Skip to contentSkip to conversation

scripts/checkpatch.pl: Add warning about leading contination tests

PATCH In review Joe Perches 5 participants 8 messages Dec 5, 2009 – Dec 8, 2009 Checkpatch

lore
Share & tools
lore

Joe Perches

author maintainer
#0 lore raw

PATCH 1/1 scripts/checkpatch.pl: Add warning about leading contination tests

scripts/checkpatch.pl +5 -0
index bc4114f..c35933a 100755
@@ -2064,6 +2064,11 @@ sub process {
 			CHK("multiple assignments should be avoided\n" . $herecurr);
 		}
 
+# Check use of leading logical continuation tests
+		if ($line =~ /^.\s*(\|\||&&)/) {
+			WARN("Continuation logic should be at end of previous line\n" . $herecurr);
+		}
+
 ## # check for multiple declarations, allowing for a function declaration
 ## # continuation.
 ## 		if ($line =~ /^.\s*$Type\s+$Ident(?:\s*=[^,{]*)?\s*,\s*$Ident.*/ &&
#1 lore raw
Quoted 10 lines · Joe Perches
scripts/checkpatch.pl +0 -0
index bc4114f..c35933a 100755
@@ -2064,6 +2064,11 @@ sub process {
 			CHK("multiple assignments should be avoided\n" . $herecurr);

scripts/checkpatch.pl sub process { · line 2074

Quoted 10 lines
scripts/checkpatch.pl +5 -0
 		}
 
+# Check use of leading logical continuation tests
+		if ($line =~ /^.\s*(\|\||&&)/) {
+			WARN("Continuation logic should be at end of previous line\n" . $herecurr);
+		}
+
 ## # check for multiple declarations, allowing for a function declaration
 ## # continuation.
 ## 		if ($line =~ /^.\s*$Type\s+$Ident(?:\s*=[^,{]*)?\s*,\s*$Ident.*/ &&

Fine with me, but please add relevant info in Documentation/CodingStyle ?

#2 lore raw
Quoted 11 lines · Eric Dumazet
Quoted 10 lines · Joe Perches
scripts/checkpatch.pl +0 -0
index bc4114f..c35933a 100755
@@ -2064,6 +2064,11 @@ sub process {
 			CHK("multiple assignments should be avoided\n" . $herecurr);
 		}
 
Quoted 12 lines

scripts/checkpatch.pl sub process { · line 2074

Quoted 10 lines
scripts/checkpatch.pl +5 -0
+# Check use of leading logical continuation tests
+		if ($line =~ /^.\s*(\|\||&&)/) {
+			WARN("Continuation logic should be at end of previous line\n" . $herecurr);
+		}
+
 ## # check for multiple declarations, allowing for a function declaration
 ## # continuation.
 ## 		if ($line =~ /^.\s*$Type\s+$Ident(?:\s*=[^,{]*)?\s*,\s*$Ident.*/ &&

Fine with me, but please add relevant info in Documentation/CodingStyle ?

Not fine with me. Placing the binary logic operator at the beginning of a line can be a deliberate choice, either to make complex binary expressions more readable, or to avoid long lines. I don't see much point in banning this style, which BTW is used over 8000 times in the current kernel tree.

Signature · 1 line
Jean Delvare
author maintainer
#3 lore raw
Quoted 5 lines · Jean Delvare

Not fine with me. Placing the binary logic operator at the beginning of a line can be a deliberate choice, either to make complex binary expressions more readable, or to avoid long lines. I don't see much point in banning this style, which BTW is used over 8000 times in the current kernel tree.

Anyone that thinks that checkpatch is the last word on linux coding style and all of its pronouncements must be followed all the time is simply wrong.

It's not a ban. It's neither a command nor an edict. It's a warning. It's a notice that leading logical continuations are not the preferred style and it can be ignored at will.

I think it's rather like the long line, >80 column warning. There are a whole lot more than 8k long lines in kernel source and no one is suggesting reformatting all of them out of existence.

cheers, Joe

#4 lore raw
Quoted 12 lines · Joe Perches
Quoted 5 lines · Jean Delvare

Not fine with me. Placing the binary logic operator at the beginning of a line can be a deliberate choice, either to make complex binary expressions more readable, or to avoid long lines. I don't see much point in banning this style, which BTW is used over 8000 times in the current kernel tree.

Anyone that thinks that checkpatch is the last word on linux coding style and all of its pronouncements must be followed all the time is simply wrong.

Quoted 5 lines

It's not a ban. It's neither a command nor an edict. It's a warning. It's a notice that leading logical continuations are not the preferred style and it can be ignored at will.

