Checkpatch errors that should be downgraded

Checkpatch errors that should be downgraded

[David Howells]

In the trace headers, the following should probably be ignored:

	CHECK: Lines should not end with a '('
	#73: FILE: include/trace/events/rxrpc.h:1947:
	+	    TP_STRUCT__entry(

	CHECK: Lines should not end with a '('
	#80: FILE: include/trace/events/rxrpc.h:1954:
	+	    TP_fast_assign(

as it's standard practice to bump the contents onto the next line for these
two macros.

Also checkpatch is wrong in giving this error:

	ERROR: Macros with complex values should be enclosed in parentheses
	#37: FILE: include/trace/events/rxrpc.h:425:
	+#define rxrpc_sack_traces \
	+	EM(rxrpc_sack_advance,			"ADV")	\
	+	EM(rxrpc_sack_fill,			"FIL")	\
	+	EM(rxrpc_sack_nack,			"NAK")	\
	+	EM(rxrpc_sack_none,			"---")	\
	+	E_(rxrpc_sack_oos,			"OOS")

as the contents, when found in a trace header, are going to be used to
generate a string table and maybe an enum.  It's standard practice, so if it
could be ignored in trace headers or at least downgraded to a warning?

Thanks,
David

Re: Checkpatch errors that should be downgraded

[Joe Perches]

On Thu, 2023-02-02 at 11:16 +0000, David Howells wrote:
> In the trace headers, the following should probably be ignored:
> 
> 	CHECK: Lines should not end with a '('
> 	#73: FILE: include/trace/events/rxrpc.h:1947:
> 	+	    TP_STRUCT__entry(
> 
> 	CHECK: Lines should not end with a '('
> 	#80: FILE: include/trace/events/rxrpc.h:1954:
> 	+	    TP_fast_assign(
> 
> as it's standard practice to bump the contents onto the next line for these
> two macros.

trace has its own grammar.
Might as well ignore all of it.

> Also checkpatch is wrong in giving this error:
> 
> 	ERROR: Macros with complex values should be enclosed in parentheses
> 	#37: FILE: include/trace/events/rxrpc.h:425:
> 	+#define rxrpc_sack_traces \
> 	+	EM(rxrpc_sack_advance,			"ADV")	\
> 	+	EM(rxrpc_sack_fill,			"FIL")	\
> 	+	EM(rxrpc_sack_nack,			"NAK")	\
> 	+	EM(rxrpc_sack_none,			"---")	\
> 	+	E_(rxrpc_sack_oos,			"OOS")
> 
> as the contents, when found in a trace header, are going to be used to
> generate a string table and maybe an enum.  It's standard practice, so if it
> could be ignored in trace headers or at least downgraded to a warning?

Don't care.

There are lots of weird macros that don't fit the typical uses
like this not just for trace weirdness.

Maybe:

---

 scripts/checkpatch.pl | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
index a2fc7d5561267..126a8d4c07ce0 100755
--- a/scripts/checkpatch.pl
+++ b/scripts/checkpatch.pl
@@ -3876,6 +3876,9 @@ sub process {
 			}
 		}
 
+# trace include files use a completely different grammar/style
+		next if ($realfile =~ m{/trace/.*\.h$} || $realfile =~ m{trace\.h$});
+
 # check for assignments on the start of a line
 		if ($sline =~ /^\+\s+($Assignment)[^=]/) {
 			my $operator = $1;