Adding support for xxx_test.exe all

[Francois Gouget]

When running the tests on Windows you can either do 'xxx_test.exe' which
gives you a list of the available tests for this executable, or
'xxx_test.exe foo' which runs test foo, and only test foo.

So I propose the following patch which makes it possible to run
'xxx_test.exe all' which runs all the tests implemented by that
executable.

Do others find this patch acceptable? Useful?

Index: include/wine/test.h
===================================================================
RCS file: /home/wine/wine/include/wine/test.h,v
retrieving revision 1.7
diff -u -r1.7 test.h
--- include/wine/test.h	30 Oct 2002 20:36:21 -0000	1.7
+++ include/wine/test.h	30 Nov 2002 04:46:23 -0000
@@ -319,7 +319,23 @@
     if ((p = getenv( "WINETEST_REPORT_SUCCESS"))) report_success =
atoi(p);
     if (!argv[1]) usage( argv[0] );

-    return run_test(argv[1]);
+    if (strcmp(argv[1], "all") == 0)
+    {
+        const struct test* test;
+        int status=0;
+
+        for (test = winetest_testlist; test->name; test++)
+        {
+            fprintf(stderr, "Testing %s\n", test->name);
+            status+=run_test(test->name);
+            fprintf(stderr, "\n");
+        }
+        return (status <= 255) ? status : 255;
+    }
+    else
+    {
+        return run_test(argv[1]);
+    }
 }

#endif  /* WINETEST_WANT_MAIN */

[Alexandre Julliard]

Francois Gouget  writes:
> So I propose the following patch which makes it possible to run
> 'xxx_test.exe all' which runs all the tests implemented by that
> executable.
>
> Do others find this patch acceptable? Useful?

The drawback I see is that with this change the tests do not start
with a clean process state, which could cause problem for some of
them. Or we need to require that all tests cleanup properly after
them, which may be a bit painful.

[Francois Gouget]

On 30 Nov 2002, Alexandre Julliard wrote:
[...]
> > Do others find this patch acceptable? Useful?
>
> The drawback I see is that with this change the tests do not start
> with a clean process state, which could cause problem for some of
> them. Or we need to require that all tests cleanup properly after
> them, which may be a bit painful.

Tests have to cleanup behind themselves to some extent otherwise it
means they cannot be run twice in a row and that would be very bad.
But yes, they might leave static variables behind or environment
variables or whatnot. That would really not be very clean. The tests
don't work well enough on Windows that I know whether that is the case
or not.

Note that it would be pretty easy to just 'system("xxx_test.exe foo")
each individual test which would solve the above issue. It may be that
with Patrik's script for running all tests the need for an 'xxx_test.exe
all' may be lower. I'll play some more with the tests and maybe I'll
submit a patch that adds an 'all' option based on 'system'.