Skip to contentSkip to conversation

coccinelle: add style check for assignment in if

RFC In review Kris Borer 3 participants 7 messages Aug 12, 2015 Coccinelle/Semantic Patches (SmPL)

lore
Share & tools
lore

Kris Borer

author
#0 lore raw

RFC 1/1 coccinelle: add style check for assignment in if

Add a semantic patch for fixing some cases of checkpatch.pl error:

ERROR: do not use assignment in if condition

---
 scripts/coccinelle/style/assignment_in_if.cocci | 82 +++++++++++++++++++++++++
 1 file changed, 82 insertions(+)
 create mode 100644 scripts/coccinelle/style/assignment_in_if.cocci
scripts/coccinelle/style/assignment_in_if.cocci new file +82 -0
new file mode 100644
index 0000000..d9895e7
@@ -0,0 +1,82 @@
+// find checkpatch.pl errors of the type:
+//	ERROR: do not use assignment in if condition
+//
+// Confidence: Moderate
+
+
+// if ( ret = call() )
+ at if1@
+identifier i;
+expression E;
+statement S1, S2;
+@@
+
++ i = E;
+  if (
+- (i = E)
++ i
+  ) S1 else S2
+
+
+// if ( (ret = call()) < 0 )
+@if2@
+identifier i;
+expression E;
+statement S1, S2;
+binary operator b;
+@@
+
++ i = E;
+  if (
+- (i = E)
++ i
+  b ... ) S1 else S2
+
+// if ( ptr->fun && (ret = ptr->fun()) < 0 )
+@if3@
+identifier i, i2;
+expression E1, E2;
+constant c;
+binary operator b;
+@@
+
++ if( E1->i ) {
++  	i2 = E2;
++ 	if (i2 < c) {
+- if( E1->i && ((i2 = E2) b c) ) {
+  ...
+- }
++ 	}
++ }
+
+// if ( (ret = call()) < 0 && ret != 0 )
+ at if4@
+identifier i;
+expression E, E2, E3;
+statement S1, S2;
+binary operator b;
+@@
+
++ i = E;
+  if (
+- (i = E)
++ i
+  b
+  ... && E2 && E3 ) S1 else S2
+
+// if ( (ret = call()) < 0 && ret != 0 && ret != 0 )
+@if5@
+identifier i;
+expression E, E2, E3;
+statement S1, S2;
+binary operator b;
+@@
+
++ i = E;
+  if (
+- (i = E)
++ i
+  b
+  ... && E2 && E3 ) S1 else S2
+
+
Signature · 1 line
1.9.1
#1 lore raw

Thanks for the contribution. Have you checked very carefully that it doesn't eg pull XXX out of if (E && XXX)? (I don't know if it does or it doesn't, but it is a common pitfall with this issue).

thanks,
julia

Quoted 102 lines · Kris Borer

Add a semantic patch for fixing some cases of checkpatch.pl error:

ERROR: do not use assignment in if condition

---
 scripts/coccinelle/style/assignment_in_if.cocci | 82 +++++++++++++++++++++++++
 1 file changed, 82 insertions(+)
 create mode 100644 scripts/coccinelle/style/assignment_in_if.cocci
scripts/coccinelle/style/assignment_in_if.cocci new file +82 -1
new file mode 100644
index 0000000..d9895e7
@@ -0,0 +1,82 @@
+// find checkpatch.pl errors of the type:
+//	ERROR: do not use assignment in if condition
+//
+// Confidence: Moderate
+
+
+// if ( ret = call() )
+ at if1@
+identifier i;
+expression E;
+statement S1, S2;
+@@
+
++ i = E;
+  if (
+- (i = E)
++ i
+  ) S1 else S2
+
+
+// if ( (ret = call()) < 0 )
+ at if2@
+identifier i;
+expression E;
+statement S1, S2;
+binary operator b;
+@@
+
++ i = E;
+  if (
+- (i = E)
++ i
+  b ... ) S1 else S2
+
+// if ( ptr->fun && (ret = ptr->fun()) < 0 )
+ at if3@
+identifier i, i2;
+expression E1, E2;
+constant c;
+binary operator b;
+@@
+
++ if( E1->i ) {
++  	i2 = E2;
++ 	if (i2 < c) {
+- if( E1->i && ((i2 = E2) b c) ) {
+  ...
+- }
++ 	}
++ }
+
+// if ( (ret = call()) < 0 && ret != 0 )
+ at if4@
+identifier i;
+expression E, E2, E3;
+statement S1, S2;
+binary operator b;
+@@
+
++ i = E;
+  if (
+- (i = E)
++ i
+  b
+  ... && E2 && E3 ) S1 else S2
+
+// if ( (ret = call()) < 0 && ret != 0 && ret != 0 )
+ at if5@
+identifier i;
+expression E, E2, E3;
+statement S1, S2;
+binary operator b;
+@@
+
++ i = E;
+  if (
+- (i = E)
++ i
+  b
+  ... && E2 && E3 ) S1 else S2
+
+
--

1.9.1

author
#2 lore raw

RFC 1/1 coccinelle: add style check for assignment in if

Quoted 7 lines · Julia Lawall

Thanks for the contribution. Have you checked very carefully that it doesn't eg pull XXX out of if (E && XXX)? (I don't know if it does or it doesn't, but it is a common pitfall with this issue).

thanks,
julia

?I tested it and it seems okay. The details are below. Let me know if you meant something else or would like a more involved test.

