[RFA] kfail longjmp.exp tests due to bug 9270

[RFA] kfail longjmp.exp tests due to bug 9270

[Doug Evans]

Anyone mind if I mark these as kfail for amd64?

2010-03-18  Doug Evans  <dje@google.com>

	* gdb.base/longjmp.exp: Mark tests that fail on x86_64 due to
	bug 9270 as kfail.

Index: longjmp.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.base/longjmp.exp,v
retrieving revision 1.4
diff -u -p -r1.4 longjmp.exp
--- longjmp.exp	1 Jan 2010 07:32:01 -0000	1.4
+++ longjmp.exp	18 Mar 2010 22:13:19 -0000
@@ -73,6 +73,7 @@ gdb_test "break $bp_miss_step_1" \
 gdb_test "next" "longjmps\\+\\+;.*" "next over setjmp (1)"
 gdb_test "next" "longjmp \\(env, 1\\);.*" "next to longjmp (1)"
 
+setup_kfail "gdb/9270" "x86_64-*-*"
 set msg "next over longjmp(1)"
 gdb_test_multiple "next" $msg {
     -re ".*patt1.*$gdb_prompt $" {
@@ -103,6 +104,7 @@ gdb_test "break $bp_miss_step_2" \
 
 gdb_test "next" "call_longjmp.*" "next over setjmp (2)"
 
+setup_kfail "gdb/9270" "x86_64-*-*"
 set msg "next over call_longjmp (2)"
 gdb_test_multiple "next" $msg {
     -re ".*patt2.*$gdb_prompt $" {
@@ -127,4 +129,5 @@ gdb_test "break $bp_start_test_3" \
     "breakpoint at pattern 3 start"
 gdb_test "continue" "patt3.*" "continue to breakpoint at pattern 3 start"
 
+setup_kfail "gdb/9270" "x86_64-*-*"
 gdb_test "next" "longjmp caught.*" "next over patt3"

Re: [RFA] kfail longjmp.exp tests due to bug 9270

[Mark Kettenis]

> Date: Thu, 18 Mar 2010 15:16:01 -0700 (PDT)
> From: dje@google.com (Doug Evans)
> 
> Anyone mind if I mark these as kfail for amd64?

What makes you think these fail on *all* amd64 platforms?

> 2010-03-18  Doug Evans  <dje@google.com>
> 
> 	* gdb.base/longjmp.exp: Mark tests that fail on x86_64 due to
> 	bug 9270 as kfail.
> 
> Index: longjmp.exp
> ===================================================================
> RCS file: /cvs/src/src/gdb/testsuite/gdb.base/longjmp.exp,v
> retrieving revision 1.4
> diff -u -p -r1.4 longjmp.exp
> --- longjmp.exp	1 Jan 2010 07:32:01 -0000	1.4
> +++ longjmp.exp	18 Mar 2010 22:13:19 -0000
> @@ -73,6 +73,7 @@ gdb_test "break $bp_miss_step_1" \
>  gdb_test "next" "longjmps\\+\\+;.*" "next over setjmp (1)"
>  gdb_test "next" "longjmp \\(env, 1\\);.*" "next to longjmp (1)"
>  
> +setup_kfail "gdb/9270" "x86_64-*-*"
>  set msg "next over longjmp(1)"
>  gdb_test_multiple "next" $msg {
>      -re ".*patt1.*$gdb_prompt $" {
> @@ -103,6 +104,7 @@ gdb_test "break $bp_miss_step_2" \
>  
>  gdb_test "next" "call_longjmp.*" "next over setjmp (2)"
>  
> +setup_kfail "gdb/9270" "x86_64-*-*"
>  set msg "next over call_longjmp (2)"
>  gdb_test_multiple "next" $msg {
>      -re ".*patt2.*$gdb_prompt $" {
> @@ -127,4 +129,5 @@ gdb_test "break $bp_start_test_3" \
>      "breakpoint at pattern 3 start"
>  gdb_test "continue" "patt3.*" "continue to breakpoint at pattern 3 start"
>  
> +setup_kfail "gdb/9270" "x86_64-*-*"
>  gdb_test "next" "longjmp caught.*" "next over patt3"
> 

Re: [RFA] kfail longjmp.exp tests due to bug 9270

[Doug Evans]

On Thu, Mar 18, 2010 at 3:16 PM, Doug Evans <dje@google.com> wrote:
> Anyone mind if I mark these as kfail for amd64?
>
> 2010-03-18  Doug Evans  <dje@google.com>
>
>        * gdb.base/longjmp.exp: Mark tests that fail on x86_64 due to
>        bug 9270 as kfail.

Here's a better patch.  amd64-darwin/*bsd set jb_pc_offset, and 9270
is glibc-specific.
This patch just kfails linux.

2010-03-18  Doug Evans  <dje@google.com>

        * gdb.base/longjmp.exp: Mark tests that fail on x86_64-linux due to
        bug 9270 as kfail.

