Remove unnecessary null pointer checks

[elfring]

An extra null pointer check is not needed in functions like the following.

• nl_cli_cls_parse_ematch
• rtnl_ematch_nbyte_set_pattern
• rtnl_ematch_text_set_pattern
• xfrm_sa_free_data

[thom311]

We don't need an issue for this. It's mostly a cosmetic change anyway.

If you care enough, please open a pull request that changes such things. Patch welcome.

---

Where possible, I think it's better to avoid explicit free and use __attribute__((cleanup)) instead. For excample, instead of explicit free , it would be better to do:

diff --git i/src/lib/cls.c w/src/lib/cls.c
index 9c94b0de7f67..fe5574ffc5d1 100644
--- i/src/lib/cls.c
+++ w/src/lib/cls.c
@@ -50,16 +50,13 @@ void nl_cli_cls_parse_proto(struct rtnl_cls *cls, char *arg)
 
 struct rtnl_ematch_tree *nl_cli_cls_parse_ematch(struct rtnl_cls *cls, char *arg)
 {
+        _nl_auto_free char *errstr = NULL;
         struct rtnl_ematch_tree *tree;
-        char *errstr = NULL;
         int err;
 
         if ((err = rtnl_ematch_parse_expr(arg, &errstr, &tree)) < 0)
                  nl_cli_fatal(err, "Unable to parse ematch expression: %s",
                                      errstr);
-        
-        if (errstr)
-                 free(errstr);
 
         return tree;
 }

the benefit is that in this case the auto variable on the stack is indicated to "own" the resource it points to.

[elfring]

💭 I am curious if another software developer (besides me) can get into the mood to apply a corresponding update suggestion which could be generated by the software “Coccinelle” (also with the help of a variant from the following script).

@Remove_unnecessary_pointer_checks@
expression x;
@@
-if (\(x != 0 \| x != NULL\))
    free(x);

[elfring]

How do you think about to improve the API documentation also for a function like nlmsg_free() (which tolerates the passing of null pointers)? 🤔