[PATCH 0/3] Small testsuite updates

[PATCH 0/3] Small testsuite updates

[Andrew Burgess]

While working on additional riscv support I ran into a few minor
annoyances in the testsuite, addressed in these patches.  Only the
second patch is really RiscV specific, one and three are pretty
generic.

---

Andrew Burgess (3):
  gdb/testsuite: Fix broken regexp in gdbstub case
  gdb/testsuite: Filter out some registers for riscv
  gdb/testsuite: Handle targets with lots of registers

 gdb/testsuite/ChangeLog              | 15 +++++++++++++++
 gdb/testsuite/gdb.base/callfuncs.exp | 10 ++++++++++
 gdb/testsuite/gdb.base/maint.exp     | 18 ++++++++++++++++--
 gdb/testsuite/lib/mi-support.exp     |  6 +++++-
 4 files changed, 46 insertions(+), 3 deletions(-)

-- 
2.12.2

[PATCH 3/3] gdb/testsuite: Handle targets with lots of registers

[Andrew Burgess]

In gdb.base/maint.exp a test calls 'maint print registers'.  If the
target has lots of registers this may overflow expect's buffers,
causing the test to fail.

After this commit we process the output line at a time until we get back
to the GDB prompt, this should prevent buffer overrun while still
testing that the command works as required.

gdb/testsuite/ChangeLog:

	* gdb.base/maint.exp: Process output from 'maint print registers'
	line at a time.
---
 gdb/testsuite/ChangeLog          |  5 +++++
 gdb/testsuite/gdb.base/maint.exp | 18 ++++++++++++++++--
 2 files changed, 21 insertions(+), 2 deletions(-)

diff --git a/gdb/testsuite/gdb.base/maint.exp b/gdb/testsuite/gdb.base/maint.exp
index 93221085653..f73495faa5b 100644
--- a/gdb/testsuite/gdb.base/maint.exp
+++ b/gdb/testsuite/gdb.base/maint.exp
@@ -62,8 +62,22 @@ gdb_test_no_output "set height 0"
 gdb_file_cmd ${binfile}
 
 # Test for a regression where this command would internal-error if the
-# program wasn't running.
-gdb_test "maint print registers" "Name.*Nr.*Rel.*Offset.*Size.*Type.*"
+# program wasn't running.  If there's a lot of registers then this
+# might overflow expect's buffers, so process the output line at a
+# time.
+send_gdb "maint print registers\n"
+gdb_expect {
+    -re "^\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[\n\r\]+" {
+	exp_continue
+    }
+    -re ".*$gdb_prompt $" {
+	pass "maint print registers"
+    }
+    timeout {
+	fail "maint print registers (timeout)"
+    }
+}
+
 
 # Test "mt expand-symtabs" here as it's easier to verify before we
 # run the program.
-- 
2.12.2

Re: [PATCH 3/3] gdb/testsuite: Handle targets with lots of registers

[Maciej W. Rozycki]

On Mon, 9 Apr 2018, Andrew Burgess wrote:

>  # Test for a regression where this command would internal-error if the
> -# program wasn't running.
> -gdb_test "maint print registers" "Name.*Nr.*Rel.*Offset.*Size.*Type.*"
> +# program wasn't running.  If there's a lot of registers then this
> +# might overflow expect's buffers, so process the output line at a
> +# time.
> +send_gdb "maint print registers\n"
> +gdb_expect {
> +    -re "^\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[\n\r\]+" {
> +	exp_continue
> +    }

 I think this changes the meaning of the test; you want to preserve the 
heading match pattern at the very least.  Also `gdb_test' handles various 
error cases gracefully (which matters for the avoidance of excessive 
timeouts with some test boards), whereas your simple matcher does not.

 Also how many is "a lot"?  Perhaps you could take the path of least 
resistance instead and simply increase the size of the buffer, like with 
commit ff604a674771 ("gdb/testsuite: Bump up `match_max'").  This could be 
done temporarily for this test only, so as to avoid slowing down `expect' 
throughout the test suite.

