scanf '%n' fix.
C. Scott Ananian
cscott at cscott.net
Thu Mar 10 00:33:57 CST 2005
Changelog:
- Allow scanf to continue parsing format string after a '%n'.
(Previously, 'st' wasn't being set, so we'd stop after %n even if
we weren't done with the format string.) Also set 'suppress'
so that the result value returned after a %n is correct.
- Add tests for scanf to check proper handling of '%s %n%s'.
This fixes a bug I'm seeing in the Windows version of 'sd', a program for
square dance callers ( http://www.lynette.org/sd/ ).
--scott
( http://cscott.net/ )
Index: dlls/msvcrt/scanf.h
===================================================================
RCS file: /home/wine/wine/dlls/msvcrt/scanf.h,v
retrieving revision 1.19
diff -u -p -r1.19 scanf.h
--- dlls/msvcrt/scanf.h 28 Jan 2005 11:27:26 -0000 1.19
+++ dlls/msvcrt/scanf.h 10 Mar 2005 06:28:04 -0000
@@ -427,6 +427,19 @@ _FUNCTION_ {
int*n = va_arg(ap, int*);
*n = consumed - (nch!=_EOF_);
}
+ /* This is an odd one: according to the standard,
+ * "Execution of a %n directive does not increment the
+ * assignment count returned at the completion of
+ * execution" even if it wasn't suppressed with the
+ * '*' flag. The Corrigendum to the standard seems
+ * to contradict this (comment out the assignment to
+ * suppress below if you want to implement these
+ * alternate semantics) but the windows program I'm
+ * looking at expects the behavior I've coded here
+ * (which happens to be what glibc does as well).
+ */
+ suppress = 1;
+ st = 1;
}
break;
case '[': {
Index: dlls/msvcrt/tests/scanf.c
===================================================================
RCS file: /home/wine/wine/dlls/msvcrt/tests/scanf.c,v
retrieving revision 1.16
diff -u -p -r1.16 scanf.c
--- dlls/msvcrt/tests/scanf.c 28 Jan 2005 11:27:26 -0000 1.16
+++ dlls/msvcrt/tests/scanf.c 10 Mar 2005 06:28:04 -0000
@@ -117,6 +117,14 @@ static void test_sscanf( void )
ok(hour == 18, "Field 1 incorrect: %d\n", hour);
ok(min == 59, "Field 2 incorrect: %d\n", min);
ok(c == 0x55, "Field 3 incorrect: 0x%02x\n", c);
+
+ /* Check %n (also whitespace in format strings and %s) */
+ buffer[0]=0; buffer1[0]=0;
+ ret = sscanf("abc def", "%s %n%s", buffer, &number_so_far, buffer1);
+ ok(strcmp(buffer, "abc")==0, "First %%s read incorrectly: %s\n", buffer);
+ ok(strcmp(buffer1,"def")==0, "Second %%s read incorrectly: %s\n", buffer1);
+ ok(number_so_far==6, "%%n yielded wrong result: %d\n", number_so_far);
+ ok(ret == 2, "%%n shouldn't count as a conversion: %d\n", ret);
}
START_TEST(scanf)
More information about the wine-patches
mailing list