From mboxrd@z Thu Jan  1 00:00:00 1970
From: sebald.ziegler.cocci@ikolus.de (sebald.ziegler.cocci at ikolus.de)
Date: Tue, 07 Mar 2017 17:12:30 +0100
Subject: [Cocci] Problems: Adding Include directives,
	Adding ({}) construct
In-Reply-To: <alpine.DEB.2.20.1703020816190.2094@hadrien>
References: <2728810.xJvpTgCk22@detux>
 <alpine.DEB.2.20.1703020816190.2094@hadrien>
Message-ID: <1767658.UB9L1PBLAS@detux>
To: cocci@systeme.lip6.fr
List-Id: cocci@systeme.lip6.fr

On Thursday, March 2, 2017 8:21:40 AM CET Julia Lawall wrote:
> On Mon, 27 Feb 2017, sebald.ziegler.cocci at ikolus.de wrote:
> > Hi,
> > 
> > I am trying to use coccinelle to add an include directive before the other
> > includes and one after the includes:
> > 
> > @PreInclude@
> > @@
> > + #include "pre.h"
> > #include ...
> > 
> > 
> > @PostInclude@
> > @@
> > #include ...
> > + #include "post.h"
> > 
> > this works well if the source file (all includes are always at the top of
> > the file) contains more than one include directive. However, if there is
> > only one include directive -> only the PreInclude rule is executed.
> 
> Sorry not to have responded sooner.
> 
> This seems quite strange.  Each rule sees the result of the previous one,
> so by the time it reaches the post include rule, there would be two
> includes in the program.  I will check on it.
> 
> > So for this program:
> > #include <stdio.h>
> > 
> > int main() {
> > 
> >     return 0;
> > 
> > }
> > 
> > This is the result:
> > #include "pre.h"
> > #include <stdio.h>
> > 
> > int main() {
> > 
> >     return 0;
> > 
> > }
> > 
> > So it seems that only one rule can match some code line/code pattern. Am I
> > mistaken?
> 
> This is not correct.  Coccinelle works on the first rule, then provides
> the resulting code to the second rule, and so on.
> 
> > What could be an approach that works for both the single include
> > directive and multiple include directives case?
> > 
> > 
> > Another Question:
> > With the following rule I want to add a macro definition just above the
> > main function:
> > @macro_square@
> > @@
> > + #define SQUARE(x) ({printf("SQUARE(x)\n"); x*x; })
> > int main(int argc, char ** argv) {
> > ...
> > }
> > 
> > If I execute spatch on the program above this error is reported
> > (coccinelle
> > version: 1.0.6):
> > 
> > plus: parse error:
> >   File "/tmp/testing/test.c-brackets.cocci", line 3, column 21, charpos =
> >   39
> >   around = '{',
> >   whole content = + #define SQUARE(x) ({printf("SQUARE(x)\n"); x*x; })
> > 
> > Is this kind of code not supported or did I make a mistake somewhere?
> 
> It is quite possible that ({ }) is not supported in the pattern matching
> language.  There is a hackish solution:
> 
> @script:python r@
> str;
> @@
> 
> coccinelle.str = "({printf("SQUARE(x)\n"); x*x; })"
> 
> @macro_square@
> idntifier r.str;
> @@
> + #define SQUARE(x) str
> int main(int argc, char ** argv) {
> ...
> }
> 
> Coccinelle won't actually check that what you have put into the variable
> str is an identifier.
> 
> julia

Thank you very much for your answer!
I used the workaround you described and can now generate those macros without 
errors. However, there is a weird thing happening: Before every second " a 
line break is added:

When I use this coccinelle script:
@script:python macros@
SQUARE_str;
@@
coccinelle.SQUARE_str = '({printf("SQUARE\n"); printf("SQUARE\n"); x*x; })'

@macro_wrapper_int_SQUARE_int@
identifier macros.SQUARE_str;
@@
+#define SQUARE(x) SQUARE_str
int main(int argc, char ** argv) {
...
}

Then this is generated as diff:
[...]
+#define SQUARE(x) ({printf("SQUARE
+"); printf("SQUARE
+"); x*x; })
[...]

Is there a simple solution for this?

Thanks,
Sebald