Re: [PATCH 9/8] lib/sort: turn off self-test

[Matt Mackall <mpm@selenic.com>]

On Mon, Jan 31, 2005 at 03:57:42AM -0800, Paul Jackson wrote:
> How about just removing the self test, not "#if 0"'ing it out.
> 
> Better to keep the kernel source code clean of development
> scaffolding.
> 
> Though your patch 1/8 hasn't arrived in my email inbox yet,
> so I don't actually know what 'self test' code it is that
> I am speaking of ;).

It's a nice self-contained unit test. It's here because I ran into a
strange regparm-related bug when developing the code in userspace and
I wanted to be sure that it was easy to diagnose in the field if a
similar bug appeared in the future. I actually think that more code
ought to have such tests, so long as they don't obscure the code in
question.

Here it is for purposes of discussion:

+#if 1
+/* a simple boot-time regression test */
+
+int cmpint(const void *a, const void *b)
+{
+	return *(int *)a - *(int *)b;
+}
+
+static int sort_test(void)
+{
+	int *a, i, r = 0;
+
+	a = kmalloc(1000 * sizeof(int), GFP_KERNEL);
+	BUG_ON(!a);
+
+	printk("testing sort()\n");
+
+	for (i = 0; i < 1000; i++) {
+		r = (r * 725861) % 6599;
+		a[i] = r;
+	}
+
+	sort(a, 1000, sizeof(int), cmpint, 0);
+
+	for (i = 0; i < 999; i++)
+		if (a[i] > a[i+1]) {
+			printk("sort() failed!\n");
+			break;
+		}
+
+	kfree(a);
+
+	return 0;
+}
+
+module_init(sort_test);
+#endif

-- 
Mathematics is the supreme nostalgia of our time.