Hi,
I might have read your answer wrongly, but I am fully satisfied with the fact
that:
void foo(int baz)
{
printf("foo %d",baz);
}
isn't changed to:
void bar(int baz)
{
printf("foo %d",baz);
}
but I don't like that
switch(foo) is changed to switch(bar)
Now if my c file only contained this:
-------------
void baz(int foo)
{
int bar = 5;
switch (foo) {
case 1 :
break ;
case 2:
foo(2);
printf("two");
break ;
default:
printf("baz %d", bar);
break;
}
}
-------------
then applying
-----
@@
@@
-foo
+bar
-----
would lead to:
---------
void baz(int foo)
{
int bar = 5;
- switch (foo) {
+ switch (bar) {
case 1 :
break ;
case 2:
- foo(2);
+ bar(2);
printf("two");
break ;
default:
---------
Note that the function parameter foo in baz() aren't changed. So now the baz()
function used the fixed bar = 5 insted of the parameter to baz() !!!
Insted using
---------
@@
expression E;
@@
-foo(E);
+bar(E);
---------
will keep it's hands of the switch(bar) statement, but it wouldn't handle the
update of:
Func *funclist [] = { foo, baz };
to
Func *funclist [] = { bar, baz };
Btw. how do I handle variable length expressions, where I wan't to insert the
same expression list again?
-foo(E...)
+bar(E...)
Regards,
Andreas Bach Aaen
-----Original Message-----
From: Julia Lawall [mailto:[email protected]]
Sent: 6. januar 2010 15:00
To: Aaen, Andreas.Bach
Cc: [email protected]
Subject: Re: [Cocci] Function replacement
Presumably there is a definition or a prototype for the function, either
locally or in a header file?
You could have one rule that matches a function declaration, eg:
@r@
identifier f;
@@
f(...) { ... }
and then use r.f in some other rule. Unfortunately, You can't put both a
function declaration and a function prototype in a single rule, so everything
would have to be duplicated for the prototpe case.
This should be run with -all_includes, and -I path, where path is the include
directory. If you are using the -dir option and the include directory is just
the argument_to_dir/include, then -I is not needed.
Another weak point is that coccinelle might not find the header file that
contains the prototype. A further refinement would be to look for calls, use
the name of the called function as well.
julia
On Wed, 6 Jan 2010, Aaen, Andreas.Bach wrote:
> Hi,
>
> The simple cocci rule:
>
> --------
> @@
> @@
> -foo
> +bar
> --------
>
> Will replace function calls of foo wityh function calls of bar - and as a
> bonus it will also update functionpointers to bar.
> Unfortunately also other identifiers are matched.
>
> Look at this example:
>
> ------------
> void foo(int baz)
> {
> printf("foo %d",baz);
> }
>
> void baz(int foo)
> {
> int bar = 5;
> switch (foo) {
> case 1 :
> break ;
> case 2:
> foo(2);
> printf("two");
> break ;
> default:
> printf("baz %d", bar);
> break;
> }
> }
>
> typedef void (Func)(int);
> Func *funclist [] = { foo, baz };
>
> void buh(int bah)
> {
> (funclist[0])(bah);
> (funclist[1])(bah);
> }
> ------------
>
> This will generate this patch:
> ----------
> --- switch.c 2010-01-06 14:13:39.742564000 +0100
> +++ /tmp/cocci-output-19759-1fb20a-switch.c 2010-01-06 14:15:52.000000000
> +0100
> @@ -6,11 +6,11 @@ void foo(int baz)
> void baz(int foo)
> {
> int bar = 5;
> - switch (foo) {
> + switch (bar) {
> case 1 :
> break ;
> case 2:
> - foo(2);
> + bar(2);
> printf("two");
> break ;
> default:
> @@ -20,7 +20,7 @@ void baz(int foo)
> }
>
> typedef void (Func)(int);
> -Func *funclist [] = { foo, baz };
> +Func *funclist [] = { bar, baz };
>
> void buh(int bah)
> {
> ----------
>
> switch(foo) is changed to switch(bar). Not good. can I avoid this and still
> get function pointers updated?
>
> regards,
> Andreas Bach Aaen
>
_______________________________________________
Cocci mailing list
[email protected]
http://lists.diku.dk/mailman/listinfo/cocci
(Web access from inside DIKUs LAN only)
