[PATCH] Handle SIGSEGV in gdb selftests

[PATCH] Handle SIGSEGV in gdb selftests

[Tom Tromey]

The gdb.gdb self-tests were timing out for me.  Looking into it, the
problem is that the version of the Boehm GC that is used by Guile on
my machine causes a SEGV during stack probing.  This unexpected stop
confuses the tests and causes repeated timeouts.

This patch adapts the two failing tests.  This makes them work for me,
and reduces the running time of gdb.gdb from 20 minutes to about 11
seconds.
---
 gdb/testsuite/gdb.gdb/python-helper.exp | 6 ++++++
 gdb/testsuite/gdb.gdb/selftest.exp      | 6 ++++++
 2 files changed, 12 insertions(+)

diff --git a/gdb/testsuite/gdb.gdb/python-helper.exp b/gdb/testsuite/gdb.gdb/python-helper.exp
index 3585ac9f0bf..fc52929489e 100644
--- a/gdb/testsuite/gdb.gdb/python-helper.exp
+++ b/gdb/testsuite/gdb.gdb/python-helper.exp
@@ -93,6 +93,12 @@ proc test_python_helper {} {
     # here.
     gdb_test_multiple "continue" "start inner gdb" {
 	-i "$inferior_spawn_id"
+	-re "received signal SIGSEGV.* in GC_.*$outer_prompt_re" {
+	    # Some versions of the GC used by Guile cause a SEGV
+	    # during stack probing.  Ignore this and carry on.
+	    send_gdb "continue\n"
+	    exp_continue
+	}
 	-re "\r\n$gdb_prompt $" {
 	    pass $gdb_test_name
 	}
diff --git a/gdb/testsuite/gdb.gdb/selftest.exp b/gdb/testsuite/gdb.gdb/selftest.exp
index ca1e7cffb4e..e2b9247f65d 100644
--- a/gdb/testsuite/gdb.gdb/selftest.exp
+++ b/gdb/testsuite/gdb.gdb/selftest.exp
@@ -75,6 +75,12 @@ proc test_with_self { } {
 	set test "xgdb is at prompt"
 	gdb_test_multiple "continue" $test {
 	    -i "$inferior_spawn_id"
+	    -re "received signal SIGSEGV.* in GC_.*$gdb_prompt" {
+		# Some versions of the GC used by Guile cause a SEGV
+		# during stack probing.  Ignore this and carry on.
+		send_gdb "continue\n"
+		exp_continue
+	    }
 	    -re "$banner" {
 		pass $test
 	    }
-- 
2.34.3

Re: [PATCH] Handle SIGSEGV in gdb selftests

[Tom de Vries]

On 12/15/22 22:14, Tom Tromey via Gdb-patches wrote:
> The gdb.gdb self-tests were timing out for me.  Looking into it, the
> problem is that the version of the Boehm GC that is used by Guile on
> my machine causes a SEGV during stack probing.  This unexpected stop
> confuses the tests and causes repeated timeouts.
> 

FWIW, that's https://sourceware.org/bugzilla/show_bug.cgi?id=29325 .