  Maciej

Re: [PATCH 3/3] gdb/testsuite: Handle targets with lots of registers

[Pedro Alves]

On 04/13/2018 12:39 AM, Maciej W. Rozycki wrote:
> On Mon, 9 Apr 2018, Andrew Burgess wrote:
> 
>>  # Test for a regression where this command would internal-error if the
>> -# program wasn't running.
>> -gdb_test "maint print registers" "Name.*Nr.*Rel.*Offset.*Size.*Type.*"
>> +# program wasn't running.  If there's a lot of registers then this
>> +# might overflow expect's buffers, so process the output line at a
>> +# time.
>> +send_gdb "maint print registers\n"
>> +gdb_expect {
>> +    -re "^\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[\n\r\]+" {
>> +	exp_continue
>> +    }
> 
>  I think this changes the meaning of the test; you want to preserve the 
> heading match pattern at the very least.  Also `gdb_test' handles various 
> error cases gracefully (which matters for the avoidance of excessive 
> timeouts with some test boards), whereas your simple matcher does not.

Yeah, there's no good reason to use gdb_expect directly here, AFAICT.
You can do the same thing with gdb_test_multiple, which handles
the timeout already, as well as other error conditions, including
the internal-error the comment the test above mentions.

> 
>  Also how many is "a lot"?  Perhaps you could take the path of least 
> resistance instead and simply increase the size of the buffer, like with 
> commit ff604a674771 ("gdb/testsuite: Bump up `match_max'").  This could be 
> done temporarily for this test only, so as to avoid slowing down `expect' 
> throughout the test suite.

I think the exp_continue trick is better for getting rid of the
problem for good.  We can still use it, with gdb_test_multiple.

A few comments more:

> +    -re "^\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[0-9\]+\[^\n\r\]+\[\n\r\]+" {

Nit: \n\r instead of \r\n kind of reads like a typo to me:

$ grep -rn "\\\\r\\\\n\\\\]" | wc -l
1036
$ grep -rn "\\\\n\\\\r\\\\]" | wc -l
28

I'd suggest flipping it around to the more usual form
just to avoid causing pause when people read the regexp.

> +	exp_continue
> +    }
> +    -re ".*$gdb_prompt $" {

No need for leading .*, it's implied.

> +	pass "maint print registers"
> +    }
> +    timeout {
> +	fail "maint print registers (timeout)"
> +    }
> +}
> +

So I'd suggest something like this:

set saw_registers 0
set test "maint print registers"
gdb_test_multiple $test $test {
    -re "^\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[\r\n\]+" {
        set saw_registers 1
	exp_continue
    }
    -re "$gdb_prompt $" {
	gdb_assert $saw_registers $test
    }
}

The "saw_registers" bit ends up serving as replacement for
seeing the heading, though you can also add a pattern to 
match the heading and check it in the gdb_assert instead if
you'd like.

Thanks,
Pedro Alves

Re: [PATCH 3/3] gdb/testsuite: Handle targets with lots of registers

[Maciej W. Rozycki]

On Fri, 13 Apr 2018, Pedro Alves wrote:

> So I'd suggest something like this:
> 
> set saw_registers 0
> set test "maint print registers"
> gdb_test_multiple $test $test {
>     -re "^\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[\r\n\]+" {
>         set saw_registers 1
> 	exp_continue
>     }
>     -re "$gdb_prompt $" {
> 	gdb_assert $saw_registers $test
>     }
> }
> 
> The "saw_registers" bit ends up serving as replacement for
> seeing the heading, though you can also add a pattern to 
> match the heading and check it in the gdb_assert instead if
> you'd like.

 FWIW I think we should keep the heading check.

  Maciej

Re: [PATCH 3/3] gdb/testsuite: Handle targets with lots of registers

[Andrew Burgess]

Thanks for the review feedback.

I've updated the patch below inline with Pedro's feedback, and
included a header check as requested by Maciej.  I tested this on
x86-64 and against RiscV which is the target that was originally
causing me problems.

