program

049e93c

justahero marked this pull request as draft September 27, 2021 14:13

japaric self-assigned this Sep 28, 2021

japaric reviewed Sep 28, 2021

View reviewed changes

justahero force-pushed the feature/add-defmt-println-macro branch from 2e04bab to 049e93c Compare September 28, 2021 10:13

justahero added 4 commits September 28, 2021 12:24

Fix return type

e2d7eef

Fix println expand function

b7e0a93

This is similar to the `function_like::log::expand` function now.

Change level in Frame struct to be an Option

bcf2e89

The `println` macro uses similar logic as the logging macros. When creating a new log frame the log level is optional now, adjusting for the case when a `Frame` is generated from a `println` call.

Update affected snapshots that use println

9ab127c

justahero marked this pull request as ready for review September 28, 2021 13:04

japaric reviewed Sep 28, 2021

View reviewed changes

justahero added 7 commits September 28, 2021 16:09

Move println test call to timestamp snapshot program

9a6da4b

Mark println call to keep log

f4b28d9

Trying to figure out how to display non-log frames

5d595ac

Add println function to display non-log frames

eadd17e

* refactor `print_location` to take separate arguments instead of a log record

Code formating

28d7db7

Remove log output

56457fa

Merge println function into log_defmt

888ba88

japaric approved these changes Sep 30, 2021

View reviewed changes

decoder/src/log.rs Outdated Show resolved Hide resolved

Refactor if conditional using if let

f06b2d7

bors bot merged commit 8a6e8ee into main Sep 30, 2021

bors bot deleted the feature/add-defmt-println-macro branch September 30, 2021 12:52

This was referenced Oct 4, 2021

[RFC] defmt::println! #541

Closed

[RFC] how to manage decoder versions? #596

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Feature: Add defmt `println!` macro #569

Feature: Add defmt `println!` macro #569

justahero commented Sep 6, 2021 •

edited

Loading

jonas-schievink left a comment

jonas-schievink Sep 10, 2021

justahero Sep 10, 2021

jonas-schievink Sep 10, 2021

jonas-schievink Sep 10, 2021

jonas-schievink Sep 10, 2021

justahero Sep 10, 2021

justahero Sep 10, 2021

justahero commented Sep 10, 2021

jonas-schievink left a comment

jonas-schievink Sep 22, 2021

japaric commented Sep 23, 2021

japaric Sep 28, 2021

japaric Sep 28, 2021

japaric Sep 28, 2021

japaric left a comment

japaric left a comment

justahero commented Sep 30, 2021

bors bot commented Sep 30, 2021

Feature: Add defmt println! macro #569

Feature: Add defmt println! macro #569

Conversation

justahero commented Sep 6, 2021 • edited Loading

jonas-schievink left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

justahero commented Sep 10, 2021

jonas-schievink left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

japaric commented Sep 23, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

japaric left a comment

Choose a reason for hiding this comment

japaric left a comment

Choose a reason for hiding this comment

justahero commented Sep 30, 2021

bors bot commented Sep 30, 2021

Feature: Add defmt `println!` macro #569

Feature: Add defmt `println!` macro #569

justahero commented Sep 6, 2021 •

edited

Loading