Skip to contentSkip to conversation

rust: macros: fix `make rusttest` build on macOS

PATCH Superseded Tamir Duberstein 3 participants 9 messages Feb 7, 2025 – Feb 8, 2025

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

cover rust: macros: fix `make rusttest` build on macOS

---
Tamir Duberstein (2):
rust: macros: improve panic messages
rust: macros: fix make rusttest build on macOS

 rust/macros/module.rs | 23 ++++++++++++-----------
 1 file changed, 12 insertions(+), 11 deletions(-)

---
base-commit: beeb78d46249cab8b2b8359a2ce8fa5376b5ad2d
change-id: 20250207-macros-section-f7cddc7bbdea

Best regards,

Signature · 1 line
Tamir Duberstein <[email protected]>
#1 lore raw

PATCH 1/2 rust: macros: improve panic messages

Include unexpected input on parsing failures. This has the side effect of avoiding a spurious rust-analyzer warning:

Variable None should have snake_case name, e.g. none

---
 rust/macros/module.rs | 21 +++++++++++----------
 1 file changed, 11 insertions(+), 10 deletions(-)
rust/macros/module.rs +11 -10
index cdf94f4982df..ca1b7e6a71ff 100644
@@ -11,12 +11,14 @@ fn expect_string_array(it: &mut token_stream::IntoIter) -> Vec<String> {
     let mut it = group.stream().into_iter();
 
     while let Some(val) = try_string(&mut it) {
-        assert!(val.is_ascii(), "Expected ASCII string");
+        assert!(val.is_ascii(), "Expected ASCII string, got {}", val);
         values.push(val);
-        match it.next() {
-            Some(TokenTree::Punct(punct)) => assert_eq!(punct.as_char(), ','),
-            None => break,
-            _ => panic!("Expected ',' or end of array"),
+        let Some(token) = it.next() else {
+            break;
+        };
+        match token {
+            TokenTree::Punct(punct) => assert_eq!(punct.as_char(), ','),
+            token => panic!("Expected ',' or end of array, got {}", token),
         }
     }
     values
@@ -116,11 +118,10 @@ fn parse(it: &mut token_stream::IntoIter) -> Self {
         const REQUIRED_KEYS: &[&str] = &["type", "name", "license"];
         let mut seen_keys = Vec::new();
 
-        loop {
-            let key = match it.next() {
-                Some(TokenTree::Ident(ident)) => ident.to_string(),
-                Some(_) => panic!("Expected Ident or end"),
-                None => break,
+        while let Some(token) = it.next() {
+            let key = match token {
+                TokenTree::Ident(ident) => ident.to_string(),
+                token => panic!("Expected Ident or end, got {}", token),
             };
 
             if seen_keys.contains(&key) {
Signature · 1 line
2.48.1
#2 lore raw

Re: [PATCH 1/2] rust: macros: improve panic messages

Quoted 5 lines · Tamir Duberstein

Include unexpected input on parsing failures. This has the side effect of avoiding a spurious rust-analyzer warning:

Variable None should have snake_case name, e.g. none

Hmm... That should be solved independently, but sure.

In any case, how is this related to the second patch in the series? i.e. do you need both to solve the macOS issue?

rust/macros/module.rs fn expect_string_array(it: &mut token_stream::IntoIter) -> Vec<String> { · line 21

Quoted 6 lines
rust/macros/module.rs +6 -0
+        let Some(token) = it.next() else {
+            break;
+        };
+        match token {
+            TokenTree::Punct(punct) => assert_eq!(punct.as_char(), ','),
+            token => panic!("Expected ',' or end of array, got {}", token),

Do we want to shadow here?

Also, I think you could write {token}. Same above.

Could you please show how the new output would look like in the commit message?

Cheers,
Miguel

#3 lore raw

Re: [PATCH 1/2] rust: macros: improve panic messages

Quoted 6 lines · Miguel Ojeda
Quoted 4 lines · Tamir Duberstein

Include unexpected input on parsing failures. This has the side effect of avoiding a spurious rust-analyzer warning:

Quoted 6 lines
Quoted 1 line

Variable None should have snake_case name, e.g. none

Hmm... That should be solved independently, but sure.

In any case, how is this related to the second patch in the series? i.e. do you need both to solve the macOS issue?

No, it's not required - you're right to point that out. I'll send it separately. Is it very inconvenient to consider the other patch on its own?