Thanks,
Andrew

---

[PATCH 1/2] gdb/testsuite: Handle targets with lots of registers

In gdb.base/maint.exp a test calls 'maint print registers'.  If the
target has lots of registers this may overflow expect's buffers,
causing the test to fail.

After this commit we process the output line at a time until we get back
to the GDB prompt, this should prevent buffer overrun while still
testing that the command works as required.

gdb/testsuite/ChangeLog:

	* gdb.base/maint.exp: Process output from 'maint print registers'
	line at a time.
---
 gdb/testsuite/ChangeLog          |  5 +++++
 gdb/testsuite/gdb.base/maint.exp | 25 +++++++++++++++++++++++--
 2 files changed, 28 insertions(+), 2 deletions(-)

diff --git a/gdb/testsuite/gdb.base/maint.exp b/gdb/testsuite/gdb.base/maint.exp
index 93221085653..1fe36405df3 100644
--- a/gdb/testsuite/gdb.base/maint.exp
+++ b/gdb/testsuite/gdb.base/maint.exp
@@ -62,8 +62,29 @@ gdb_test_no_output "set height 0"
 gdb_file_cmd ${binfile}
 
 # Test for a regression where this command would internal-error if the
-# program wasn't running.
-gdb_test "maint print registers" "Name.*Nr.*Rel.*Offset.*Size.*Type.*"
+# program wasn't running.  If there's a lot of registers then this
+# might overflow expect's buffers, so process the output line at a
+# time.
+set saw_registers 0
+set saw_headers 0
+set test "maint print registers"
+gdb_test_multiple $test $test {
+    -re "\[^\r\n\]+Name\[^\r\n\]+Nr\[^\r\n\]+Rel\[^\r\n\]+Offset\[^\r\n\]+Size\[^\r\n\]+Type\[^\r\n\]+\[\r\n\]+" {
+	set saw_headers 1
+	exp_continue
+    }
+    -re "^\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[\r\n\]+" {
+        set saw_registers 1
+        exp_continue
+    }
+    -re "^\\*\[0-9\]+\[^\r\n\]+\[\r\n\]+" {
+        exp_continue
+    }
+    -re "$gdb_prompt $" {
+        gdb_assert $saw_registers "$test: saw some registers"
+        gdb_assert $saw_headers "$test: saw header line"
+    }
+}
 
 # Test "mt expand-symtabs" here as it's easier to verify before we
 # run the program.
-- 
2.14.3

Re: [PATCH 3/3] gdb/testsuite: Handle targets with lots of registers

[Pedro Alves]

On 05/04/2018 01:01 PM, Andrew Burgess wrote:
> Thanks for the review feedback.
> 
> I've updated the patch below inline with Pedro's feedback, and
> included a header check as requested by Maciej.  I tested this on
> x86-64 and against RiscV which is the target that was originally
> causing me problems.

Thanks.

>  # Test for a regression where this command would internal-error if the
> -# program wasn't running.
> -gdb_test "maint print registers" "Name.*Nr.*Rel.*Offset.*Size.*Type.*"
> +# program wasn't running.  If there's a lot of registers then this
> +# might overflow expect's buffers, so process the output line at a
> +# time.
> +set saw_registers 0
> +set saw_headers 0
> +set test "maint print registers"
> +gdb_test_multiple $test $test {
> +    -re "\[^\r\n\]+Name\[^\r\n\]+Nr\[^\r\n\]+Rel\[^\r\n\]+Offset\[^\r\n\]+Size\[^\r\n\]+Type\[^\r\n\]+\[\r\n\]+" {
> +	set saw_headers 1
> +	exp_continue
> +    }
> +    -re "^\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[0-9\]+\[^\r\n\]+\[\r\n\]+" {
> +        set saw_registers 1
> +        exp_continue
> +    }
> +    -re "^\\*\[0-9\]+\[^\r\n\]+\[\r\n\]+" {
> +        exp_continue
> +    }
> +    -re "$gdb_prompt $" {
> +        gdb_assert $saw_registers "$test: saw some registers"
> +        gdb_assert $saw_headers "$test: saw header line"