I don't disagree with that. However, adding more and more checks in checkpatch.pl has its downsides. For example, I want to be able to tell people who submit patches to me: "run checkpatch.pl on your patch and solve every problem it reports before sending it to me again". If I must instead tell them: "run checkpatch.pl on your patch and fix what you want" then they might as well not fix anything, because they will not know which warnings _I_ find relevant and which I don't. Then the checkpatch.pl script becomes useless for that use case.

More generally, if checkpatch.pl starts reporting too many warnings which people consider false positives, then developers may simply stop using it. This especially holds for newcomers. If they check their patch and they get 10 errors or warnings, they'll fix them. If they get 100 they may just give up. checkpatch.pl is a wonderful tool and I don't want to lose that.

So if you are going to add checks which are icing on the cake, please disable them by default and only show them if the user explicitly asks for it by passing --strict or some such. As a matter of fact, the rule you are trying to add is not in Documentation/CodingStyle, as Eric noticed already, so it can't be that important, can it?

Quoted 5 lines

I think it's rather like the long line, >80 column warning. There are a whole lot more than 8k long lines in kernel source and no one is suggesting reformatting all of them out of existence.

Lines longer than 8_k_? I hope not ;)

Signature · 1 line
Jean Delvare
author maintainer
#5 lore raw
Quoted 7 lines · Jean Delvare

I want to be able to tell
people who submit patches to me: "run checkpatch.pl on your patch and solve every problem it reports before sending it to me again". If I must instead tell them: "run checkpatch.pl on your patch and fix what you want" then they might as well not fix anything, because they will not know which warnings _I_ find relevant and which I don't. Then the checkpatch.pl script becomes useless for that use case.

If you actually do that, you probably want them to fix _every last problem_ because the patch is either trivial or has so many broken elements that asking that contributor to fix them all is a learning experience for them.

Quoted 2 lines

So if you are going to add checks which are icing on the cake, please disable them by default and only show them if the user explicitly asks

Making the test use the CHK function rather than WARNING one seems sensible.

Quoted 7 lines
Quoted 5 lines

I think it's rather like the long line, >80 column warning. There are a whole lot more than 8k long lines in kernel source and no one is suggesting reformatting all of them out of existence.

Lines longer than 8_k_? I hope not ;)

Interpretive reading is like interpretive dance. I've used compilers like that...

cheers, Joe

#6 lore raw
In reply to Joe Perches
Quoted 8 lines · Joe Perches
scripts/checkpatch.pl +0 -0
index bc4114f..c35933a 100755
@@ -2064,6 +2064,11 @@ sub process {
 			CHK("multiple assignments should be avoided\n" . $herecurr);

scripts/checkpatch.pl sub process { · line 2074

Quoted 10 lines
scripts/checkpatch.pl +5 -0
 		}
 
+# Check use of leading logical continuation tests
+		if ($line =~ /^.\s*(\|\||&&)/) {
+			WARN("Continuation logic should be at end of previous line\n" . $herecurr);
+		}
+
 ## # check for multiple declarations, allowing for a function declaration
 ## # continuation.
 ## 		if ($line =~ /^.\s*$Type\s+$Ident(?:\s*=[^,{]*)?\s*,\s*$Ident.*/ &&

Where does this preference come from?

In

	excessivelylongcondition
	&& anotherreallylongcondition
	&& yetanotherunbelievablylongcondition
	&& yetanotherwellyougettheidea

I want to be able to keep the &&'s all justified.

Or look for well-typeset math or CS texts and try to find any that leave operators dangling on the right.

I don't really care much about this particular point, but: the checkpatch output is already getting too verbose to be useful, without adding advice that's actually the opposite of what I'd normally want to do....

--b.

#7 lore raw
Quoted 2 lines · J. Bruce Fields

Where does this preference come from?

David Miller -- in response to a patch of mine that used:
- trailing && on existing lines that already had trailing &&, and
- leading && on existing lines that already had leading &&, and
- leading && on new code.

He decided he wants "consistency", existing code be damned.

Quoted 9 lines

In

	excessivelylongcondition
	&& anotherreallylongcondition
	&& yetanotherunbelievablylongcondition
	&& yetanotherwellyougettheidea

I want to be able to keep the &&'s all justified.

Agree with you and Jean Delvare and thousands of other developers.

Quoted 3 lines

Or look for well-typeset math or CS texts and try to find any that leave operators dangling on the right.

Agreed.

Quoted 5 lines

I don't really care much about this particular point, but: the checkpatch output is already getting too verbose to be useful, without adding advice that's actually the opposite of what I'd normally want to do....

Yes, you are agreeing with a point Jean raised here, too.

Count me as opposed to this patch.

When I first looked at CodingStyle back in August, one thing that appealed to me was the laid-back simpler style -- very few, very clear rules.

I'd prefer an addition to CodingStyle clarifying that we should not argue about this minutiae.