Thanks,
- Tom

> This patch adapts the two failing tests.  This makes them work for me,
> and reduces the running time of gdb.gdb from 20 minutes to about 11
> seconds.
> ---
>   gdb/testsuite/gdb.gdb/python-helper.exp | 6 ++++++
>   gdb/testsuite/gdb.gdb/selftest.exp      | 6 ++++++
>   2 files changed, 12 insertions(+)
> 
> diff --git a/gdb/testsuite/gdb.gdb/python-helper.exp b/gdb/testsuite/gdb.gdb/python-helper.exp
> index 3585ac9f0bf..fc52929489e 100644
> --- a/gdb/testsuite/gdb.gdb/python-helper.exp
> +++ b/gdb/testsuite/gdb.gdb/python-helper.exp
> @@ -93,6 +93,12 @@ proc test_python_helper {} {
>       # here.
>       gdb_test_multiple "continue" "start inner gdb" {
>   	-i "$inferior_spawn_id"
> +	-re "received signal SIGSEGV.* in GC_.*$outer_prompt_re" {
> +	    # Some versions of the GC used by Guile cause a SEGV
> +	    # during stack probing.  Ignore this and carry on.
> +	    send_gdb "continue\n"
> +	    exp_continue
> +	}
>   	-re "\r\n$gdb_prompt $" {
>   	    pass $gdb_test_name
>   	}
> diff --git a/gdb/testsuite/gdb.gdb/selftest.exp b/gdb/testsuite/gdb.gdb/selftest.exp
> index ca1e7cffb4e..e2b9247f65d 100644
> --- a/gdb/testsuite/gdb.gdb/selftest.exp
> +++ b/gdb/testsuite/gdb.gdb/selftest.exp
> @@ -75,6 +75,12 @@ proc test_with_self { } {
>   	set test "xgdb is at prompt"
>   	gdb_test_multiple "continue" $test {
>   	    -i "$inferior_spawn_id"
> +	    -re "received signal SIGSEGV.* in GC_.*$gdb_prompt" {
> +		# Some versions of the GC used by Guile cause a SEGV
> +		# during stack probing.  Ignore this and carry on.
> +		send_gdb "continue\n"
> +		exp_continue
> +	    }
>   	    -re "$banner" {
>   		pass $test
>   	    }

Re: [PATCH] Handle SIGSEGV in gdb selftests

[Tom Tromey]

>>>>> "Tom" == Tom de Vries <tdevries@suse.de> writes:

Tom> On 12/15/22 22:14, Tom Tromey via Gdb-patches wrote:
>> The gdb.gdb self-tests were timing out for me.  Looking into it, the
>> problem is that the version of the Boehm GC that is used by Guile on
>> my machine causes a SEGV during stack probing.  This unexpected stop
>> confuses the tests and causes repeated timeouts.
>> 

Tom> FWIW, that's https://sourceware.org/bugzilla/show_bug.cgi?id=29325 .

Thanks, I updated the commit message.

Tom

Re: [PATCH] Handle SIGSEGV in gdb selftests

[tdevries]

On 2022-12-16 16:11, Tom Tromey via Gdb-patches wrote:
>>>>>> "Tom" == Tom de Vries <tdevries@suse.de> writes:
> 
> Tom> On 12/15/22 22:14, Tom Tromey via Gdb-patches wrote:
>>> The gdb.gdb self-tests were timing out for me.  Looking into it, the
>>> problem is that the version of the Boehm GC that is used by Guile on
>>> my machine causes a SEGV during stack probing.  This unexpected stop
>>> confuses the tests and causes repeated timeouts.
>>> 
> 
> Tom> FWIW, that's https://sourceware.org/bugzilla/show_bug.cgi?id=29325 
> .
> 
> Thanks, I updated the commit message.

I ran into this again, tried out the patch and checked that it fixes the 
problem.

LGTM.

Thanks,
- Tom

Re: [PATCH] Handle SIGSEGV in gdb selftests

[Simon Marchi]

On 12/26/22 03:02, tdevries via Gdb-patches wrote:
> On 2022-12-16 16:11, Tom Tromey via Gdb-patches wrote:
>>>>>>> "Tom" == Tom de Vries <tdevries@suse.de> writes:
>>
>> Tom> On 12/15/22 22:14, Tom Tromey via Gdb-patches wrote:
>>>> The gdb.gdb self-tests were timing out for me.  Looking into it, the
>>>> problem is that the version of the Boehm GC that is used by Guile on
>>>> my machine causes a SEGV during stack probing.  This unexpected stop
>>>> confuses the tests and causes repeated timeouts.
>>>>
>>
>> Tom> FWIW, that's https://sourceware.org/bugzilla/show_bug.cgi?id=29325 .
>>
>> Thanks, I updated the commit message.
> 
> I ran into this again, tried out the patch and checked that it fixes the problem.
> 
> LGTM.
> 
> Thanks,
> - Tom

Hi both Toms,

I still see the issue on Ubuntu 22.04, when testing with the
native-extended-gdbserver board:

...
Thread 1 "xgdb" received signal SIGSEGV, Segmentation fault.
0x00007ffff6d828f2 in GC_find_limit_with_bound () from /usr/lib/x86_64-linux-gnu/libgc.so.1
(gdb) FAIL: gdb.gdb/selftest.exp: xgdb is at prompt