Index: longjmp.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.base/longjmp.exp,v
retrieving revision 1.4
diff -u -p -r1.4 longjmp.exp
--- longjmp.exp 1 Jan 2010 07:32:01 -0000       1.4
+++ longjmp.exp 18 Mar 2010 23:38:47 -0000
@@ -73,6 +73,7 @@ gdb_test "break $bp_miss_step_1" \
 gdb_test "next" "longjmps\\+\\+;.*" "next over setjmp (1)"
 gdb_test "next" "longjmp \\(env, 1\\);.*" "next to longjmp (1)"

+setup_kfail "gdb/9270" "x86_64-*-linux*"
 set msg "next over longjmp(1)"
 gdb_test_multiple "next" $msg {
     -re ".*patt1.*$gdb_prompt $" {
@@ -103,6 +104,7 @@ gdb_test "break $bp_miss_step_2" \

 gdb_test "next" "call_longjmp.*" "next over setjmp (2)"

+setup_kfail "gdb/9270" "x86_64-*-linux*"
 set msg "next over call_longjmp (2)"
 gdb_test_multiple "next" $msg {
     -re ".*patt2.*$gdb_prompt $" {
@@ -127,4 +129,5 @@ gdb_test "break $bp_start_test_3" \
     "breakpoint at pattern 3 start"
 gdb_test "continue" "patt3.*" "continue to breakpoint at pattern 3 start"

+setup_kfail "gdb/9270" "x86_64-*-linux*"
 gdb_test "next" "longjmp caught.*" "next over patt3"

Re: [RFA] kfail longjmp.exp tests due to bug 9270

[Doug Evans]

On Thu, Mar 18, 2010 at 3:33 PM, Mark Kettenis <mark.kettenis@xs4all.nl> wrote:
>> Date: Thu, 18 Mar 2010 15:16:01 -0700 (PDT)
>> From: dje@google.com (Doug Evans)
>>
>> Anyone mind if I mark these as kfail for amd64?
>
> What makes you think these fail on *all* amd64 platforms?

I don't.  Revised patch submitted.

Re: [RFA] kfail longjmp.exp tests due to bug 9270

[Joseph S. Myers]

On Thu, 18 Mar 2010, Doug Evans wrote:

> +setup_kfail "gdb/9270" "x86_64-*-linux*"

What will happen for i686-*-linux* --enable-64-bit-bfd (running tests with 
-m64)?  What about x86_64 running tests with -m32?  Both i?86-* and 
x86_64-* can cover both 32-bit and 64-bit configurations.

What you want is (IA32 or x86_64 GNU/Linux) and (64-bit multilib), I 
think.  In GCC such tests would use the lp64 effective-target, but I don't 
know what the GDB equivalent is.

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: [RFA] kfail longjmp.exp tests due to bug 9270

[Daniel Jacobowitz]

On Thu, Mar 18, 2010 at 03:16:01PM -0700, Doug Evans wrote:
> 2010-03-18  Doug Evans  <dje@google.com>
> 
> 	* gdb.base/longjmp.exp: Mark tests that fail on x86_64 due to
> 	bug 9270 as kfail.

FWIW, my experience is that these tests fail a ton of places right
now.  So centralizing the check into a function would be nice if we're
going to start making a list.

For instance why do they pass on i386 GNU/Linux?

-- 
Daniel Jacobowitz
CodeSourcery

Re: [RFA] kfail longjmp.exp tests due to bug 9270

[Tom Tromey]

>>>>> "Doug" == Doug Evans <dje@google.com> writes:

Doug> 2010-03-18  Doug Evans  <dje@google.com>
Doug>         * gdb.base/longjmp.exp: Mark tests that fail on x86_64-linux due to
Doug>         bug 9270 as kfail.

I suspect these will fail on all recent-enough glibc systems as-is.
They fail on my plain old x86 box, for instance.

Tom

Re: [RFA] kfail longjmp.exp tests due to bug 9270

[Joseph S. Myers]

On Thu, 18 Mar 2010, Tom Tromey wrote:

> >>>>> "Doug" == Doug Evans <dje@google.com> writes:
> 
> Doug> 2010-03-18  Doug Evans  <dje@google.com>
> Doug>         * gdb.base/longjmp.exp: Mark tests that fail on x86_64-linux due to
> Doug>         bug 9270 as kfail.
> 
> I suspect these will fail on all recent-enough glibc systems as-is.
> They fail on my plain old x86 box, for instance.

If so, XFAILing for both 32-bit and 64-bit would be a simple way to avoid 
needing to solve my effective-target question.

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: [RFA] kfail longjmp.exp tests due to bug 9270

[Doug Evans]

On Thu, Mar 18, 2010 at 5:58 PM, Joseph S. Myers
<joseph@codesourcery.com> wrote:
> On Thu, 18 Mar 2010, Doug Evans wrote:
>
>> +setup_kfail "gdb/9270" "x86_64-*-linux*"
>
> What will happen for i686-*-linux* --enable-64-bit-bfd (running tests with
> -m64)?  What about x86_64 running tests with -m32?  Both i?86-* and
> x86_64-* can cover both 32-bit and 64-bit configurations.
>
> What you want is (IA32 or x86_64 GNU/Linux) and (64-bit multilib), I
> think.  In GCC such tests would use the lp64 effective-target, but I don't
> know what the GDB equivalent is.

I run 64x32 tests regularly here.
The way I do it is to hack site.exp to set host_triplet, etc. to i686-linux.
Done this way, nothing further is needed, in so far as handling
{32,64}-cross-{64,x32}.

[I'm leaving as a separate discussion whether we need to xfail these
for i686-linux too.  It passes here, but I gather it fails with more
recent glibc's]