We try to avoid the potential for different FAIL / PASS
names/messages.  I.e., if the test times out, or gdb crashes,
you'll get "FAIL: $test" only , while if it reaches the prompt,
you'll get two tests recorded in gdb.sum.  The idea of matching
FAIL/PASS is to make it easier for scripts to distinguish
regressions/progressions vs new failures/passes.  (Text in
trailing ()s, like " (timeout)" is considered informational, and
can/should be stripped for test-matching purposes).  If you want to
record that the register and headers were seen, better put it in gdb.log,
with send_log or verbose -log, and do:

        gdb_assert {$saw_registers && $saw_headers} $test

OK with that change.

Thanks,
Pedro Alves

[PATCH 1/3] gdb/testsuite: Fix broken regexp in gdbstub case

[Andrew Burgess]

When $use_gdb_stub is true then, when we start an MI target there's a
regexp to match GDB's startup pattern.  Unfortunately the pattern is
broken, and we're also missing a timeout case in the match list (which
would have helped point out that the regexp was broken).

gdb/testsuite/ChangeLog:

	* lib/mi-support.exp (mi_run_cmd_full): Fix regexp and add a
	timeout.
---
 gdb/testsuite/ChangeLog          | 5 +++++
 gdb/testsuite/lib/mi-support.exp | 6 +++++-
 2 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/gdb/testsuite/lib/mi-support.exp b/gdb/testsuite/lib/mi-support.exp
index 2846da74e47..851e490f4de 100644
--- a/gdb/testsuite/lib/mi-support.exp
+++ b/gdb/testsuite/lib/mi-support.exp
@@ -979,7 +979,11 @@ proc mi_run_cmd_full {use_mi_command args} {
 	send_gdb  "jump *$start\n"
 	warning "Using CLI jump command, expect run-to-main FAIL"
 	gdb_expect {
-	    -re "${run_match}&\"jump \\*${start}\\n\"\[\r\n\]+~\"Continuing at 0x\[0-9A-Fa-f\]+\\n.\"\[\r\n\]+\^running\[\r\n\]+\\*running,thread-id=\"\[^\"\]+\"\r\n${mi_gdb_prompt}" {}
+	    -re "&\"jump \\*${start}\\\\n\"\[\r\n\]+~\"Continuing at 0x\[0-9A-Fa-f\]+\.\\\\n\"\[\r\n\]+\\^running\[\r\n\]+\\*running,thread-id=\"\[^\"\]+\"\[\r\n\]+${mi_gdb_prompt}" {}
+	    timeout {
+		perror "Unable to start target"
+		return -1
+	    }
 	}
 	return 0
     }
-- 
2.12.2

Re: [PATCH 1/3] gdb/testsuite: Fix broken regexp in gdbstub case

[Pedro Alves]

On 04/09/2018 04:15 PM, Andrew Burgess wrote:
> When $use_gdb_stub is true then, when we start an MI target there's a
> regexp to match GDB's startup pattern.  Unfortunately the pattern is
> broken, and we're also missing a timeout case in the match list (which
> would have helped point out that the regexp was broken).

That seems to have been added by:
 https://sourceware.org/ml/gdb-patches/2015-12/msg00357.html

curious how it seems to have worked back then.  I wonder whether
it was GDB's CLI-in-MI output that changed meanwhile?

I'd have appreciated a bit more detail on what is actually broken in
the pattern.  The git log above gives no clue, and the diff is not
exactly easy to read.  I trust that it's now correct though.  Thus, OK.

