[Patches] [patch] Fix for debug/tst-backtrace5 failure

[Tom de Vries <Tom_deVries@xxxxxxxxxx>]

Joseph,

I'm reposting the tst-backtrace5.c part of
http://www.eglibc.org/archives/patches/msg01077.html.

The patch fixes a problem with tst-backtrace5.c where the test fails because it
doesn't take VDSO into account.

I've reproduced the problem on an i686-pc-linux-gnu host with
i686-unknown-linux-gnu target:
...
Obtained backtrace with 7 functions
Function 0: /build/debug/tst-backtrace5(handle_signal+0x1a) [0x804930a]
Function 1: linux-gate.so.1(__kernel_sigreturn+0) [0x55577400]
Function 2: linux-gate.so.1(__kernel_vsyscall+0x10) [0x55577430]
Function 3: /build/libc.so.6(__read+0x23) [0x55659c43]
Function 4: /build/debug/tst-backtrace5(fn+0xb8) [0x80496f8]
Function 5: /build/debug/tst-backtrace5(fn+0x1e) [0x804965e]
Function 6: /build/debug/tst-backtrace5(fn+0x1e) [0x804965e]
Failure on line 87
...

I've retested the patch, and checked that it fixes the tst-backtrace5.c failure.

OK for trunk?

Thanks,
- Tom

2012-02-20  Paul Pluzhnikov  <ppluzhnikov@xxxxxxxxxx>

	* debug/tst-backtrace5.c (handle_signal): Adjust for VDSO.

Index: debug/tst-backtrace5.c
===================================================================
--- debug/tst-backtrace5.c	(revision 17223)
+++ debug/tst-backtrace5.c	(working copy)
@@ -82,12 +82,17 @@
       return;
     }
   /* Do not check name for signal trampoline.  */
-  if (strstr (symbols[2], "read") == NULL)
+  i = 2;
+  if (strstr (symbols[i++], "read") == NULL)
     {
-      FAIL ();
-      return;
+      /* Perhaps symbols[2] is __kernel_vsyscall?  */
+      if (strstr (symbols[i++], "read") == NULL)
+	{
+	  FAIL ();
+	  return;
+	}
     }
-  for (i = 3; i < n - 1; i++)
+  for (; i < n - 1; i++)
     if (strstr (symbols[i], "fn") == NULL)
       {
 	FAIL ();

_______________________________________________
Patches mailing list
Patches@xxxxxxxxxx
http://eglibc.org/cgi-bin/mailman/listinfo/patches