Norihiro Tanaka wrote:
I can't find any reasons that `accept' variable should be uninitialized.
Two reasons. First, if we add initialization to the source, that might
cause some compilers to generate less-efficient code. More important,
adding initialization might cause some human readers of the code to
become confused, and to think that the initialization is necessary.
If you're using -Wall, I suggest configuring with './configure
--enable-gcc-warnings', as that should avoid the problem.
Come to think of it, perhaps we should get rid of the '#ifdef lint' code
here, as in the attached patch. These days GCC is smart enough to
figure this stuff out without that code, if one uses
--enable-gcc-warnings. This works for me with GCC 4.9.0 and with GCC
4.8.2. --enable-gcc-warnings is intended for relatively-recent GCC
versions; we needn't worry about supporting old versions.
diff --git a/src/kwset.c b/src/kwset.c
index f86ee03..617db2f 100644
--- a/src/kwset.c
+++ b/src/kwset.c
@@ -630,10 +630,6 @@ cwexec (kwset_t kwset, char const *text, size_t len,
struct kwsmatch *kwsmatch)
struct tree const *tree;
char const *trans;
-#ifdef lint
- accept = NULL;
-#endif
-
/* Initialize register copies and look for easy ways out. */
if (len < kwset->mind)
return -1;