Thanks,
Pedro Alves

Re: [PATCH 1/3] gdb/testsuite: Fix broken regexp in gdbstub case

[Andrew Burgess]

* Pedro Alves <palves@redhat.com> [2018-04-13 13:12:30 +0100]:

> On 04/09/2018 04:15 PM, Andrew Burgess wrote:
> > When $use_gdb_stub is true then, when we start an MI target there's a
> > regexp to match GDB's startup pattern.  Unfortunately the pattern is
> > broken, and we're also missing a timeout case in the match list (which
> > would have helped point out that the regexp was broken).
> 
> That seems to have been added by:
>  https://sourceware.org/ml/gdb-patches/2015-12/msg00357.html
> 
> curious how it seems to have worked back then.  I wonder whether
> it was GDB's CLI-in-MI output that changed meanwhile?
> 
> I'd have appreciated a bit more detail on what is actually broken in
> the pattern.  The git log above gives no clue, and the diff is not
> exactly easy to read.  I trust that it's now correct though.  Thus, OK.

I aim to please, so here's a revised commit message detailing what
changed.  The actual patch content hasn't changed.  I'll wait a few
days, but if I don't hear anything I'll assume your approval above
still stands and apply the patch.

Thanks,
Andrew

---

gdb/testsuite: Fix broken regexp in gdbstub case

When $use_gdb_stub is true then, when we start an MI target there's a
regexp to match GDB's startup pattern.  Unfortunately the pattern is
broken, and we're also missing a timeout case in the match list (which
would have helped point out that the regexp was broken).

The changes to the regexp are:

  1. Remove '${run_match}' prefix, the issued command doesn't include
  '${run_prefix}' so expecting '${run_match}' is wrong.

  2. Replaced '\\n' with '\\\\n' in order to match literal '\n' in
  GDBs output (that is, match a backslash followed by 'n', not a
  newline character).

  3. Replaced a '.' (matching any character) with '\.' to match a '.'
  and moved the '\.' into the correct place in the regexp.

  4. Replaced '\r\n' with '[\r\n]+' to match the end of a line.  This
  change isn't esential, but matches the other end of line patterns
  within this regexp.

Here's an example of the output that the regexp should match taken
from a testfile log, the first line is the command sent to GDB, and
the remaining lines are the response from GDB:

  jump *_start
  &"jump *_start\n"
  ~"Continuing at 0x10074.\n"
  ^running
  *running,thread-id="all"
  (gdb)

gdb/testsuite/ChangeLog:

	* lib/mi-support.exp (mi_run_cmd_full): Fix regexp and add a
	timeout.

---
 gdb/testsuite/ChangeLog          | 5 +++++
 gdb/testsuite/lib/mi-support.exp | 6 +++++-
 2 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/gdb/testsuite/lib/mi-support.exp b/gdb/testsuite/lib/mi-support.exp
index 2846da74e47..851e490f4de 100644
--- a/gdb/testsuite/lib/mi-support.exp
+++ b/gdb/testsuite/lib/mi-support.exp
@@ -979,7 +979,11 @@ proc mi_run_cmd_full {use_mi_command args} {
 	send_gdb  "jump *$start\n"
 	warning "Using CLI jump command, expect run-to-main FAIL"
 	gdb_expect {
-	    -re "${run_match}&\"jump \\*${start}\\n\"\[\r\n\]+~\"Continuing at 0x\[0-9A-Fa-f\]+\\n.\"\[\r\n\]+\^running\[\r\n\]+\\*running,thread-id=\"\[^\"\]+\"\r\n${mi_gdb_prompt}" {}
+	    -re "&\"jump \\*${start}\\\\n\"\[\r\n\]+~\"Continuing at 0x\[0-9A-Fa-f\]+\.\\\\n\"\[\r\n\]+\\^running\[\r\n\]+\\*running,thread-id=\"\[^\"\]+\"\[\r\n\]+${mi_gdb_prompt}" {}
+	    timeout {
+		perror "Unable to start target"
+		return -1
+	    }
 	}
 	return 0
     }
-- 
2.14.3

Re: [PATCH 1/3] gdb/testsuite: Fix broken regexp in gdbstub case

[Pedro Alves]