Quoted 8 lines

rust/macros/module.rs fn expect_string_array(it: &mut token_stream::IntoIter) -> Vec<String> { · line 21

Quoted 6 lines
rust/macros/module.rs +6 -0
+        let Some(token) = it.next() else {
+            break;
+        };
+        match token {
+            TokenTree::Punct(punct) => assert_eq!(punct.as_char(), ','),
+            token => panic!("Expected ',' or end of array, got {}", token),

Do we want to shadow here?

I personally like this style because the shadow and shadowed are the same variable and the presence of a binding in the match expression hints to the reader that it's used somewhere in the body of the match arm. Do you have another preference?

Quoted 1 line

Also, I think you could write {token}. Same above.

Good call. Will make it so.

Quoted 1 line

Could you please show how the new output would look like in the commit message?

Yes, I will include before and after output.

Quoted 2 lines

Cheers,
Miguel

#4 lore raw

PATCH 2/2 rust: macros: fix `make rusttest` build on macOS

In reply to Tamir Duberstein

Do not emit #[link_section = ".modinfo"] on macOS (i.e. when building userspace tests); .modinfo is not a legal section specifier in mach-o.

Before this change tests failed to compile:

---- ../rust/macros/lib.rs - module (line 66) stdout ----
rustc-LLVM ERROR: Global variable '_ZN8rust_out13__module_init13__module_init27__MY_DEVICE_DRIVER_MODULE_017h141f80536770e0d4E' has an invalid section specifier '.modinfo': mach-o section specifier requires a segment and section separated by a comma. Couldn't compile the test.
---- ../rust/macros/lib.rs - module (line 33) stdout ----
rustc-LLVM ERROR: Global variable '_ZN8rust_out13__module_init13__module_init20__MY_KERNEL_MODULE_017h5d79189564b41e07E' has an invalid section specifier '.modinfo': mach-o section specifier requires a segment and section separated by a comma. Couldn't compile the test.

---
 rust/macros/module.rs | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
rust/macros/module.rs +1 -1
index ca1b7e6a71ff..f09431614f2b 100644
@@ -58,7 +58,7 @@ fn emit_base(&mut self, field: &str, content: &str, builtin: bool) {
             "
                 {cfg}
                 #[doc(hidden)]
-                #[link_section = \".modinfo\"]
+                #[cfg_attr(not(target_os = \"macos\"), link_section = \".modinfo\")]
                 #[used]
                 pub static __{module}_{counter}: [u8; {length}] = *{string};
             ",
Signature · 1 line
2.48.1
#5 lore raw
Quoted 3 lines · Tamir Duberstein

Do not emit #[link_section = ".modinfo"] on macOS (i.e. when building userspace tests); .modinfo is not a legal section specifier in mach-o.

I thought the changes we already applied were the ones needed for macOS support. Oh, well...

Cheers,
Miguel

#6 lore raw
Quoted 11 lines · Miguel Ojeda
Quoted 3 lines · Tamir Duberstein

Do not emit #[link_section = ".modinfo"] on macOS (i.e. when building userspace tests); .modinfo is not a legal section specifier in mach-o.

I thought the changes we already applied were the ones needed for macOS support. Oh, well...

Cheers,
Miguel

Yeah, that's on me. I didn't know about (and hadn't tried) make rusttest until the cstr series[0].

[0] https://lore.kernel.org/all/[email protected]/t/#u

#7 lore raw
In reply to Tamir Duberstein
Quoted 5 lines · Tamir Duberstein

---
Tamir Duberstein (2):
rust: macros: improve panic messages
rust: macros: fix make rusttest build on macOS

Very minor nit, patch 0/X does not need a signed-off-by, and if you find that you don't need any text for patch 0/X, there's no need to send a patch 0/X at all!

Hopet this helps for future submissions.

thanks,

greg k-h

#8 lore raw
Quoted 12 lines · Greg KH
Quoted 5 lines · Tamir Duberstein

---
Tamir Duberstein (2):
rust: macros: improve panic messages
rust: macros: fix make rusttest build on macOS

Very minor nit, patch 0/X does not need a signed-off-by, and if you find that you don't need any text for patch 0/X, there's no need to send a patch 0/X at all!

Quoted 5 lines

Hopet this helps for future submissions.

thanks,

greg k-h

Thanks Greg. This was me holding b4 wrong :(
Tamir