Re: [cocci] Question regarding Coccinelle

On Tue, 26 May 2026, Pierrick Philippe wrote:

>
> Hi Julia,
>
> First of all thanks for your answer.
>
> On 5/26/26 10:23, Julia Lawall wrote:
>
> On Fri, 22 May 2026, Pierrick Philippe wrote:
>
> Hi everyone,
>
> I am currently trying to assess if Coccinelle could be used to match my
> use case.
> What I am trying to do is to modify the declaration of a specific
> variable by adding an attribute to it, if and only if it is passed to a
> specific function (memcmp in this case).
> i am using Coccinelle version 1.3.0 (through nix-shell:
> https://search.nixos.org/packages?channel=25.11&query=coccinelle#show=coccinelle
> <https://search.nixos.org/packages?channel=25.11&query=coccinelle#show=coccinelle> ).
>
> My input test program is the following:
>
> ```c
> #include <string.h>
> int main(void) {
>     int a = 42; // This is the variable of interest
>     int b = 43;
>     int z = memcmp(&a, &b, sizeof(int));
>     return z;
> }
> ```
>
> Which should be modify to:
>
> ```c
> #include <string.h>
> int main(void) {
>     int __attribute__((myattr)) a = 42; // This is the variable of interest
>     int b = 43;
>     int z = memcmp(&a, &b, sizeof(int));
>     return z;
> }
> ```
>
> In order to do this, I played around with Coccinelle, but I keep having
> weird behavior from spatch (at least from where I stand).
> Here are two attempt to cocci rules:
>
> ```cocci
> @first@
> type T;
> T VAR;
> expression E;
> @@
> -VAR = E;
> +__attribute__((myattr)) VAR = E; // Parse error here
> ...
> memcmp(&VAR, ...);
> ```
>
> But for some reason, I got the following parsing error for the first
> attempt and I couldn't wrap my head around it:
>
> ```
> plus: parse error: 
>   File "test.cocci", line 7, column 25, charpos = 76
>   around = 'VAR',
>   whole content = +__attribute__((myattr)) VAR = E;
> ```
>
> Attributes are only allowed in variable declarations, which declare
> identifiers, not arbitrary expressions.
>
> In int x = 3;
>
> x somehow doesn't yet have type int.  It's just a name.
>
> Also, Coccinelle requires you specify a complete term, not just the
> subpart of the term that is convenient for you.
>
> I think I understand, also I do have a question regarding the `declaration` keyword of the grammar.
> Is there any way to access inner information of the declaration? Such as the type or the presence of an initializer or not?
> For example in a python or a ocaml script?


Not sure what you mean.  There is a metavariable of type declaration taht
will take the whole declaration at once.  If you also want the pieces you
can write:

(
d
&
T x = ...; // matches the case with and without an initializer
)

>
> My second attempt is successfully parsed, but leave the input program
> untouched:
>
> ```cocci
> @second@
> type T;
> identifier VAR;
> expression E;
> @@
> -T VAR = E;
> +T __attribute__((myattr)) VAR = E;
> ...
> memcmp(&VAR, ...);
>
> This again fails because you have only specified part of the memcmp
> statement.  If you want to put a complete statement, you have to put the
> whole thing.  I think you can also just put a memcmp call, without the
> semicolon.
>
> It worked with the whole `memcmp` call statement indeed, I probably misunderstood the `...` operator then.

... on the line before memcmp only jumps over statements in the control
flow graph, not arbitrary tokens at the beginning of a statement.

julia

> Pierrick
>
> julia
>
> ```
>
> This is probably an easy use case, but I am completely new to Coccinelle
> and I honestly can't understand what is going wrong here.
> To be honest, I also do not understand the difference between the two
> rules, and why one is successfully parsed and not the other one.
> Although this is probably due to a lack of experience with Coccinelle
> and of its internals.
>
> Cheers,
>
> Pierrick
>
> >
>
>
>