On 05/03/2018 08:41 PM, Andrew Burgess wrote:
> * Pedro Alves <palves@redhat.com> [2018-04-13 13:12:30 +0100]:
> 
>> On 04/09/2018 04:15 PM, Andrew Burgess wrote:
>>> When $use_gdb_stub is true then, when we start an MI target there's a
>>> regexp to match GDB's startup pattern.  Unfortunately the pattern is
>>> broken, and we're also missing a timeout case in the match list (which
>>> would have helped point out that the regexp was broken).
>>
>> That seems to have been added by:
>>  https://sourceware.org/ml/gdb-patches/2015-12/msg00357.html
>>
>> curious how it seems to have worked back then.  I wonder whether
>> it was GDB's CLI-in-MI output that changed meanwhile?
>>
>> I'd have appreciated a bit more detail on what is actually broken in
>> the pattern.  The git log above gives no clue, and the diff is not
>> exactly easy to read.  I trust that it's now correct though.  Thus, OK.
> 
> I aim to please, 

:-)

> so here's a revised commit message detailing what
> changed.  The actual patch content hasn't changed.  I'll wait a few
> days, but if I don't hear anything I'll assume your approval above
> still stands and apply the patch.

Perfect.  This is OK.

Thanks,
Pedro Alves

[PATCH 2/3] gdb/testsuite: Filter out some registers for riscv

[Andrew Burgess]

On riscv the cycle counter, and instructions retired counter CSRs are
read only, this causes problems in the gdb.base/callfuncs.exp test, as
the values in these CSRs change after an inferior call, the check that
no target registers have been modified then fails.

Luckily the test already has a mechanism in place for filtering out
registers that are modified (and can't be restored) by an inferior call,
so this commit adds the problem registers into this list for riscv.

In the future we may end up needing to filter out more CSRs, but right
now, for the targets I have access too, these are the only ones causing
problems.

gdb/testsuite/ChangeLog:

	* gdb.base/callfuncs.exp (fetch_all_registers): Add riscv register
	filter pattern.
---
 gdb/testsuite/ChangeLog              |  5 +++++
 gdb/testsuite/gdb.base/callfuncs.exp | 10 ++++++++++
 2 files changed, 15 insertions(+)

diff --git a/gdb/testsuite/gdb.base/callfuncs.exp b/gdb/testsuite/gdb.base/callfuncs.exp
index 94636938752..c5e39918c2a 100644
--- a/gdb/testsuite/gdb.base/callfuncs.exp
+++ b/gdb/testsuite/gdb.base/callfuncs.exp
@@ -285,6 +285,16 @@ proc fetch_all_registers {test} {
 	    }
 	    exp_continue
 	}
+	-re "^\(?:cycle\|instret\)\[ \t\]+\[^\r\n\]+\r\n" {
+	    if [istarget "riscv*-*-*"] {
+		# Filter out the cycle counter and instructions
+		# retired counter CSRs which are read-only, giving
+		# spurious differences.
+	    } else {
+		lappend all_registers_lines $expect_out(0,string)
+	    }
+	    exp_continue
+	}
 	-re "^\[^ \t\]+\[ \t\]+\[^\r\n\]+\r\n" {
 	    lappend all_registers_lines $expect_out(0,string)
 	    exp_continue
-- 
2.12.2

Re: [PATCH 2/3] gdb/testsuite: Filter out some registers for riscv

[Palmer Dabbelt]

