This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH] Fix trace file fails on powerpc64
- From: Yao Qi <yao at codesourcery dot com>
- To: Ulrich Weigand <uweigand at de dot ibm dot com>
- Cc: <gdb-patches at sourceware dot org>
- Date: Mon, 27 Oct 2014 20:11:03 +0800
- Subject: Re: [PATCH] Fix trace file fails on powerpc64
- Authentication-results: sourceware.org; auth=none
- References: <201410271107 dot s9RB7xto005844 at d06av02 dot portsmouth dot uk dot ibm dot com>
Ulrich Weigand <uweigand@de.ibm.com> writes:
> This will break powerpc64le, which uses the ELFv2 ABI without function
> descriptors. The correct test should be something like
>
> #elif defined __powerpc64__ && _CALL_ELF != 2
OK, fixed.
>
> OK with that change.
Patch below is pushed in.
--
Yao (éå)
Subject: [PATCH] Fix trace file fails on powerpc64
I see the following fails on powerpc64-linux,
(gdb) target tfile tfile-basic.tf^M
warning: Uploaded tracepoint 1 has no source location, using raw address^M
Tracepoint 1 at 0x10012358^M
Created tracepoint 1 for target's tracepoint 1 at 0x10012358.^M
(gdb) PASS: gdb.trace/tfile.exp: target tfile tfile-basic.tf
info trace^M
Num Type Disp Enb Address What^M
1 tracepoint keep y 0x0000000010012358 <write_basic_trace_file>^M
installed on target^M
(gdb) FAIL: gdb.trace/tfile.exp: info tracepoints on trace file
-target-select tfile tfile-basic.tf^M
=thread-group-started,id="i1",pid="1"^M
=thread-created,id="1",group-id="i1"^M
&"warning: Uploaded tracepoint 1 has no source location, using raw address\n"^M
=breakpoint-created,bkpt={number="1",type="tracepoint",disp="keep",enabled="y",
addr="0x0000000010012358",at="<write_basic_trace_file>",thread-groups=["i1"],
times="0",installed="y",original-location="*0x10012358"}^M
~"Created tracepoint 1 for target's tracepoint 1 at 0x10012358.\n"^M
^connected^M
(gdb) ^M
FAIL: gdb.trace/mi-traceframe-changed.exp: tfile: select trace file
These fails are caused by writing function descriptor address into trace
file instead of function address. This patch is to teach tfile.c to
write function address on powerpc64 target. With this patch applied,
fails in tfile.exp and mi-traceframe-changed.exp are fixed. Is it
OK?
gdb/testsuite:
2014-10-27 Yao Qi <yao@codesourcery.com>
* gdb.trace/tfile.c (adjust_function_address)
[__powerpc64__ && _CALL_ELF != 2]: Get function address from
function descriptor.
diff --git a/gdb/testsuite/ChangeLog b/gdb/testsuite/ChangeLog
index 036da6d..2eea791 100644
--- a/gdb/testsuite/ChangeLog
+++ b/gdb/testsuite/ChangeLog
@@ -1,3 +1,9 @@
+2014-10-27 Yao Qi <yao@codesourcery.com>
+
+ * gdb.trace/tfile.c (adjust_function_address)
+ [__powerpc64__ && _CALL_ELF != 2]: Get function address from
+ function descriptor.
+
2014-10-24 Don Breazeal <donb@codesourcery.com>
* gdb.base/foll-fork.exp (test_follow_fork,
diff --git a/gdb/testsuite/gdb.trace/tfile.c b/gdb/testsuite/gdb.trace/tfile.c
index e69240a..b673211 100644
--- a/gdb/testsuite/gdb.trace/tfile.c
+++ b/gdb/testsuite/gdb.trace/tfile.c
@@ -130,6 +130,9 @@ adjust_function_address (uintptr_t func_addr)
/* Although Thumb functions are two-byte aligned, function
pointers have the Thumb bit set. Clear it. */
return func_addr & ~1;
+#elif defined __powerpc64__ && _CALL_ELF != 2
+ /* Get function address from function descriptor. */
+ return *(uintptr_t *) func_addr;
#else
return func_addr;
#endif