----- "Caspar Zhang" <[email protected]> wrote:
> ----- "Garrett Cooper" <[email protected]> wrote:
>
> > On Tue, Nov 9, 2010 at 7:32 PM, Caspar Zhang <[email protected]>
> > wrote:
> > > Hi all,
> > >
> > > I met a strange error (only) in RHEL4, syscalls/sysconf
> > > failed the last test case because of the variable `actual'
> > > was not evaluated by `errno':
> > >
> > > sysconf01 56 TFAIL : sysconf correctly failed, but expected
> > errno (22) != actual (0)
> > >
> > > I remov the var `actual' and use `errno' directly, seems it
> > > works. Re-test under RHEL4/5/6 and all of them pass the test.
> > >
> > > I have no idea why errno value couldn't be assigned to
> > > `actual', either not sure whether my patch works on other
> > > non-Red Hat distos. Please have a look at this patch if
> > > you want, thanks.
> > >
> > > Signed-off-by: Caspar Zhang <[email protected]>
> > >
> > > ---
> > > testcases/kernel/syscalls/sysconf/sysconf01.c | 9 ++++-----
> > > 1 files changed, 4 insertions(+), 5 deletions(-)
> > >
> > > diff --git a/testcases/kernel/syscalls/sysconf/sysconf01.c
> > b/testcases/kernel/syscalls/sysconf/sysconf01.c
> > > index 783c6f5..42246f0 100644
> > > --- a/testcases/kernel/syscalls/sysconf/sysconf01.c
> > > +++ b/testcases/kernel/syscalls/sysconf/sysconf01.c
> > > @@ -153,19 +153,18 @@ int main()
> > >
> > > /* 56 */
> > > {
> > > - int retval, actual;
> > > + int retval;
> > > errno = 0;
> > > retval = sysconf(INVAL_FLAG);
> > > - actual = errno;
> > > if (retval != -1) {
> > > tst_resm(TFAIL,
> > > "sysconf succeeded for invalid
> flag
> > (%i), "
> > > " retval=%d errno=%d: %s",
> > > - INVAL_FLAG, retval, actual,
> > strerror(actual));
> > > - } else if (actual != EINVAL) {
> > > + INVAL_FLAG, retval, errno,
> > strerror(errno));
> > > + } else if (errno != EINVAL) {
> > > tst_resm(TFAIL,
> > > "sysconf correctly failed, but
> > expected "
> > > - "errno (%i) != actual (%i)",
> > EINVAL, actual);
> > > + "errno (%i) != actual (%i)",
> > EINVAL, errno);
> > > } else
> > > tst_resm(TPASS, "The invalid sysconf key
> was
> > trapped "
> > > "appropriately");
> >
> > Hmmm... well, for starters the definitions aren't the same:
> >
> > extern int * __error();
> > #define errno (* __error())
> >
> > I wonder if that has to do something with it... try out the
> following
> > program:
> >
> > #define _GNU_SOURCE 1
> > #include <errno.h>
> > #include <stdio.h>
> >
> > int
> > main(void)
> > {
> > open(NULL, 0);
> > int actual = errno;
> > printf("%i %d %i %d\n", actual, actual, errno, errno);
> > return (0);
> > }
>
> I've tried a similar simple program before in RHEL4, it worked
> out that actual == errno. But when turned to the program in
> this testcase, it always gave the error result.
Sorry I made a mistake.
My reproducer is:
#define _GNU_SOURCE 1
#include <stdio.h>
#include <errno.h>
#include <unistd.h>
int main()
{
int retval;
errno = 0;
retval = sysconf(-1);
printf("errno=%d\n", errno);
return 0;
}
This failure was caused by -O2 option. Nothing to do
with `actual' variable.
Here is the gas diff of with/without -O2 option:
--- with-O2.s 2010-11-10 02:01:45.000000000 -0500
+++ without-O2.s 2010-11-10 02:01:53.000000000 -0500
@@ -1,9 +1,8 @@
.file "main.c"
- .section .rodata.str1.1,"aMS",@progbits,1
+ .section .rodata
.LC0:
.string "errno=%d\n"
.text
- .p2align 2,,3
.globl main
.type main, @function
main:
@@ -11,19 +10,28 @@
movl %esp, %ebp
subl $8, %esp
andl $-16, %esp
- subl $16, %esp
+ movl $0, %eax
+ addl $15, %eax
+ addl $15, %eax
+ shrl $4, %eax
+ sall $4, %eax
+ subl %eax, %esp
call __errno_location
- subl $12, %esp
movl $0, (%eax)
+ subl $12, %esp
pushl $-1
call sysconf
+ addl $16, %esp
+ movl %eax, -4(%ebp)
+ subl $8, %esp
+ subl $8, %esp
call __errno_location
- popl %eax
- popl %edx
- pushl $0
+ addl $8, %esp
+ pushl (%eax)
pushl $.LC0
call printf
- xorl %eax, %eax
+ addl $16, %esp
+ movl $0, %eax
leave
ret
.size main, .-main
>
> >
> > It spit out this on FreeBSD at least:
> >
> > $ gcc -Wall -o test_errno_assignment test_errno_assignment.c
> > $ ./test_errno_assignment
> > 14 14 14 14
> >
> > HTH,
> > -Garrett
>
> --
> Quality Assurance Associate (Kernel) in
> Red Hat Software (Beijing) Co., R&D Branch
>
> TEL: +86-10-62608150
> WEB: http://www.redhat.com/
--
Quality Assurance Associate (Kernel) in
Red Hat Software (Beijing) Co., R&D Branch
TEL: +86-10-62608150
WEB: http://www.redhat.com/
------------------------------------------------------------------------------
The Next 800 Companies to Lead America's Growth: New Video Whitepaper
David G. Thomson, author of the best-selling book "Blueprint to a
Billion" shares his insights and actions to help propel your
business during the next growth cycle. Listen Now!
http://p.sf.net/sfu/SAP-dev2dev
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list