The modified code looks like this:


        set test "xgdb is at prompt"
        gdb_test_multiple "continue" $test {
            -i "$inferior_spawn_id"
            -re "received signal SIGSEGV.* in GC_.*$gdb_prompt" {
                # Some versions of the GC used by Guile cause a SEGV
                # during stack probing.  Ignore this and carry on.
                send_gdb "continue\n"
                exp_continue
            }
            -re "$banner" {
                pass $test
            }
        }

Shouldn't the SIGSEGV -re be before the `-i $inferior_spawn_id`?
Otherwise we expect the "received signal SIGSEGV" message to come from
the debugged GDB, when it's the top GDB that will print it (like when
debugging any other test program).

I think we don't see the problem when testing with native because
the top and bottom GDBs share the same spawn id.

Simon

Re: [PATCH] Handle SIGSEGV in gdb selftests

[Simon Marchi]

On 1/13/23 20:11, Simon Marchi via Gdb-patches wrote:
> 
> 
> On 12/26/22 03:02, tdevries via Gdb-patches wrote:
>> On 2022-12-16 16:11, Tom Tromey via Gdb-patches wrote:
>>>>>>>> "Tom" == Tom de Vries <tdevries@suse.de> writes:
>>>
>>> Tom> On 12/15/22 22:14, Tom Tromey via Gdb-patches wrote:
>>>>> The gdb.gdb self-tests were timing out for me.  Looking into it, the
>>>>> problem is that the version of the Boehm GC that is used by Guile on
>>>>> my machine causes a SEGV during stack probing.  This unexpected stop
>>>>> confuses the tests and causes repeated timeouts.
>>>>>
>>>
>>> Tom> FWIW, that's https://sourceware.org/bugzilla/show_bug.cgi?id=29325 .
>>>
>>> Thanks, I updated the commit message.
>>
>> I ran into this again, tried out the patch and checked that it fixes the problem.
>>
>> LGTM.
>>
>> Thanks,
>> - Tom
> 
> Hi both Toms,
> 
> I still see the issue on Ubuntu 22.04, when testing with the
> native-extended-gdbserver board:
> 
> ...
> Thread 1 "xgdb" received signal SIGSEGV, Segmentation fault.
> 0x00007ffff6d828f2 in GC_find_limit_with_bound () from /usr/lib/x86_64-linux-gnu/libgc.so.1
> (gdb) FAIL: gdb.gdb/selftest.exp: xgdb is at prompt
> 
> The modified code looks like this:
> 
> 
>         set test "xgdb is at prompt"
>         gdb_test_multiple "continue" $test {
>             -i "$inferior_spawn_id"
>             -re "received signal SIGSEGV.* in GC_.*$gdb_prompt" {
>                 # Some versions of the GC used by Guile cause a SEGV
>                 # during stack probing.  Ignore this and carry on.
>                 send_gdb "continue\n"
>                 exp_continue
>             }
>             -re "$banner" {
>                 pass $test
>             }
>         }
> 
> Shouldn't the SIGSEGV -re be before the `-i $inferior_spawn_id`?
> Otherwise we expect the "received signal SIGSEGV" message to come from
> the debugged GDB, when it's the top GDB that will print it (like when
> debugging any other test program).
> 
> I think we don't see the problem when testing with native because
> the top and bottom GDBs share the same spawn id.
> 
> Simon


Here's a patch to fix it, with a commit message.


From f2d47cdd63ea9abddabdccd8097b2a6e7d1e64b1 Mon Sep 17 00:00:00 2001
From: Simon Marchi <simon.marchi@efficios.com>
Date: Fri, 13 Jan 2023 20:08:41 -0500
Subject: [PATCH] gdb/testsuite: expect SIGSEGV from top GDB spawn id

When testing with the native-extended-gdbserver, I get:

    Thread 1 "xgdb" received signal SIGSEGV, Segmentation fault.
    0x00007ffff6d828f2 in GC_find_limit_with_bound () from /usr/lib/x86_64-linux-gnu/libgc.so.1
    (gdb) FAIL: gdb.gdb/selftest.exp: xgdb is at prompt

This is because the -re that is supposed to match this SIGSEGV is after
`-i $inferior_spawn_id`.  On native, the top and bottom GDB are on the
same spawn id, so it ends up working.  But with a gdbserver board,
that's not the case.  Move the SIGSEGV -re before the `-i
$inferior_spawn_id` line, such that it matches what the top GDB outputs.

