threads / patch / 43037

patch, 2 partsconfig.c: avoid duplicated global static variables

Subject: [PATCH 1/2] config.c: avoid duplicated global static variables

## tl;dr

4 messages between Aug 9, 2016 and Aug 9, 2016. Diffs are folded; open one to read it.

replies: 3people: 2as markdown or json

Johannes Sixt· Aug 9, 2016, 14:16 UTC · lore

Repeating the definition of a static variable seems to be valid in C. Nevertheless, it is bad style because it can cause confusion, definitely when it becomes necessary to change the type.

d64ec16 (git config: reorganize to use parseopt, 2009-02-21) added two static variables near the top of the file config.c without removing the definitions of the two variables that occurs later in the file.

The two variables were needed earlier in the file in the newly introduced parseopt structure. These references were removed later in d0e08d6 (config: fix parsing of "git config --get-color some.key -1", 2014-11-20).

Remove the redundant, younger, definitions near the top of the file and keep the original definitions that occur later.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
---
 Why the heck are duplicated static variables not an error in C?
 Since they aren't (as it seems), this and the next patch are just
 a matter of taste.
 builtin/config.c | 1 -
 1 file changed, 1 deletion(-)
Show changes to builtin/config.c +0 −1
diff --git a/builtin/config.c b/builtin/config.c
index a991a53..6cbf733 100644
--- a/builtin/config.c
+++ b/builtin/config.c
@@ -25,7 +25,6 @@ static char term = '\n';
 static int use_global_config, use_system_config, use_local_config;
 static struct git_config_source given_config_source;
 static int actions, types;
-static const char *get_color_slot, *get_colorbool_slot;
 static int end_null;
 static int respect_includes = -1;
 static int show_origin;

-- 
2.9.2.935.gccae72a.dirty
Johannes Sixt· Aug 9, 2016, 14:17 UTC · re: Johannes Sixt · lore

[PATCH 2/2] commit-slab.h: avoid duplicated global static variables

The gigantic define_commit_slab() macro repeats the definition of a static variable that occurs earlier in the macro text. The purpose of the repeated definition at the end of the macro is that it takes the semicolon that occurs where the macro is used.

We cannot just remove the first definition of the variable because it is referenced elsewhere in the macro text, and defining the macro later would produce undefined identifier errors. We cannot have a "forward" declaration, either. (This works only with "extern" global variables.)

The solution is to use a declaration of a struct that is already defined earlier. This language construct can serve the same purpose as the duplicated static variable definition, but without the confusion.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
---
 commit-slab.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
Show changes to commit-slab.h +3 −3
diff --git a/commit-slab.h b/commit-slab.h
index f84b449..006a50b 100644
--- a/commit-slab.h
+++ b/commit-slab.h
@@ -126,16 +126,16 @@ static MAYBE_UNUSED elemtype *slabname## _peek(struct slabname *s,	\
 	return slabname##_at_peek(s, c, 0);				\
 }									\
 									\
-static int stat_ ##slabname## realloc
+struct slabname
 
 /*
- * Note that this seemingly redundant second declaration is required
+ * Note that this redundant forward declaration is required
  * to allow a terminating semicolon, which makes instantiations look
  * like function declarations.  I.e., the expansion of
  *
  *    define_commit_slab(indegree, int);
  *
- * ends in 'static int stat_indegreerealloc;'.  This would otherwise
+ * ends in 'struct indegree;'.  This would otherwise
  * be a syntax error according (at least) to ISO C.  It's hard to
  * catch because GCC silently parses it by default.
  */

-- 
2.9.2.935.gccae72a.dirty
Johannes Sixt· Aug 9, 2016, 14:25 UTC · re: Johannes Sixt · lore

Re: [PATCH 2/2] commit-slab.h: avoid duplicated global static variables

BTW, these are all instances of duplicated global static variables that can be found in a standard Linux build.

How I found them? I waded through the error messages produced by compiling the code base as C++ code for the fun of it (basically CFLAGS='-x c++ -fpermissive').

-- Hannes
Junio C Hamano· Aug 9, 2016, 17:17 UTC · re: Johannes Sixt · lore

Re: [PATCH 2/2] commit-slab.h: avoid duplicated global static variables

Johannes Sixt <j6t@kdbg.org> writes:
Show 9 quoted lines
>   * like function declarations.  I.e., the expansion of
>   *
>   *    define_commit_slab(indegree, int);
>   *
> - * ends in 'static int stat_indegreerealloc;'.  This would otherwise
> + * ends in 'struct indegree;'.  This would otherwise
>   * be a syntax error according (at least) to ISO C.  It's hard to
>   * catch because GCC silently parses it by default.
>   */

I briefly wondered what "otherwise" and "it" in "silently parses it" refer to; it is having a hanging semicolon after the definition of the *_peek() function, i.e.

	...
        static int *indegree_peek(...)
        {
		return indegree_at_peek(s, c, 0);
	}
	;
so all is fine.

Needless to say, using "struct slabname;" is indeed a much nicer solution you found to the original issue. Thanks.

← back to recent threads