?bash> cat test.c

int f() {
    return 0;
}

int main() {

    int a;

    // if1 good
    if( a = f() ) {
        a = 1;
    }

    // if1 bad
    if( a == 1 && (a = f()) ) {
        a = 1;
    }

    // if2 good
    if( (a = f()) < 0 ) {
        a = 1;
    }

    // if2 bad
    if( a == 1 && (a = f()) < 0 ) {
        a = 1;
    }

    // if4 good
    if( (a = f()) < 0 && a != 0 ) {
        a = 1;
    }
    // if4 bad
    if( a == 1 && (a = f()) < 0 && a != 0 ) {
        a = 1;
    }

    // if5 good
    if( (a = f()) < 0 && a != 0 ) {
        a = 1;
    }
    // if5 bad
    if( a == 1 && (a = f()) < 0 && a != 0 ) {
        a = 1;
    }

    return 0;
}

bash> spatch --sp-file assignment_in_if.cocci test.c
init_defs_builtins: /usr/local/lib/coccinelle/standard.h
warning: if4: metavariable E3 not used in the - or context code
HANDLING: test.c

tmp/cocci-output-11515-32ee81-test.c renamed from test.c +8 -4
diff =
@@ -7,7 +7,8 @@ int main() {
     int a;
 
     // if1 good
-    if( a = f() ) {
+    a = f();
+    if(a) {
         a = 1;
     }
 
@@ -17,7 +18,8 @@ int main() {
     }
 
     // if2 good
-    if( (a = f()) < 0 ) {
+    a = f();
+    if(a < 0 ) {
         a = 1;
     }
 
@@ -27,7 +29,8 @@ int main() {
     }
 
     // if4 good
-    if( (a = f()) < 0 && a != 0 ) {
+    a = f();
+    if(a < 0 && a != 0 ) {
         a = 1;
     }
     // if4 bad
@@ -36,7 +39,8 @@ int main() {
     }
 
     // if5 good
-    if( (a = f()) < 0 && a != 0 ) {
+    a = f();
+    if(a < 0 && a != 0 ) {
         a = 1;
     }
     // if5 bad

?
Thanks,

?Kris Borer?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://systeme.lip6.fr/pipermail/cocci/attachments/20150812/b53e6227/attachment.html>

#3 lore raw
In reply to Kris Borer
Quoted 3 lines · Kris Borer

Add a semantic patch for fixing some cases of checkpatch.pl error:

ERROR: do not use assignment in if condition

There is a gcc warning for this already.

Michal

author
#4 lore raw
Quoted 10 lines · Michal Marek
Quoted 3 lines · Kris Borer

Add a semantic patch for fixing some cases of checkpatch.pl error:

ERROR: do not use assignment in if condition

There is a gcc warning for this already.

Michal

?My intention was not to create another way to uncover problems but rather
to ?provide a tool for people to use to fix them. Let me know if I am misunderstanding the purpose of this subsystem.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://systeme.lip6.fr/pipermail/cocci/attachments/20150812/cc990076/attachment.html>

#5 lore raw
Quoted 8 lines · Kris Borer

On Wed, Aug 12, 2015 at 10:12 AM, Michal Marek <[email protected] <mailto:[email protected]>> wrote:

On 2015-08-12 15:51, Kris Borer wrote:
> Add a semantic patch for fixing some cases of checkpatch.pl <http://checkpatch.pl> error:
>
> ERROR: do not use assignment in if condition

Quoted 8 lines

There is a gcc warning for this already.

Michal

?My intention was not to create another way to uncover problems but rather to ?provide a tool for people to use to fix them. Let me know if I am misunderstanding the purpose of this subsystem.

OK, so this is fixing a style issue, and not cases of accidental assignment instead of '==' (for which there is a gcc warning and we hopefully do not have such errors in the kernel). While I'm probably ignorant and no not see how one style is better than the other, I see that some maintainers already applied your patches based on this check. So I'll merge it once Julia acks it.

P.S.: Please switch of HTML email, otherwise vger.kernel.org won't accept your messages.

Michal

#6 lore raw
Quoted 18 lines · Michal Marek
Quoted 8 lines · Kris Borer

On Wed, Aug 12, 2015 at 10:12 AM, Michal Marek <[email protected] <mailto:[email protected]>> wrote:

On 2015-08-12 15:51, Kris Borer wrote:
> Add a semantic patch for fixing some cases of checkpatch.pl <http://checkpatch.pl> error:
>
> ERROR: do not use assignment in if condition

Quoted 8 lines

There is a gcc warning for this already.

Michal

?My intention was not to create another way to uncover problems but rather to ?provide a tool for people to use to fix them. Let me know if I am misunderstanding the purpose of this subsystem.

Quoted 6 lines

OK, so this is fixing a style issue, and not cases of accidental assignment instead of '==' (for which there is a gcc warning and we hopefully do not have such errors in the kernel). While I'm probably ignorant and no not see how one style is better than the other, I see that some maintainers already applied your patches based on this check. So I'll merge it once Julia acks it.

Actually, assignments inside if tests are really annoying for Coccinelle, because there become two different control flows from the assignment to the test on the result. So I would be happy to see these go away.

I'll check the semantic patch as soon as possible.

julia