Do the same fix in gdb.gdb/python-helper.exp.

Change-Id: I3291630e218a5a3a6a47805b999ddbc9b968c927
---
 gdb/testsuite/gdb.gdb/python-helper.exp | 2 +-
 gdb/testsuite/gdb.gdb/selftest.exp      | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/gdb/testsuite/gdb.gdb/python-helper.exp b/gdb/testsuite/gdb.gdb/python-helper.exp
index c147c6bcbb03..98f03ef456f3 100644
--- a/gdb/testsuite/gdb.gdb/python-helper.exp
+++ b/gdb/testsuite/gdb.gdb/python-helper.exp
@@ -86,13 +86,13 @@ proc test_python_helper {} {
     # stop is being detected from the inner GDB, hence the use of -i
     # here.
     gdb_test_multiple "continue" "start inner gdb" {
-	-i "$inferior_spawn_id"
 	-re "received signal SIGSEGV.* in GC_.*$outer_prompt_re" {
 	    # Some versions of the GC used by Guile cause a SEGV
 	    # during stack probing.  Ignore this and carry on.
 	    send_gdb "continue\n"
 	    exp_continue
 	}
+	-i "$inferior_spawn_id"
 	-re "\r\n$gdb_prompt $" {
 	    pass $gdb_test_name
 	}
diff --git a/gdb/testsuite/gdb.gdb/selftest.exp b/gdb/testsuite/gdb.gdb/selftest.exp
index 177c29fb224d..df12602b64bb 100644
--- a/gdb/testsuite/gdb.gdb/selftest.exp
+++ b/gdb/testsuite/gdb.gdb/selftest.exp
@@ -74,13 +74,13 @@ proc test_with_self { } {
 			"$gdb_prompt $"]
 	set test "xgdb is at prompt"
 	gdb_test_multiple "continue" $test {
-	    -i "$inferior_spawn_id"
 	    -re "received signal SIGSEGV.* in GC_.*$gdb_prompt" {
 		# Some versions of the GC used by Guile cause a SEGV
 		# during stack probing.  Ignore this and carry on.
 		send_gdb "continue\n"
 		exp_continue
 	    }
+	    -i "$inferior_spawn_id"
 	    -re "$banner" {
 		pass $test
 	    }

base-commit: 325ab797da3b71e03eba184930f08ee3179a30d6
-- 
2.39.0

Re: [PATCH] Handle SIGSEGV in gdb selftests

[Tom Tromey]

>>>>> "Simon" == Simon Marchi <simark@simark.ca> writes:

>> Shouldn't the SIGSEGV -re be before the `-i $inferior_spawn_id`?
>> Otherwise we expect the "received signal SIGSEGV" message to come from
>> the debugged GDB, when it's the top GDB that will print it (like when
>> debugging any other test program).
>> 
>> I think we don't see the problem when testing with native because
>> the top and bottom GDBs share the same spawn id.

Makes sense to me.

Simon> Here's a patch to fix it, with a commit message.

Looks good to me.

Tom

Re: [PATCH] Handle SIGSEGV in gdb selftests

[Simon Marchi]

On 1/17/23 11:20, Tom Tromey wrote:
>>>>>> "Simon" == Simon Marchi <simark@simark.ca> writes:
> 
>>> Shouldn't the SIGSEGV -re be before the `-i $inferior_spawn_id`?
>>> Otherwise we expect the "received signal SIGSEGV" message to come from
>>> the debugged GDB, when it's the top GDB that will print it (like when
>>> debugging any other test program).
>>>
>>> I think we don't see the problem when testing with native because
>>> the top and bottom GDBs share the same spawn id.
> 
> Makes sense to me.
> 
> Simon> Here's a patch to fix it, with a commit message.
> 
> Looks good to me.
> 
> Tom

Thanks, pushed.

Simon