On Mon, 09 Apr 2018 08:15:28 PDT (-0700), andrew.burgess@embecosm.com wrote:
> On riscv the cycle counter, and instructions retired counter CSRs are
> read only, this causes problems in the gdb.base/callfuncs.exp test, as
> the values in these CSRs change after an inferior call, the check that
> no target registers have been modified then fails.
>
> Luckily the test already has a mechanism in place for filtering out
> registers that are modified (and can't be restored) by an inferior call,
> so this commit adds the problem registers into this list for riscv.
>
> In the future we may end up needing to filter out more CSRs, but right
> now, for the targets I have access too, these are the only ones causing
> problems.
>
> gdb/testsuite/ChangeLog:
>
> 	* gdb.base/callfuncs.exp (fetch_all_registers): Add riscv register
> 	filter pattern.
> ---
>  gdb/testsuite/ChangeLog              |  5 +++++
>  gdb/testsuite/gdb.base/callfuncs.exp | 10 ++++++++++
>  2 files changed, 15 insertions(+)
>
> diff --git a/gdb/testsuite/gdb.base/callfuncs.exp b/gdb/testsuite/gdb.base/callfuncs.exp
> index 94636938752..c5e39918c2a 100644
> --- a/gdb/testsuite/gdb.base/callfuncs.exp
> +++ b/gdb/testsuite/gdb.base/callfuncs.exp
> @@ -285,6 +285,16 @@ proc fetch_all_registers {test} {
>  	    }
>  	    exp_continue
>  	}
> +	-re "^\(?:cycle\|instret\)\[ \t\]+\[^\r\n\]+\r\n" {
> +	    if [istarget "riscv*-*-*"] {
> +		# Filter out the cycle counter and instructions
> +		# retired counter CSRs which are read-only, giving
> +		# spurious differences.
> +	    } else {
> +		lappend all_registers_lines $expect_out(0,string)
> +	    }
> +	    exp_continue
> +	}
>  	-re "^\[^ \t\]+\[ \t\]+\[^\r\n\]+\r\n" {
>  	    lappend all_registers_lines $expect_out(0,string)
>  	    exp_continue

I think we only want to check the X and F registers here -- essentially every 
CSR is a special register where you can't really rely on the value not being 
changed somewhere by hardware.  For example:

* The interrupt pending bits could flip at any point, even if interrupts are 
  disabled.
* The floating-point dirty and exception state bits could change if a
  floating-point instruction executes.
* The various trap CSRs (epc, badaddr, cause, etc) get set whenever a trap is 
  executed.

Re: [PATCH 2/3] gdb/testsuite: Filter out some registers for riscv

[Andrew Burgess]

* Palmer Dabbelt <palmer@sifive.com> [2018-04-09 14:28:33 -0700]:

> On Mon, 09 Apr 2018 08:15:28 PDT (-0700), andrew.burgess@embecosm.com wrote:
> > On riscv the cycle counter, and instructions retired counter CSRs are
> > read only, this causes problems in the gdb.base/callfuncs.exp test, as
> > the values in these CSRs change after an inferior call, the check that
> > no target registers have been modified then fails.
> > 
> > Luckily the test already has a mechanism in place for filtering out
> > registers that are modified (and can't be restored) by an inferior call,
> > so this commit adds the problem registers into this list for riscv.
> > 
> > In the future we may end up needing to filter out more CSRs, but right
> > now, for the targets I have access too, these are the only ones causing
> > problems.
> > 
> > gdb/testsuite/ChangeLog:
> > 
> > 	* gdb.base/callfuncs.exp (fetch_all_registers): Add riscv register
> > 	filter pattern.
> > ---
> >  gdb/testsuite/ChangeLog              |  5 +++++
> >  gdb/testsuite/gdb.base/callfuncs.exp | 10 ++++++++++
> >  2 files changed, 15 insertions(+)
> > 
> > diff --git a/gdb/testsuite/gdb.base/callfuncs.exp b/gdb/testsuite/gdb.base/callfuncs.exp
> > index 94636938752..c5e39918c2a 100644
> > --- a/gdb/testsuite/gdb.base/callfuncs.exp
> > +++ b/gdb/testsuite/gdb.base/callfuncs.exp
> > @@ -285,6 +285,16 @@ proc fetch_all_registers {test} {
> >  	    }
> >  	    exp_continue
> >  	}
> > +	-re "^\(?:cycle\|instret\)\[ \t\]+\[^\r\n\]+\r\n" {
> > +	    if [istarget "riscv*-*-*"] {
> > +		# Filter out the cycle counter and instructions
> > +		# retired counter CSRs which are read-only, giving
> > +		# spurious differences.
> > +	    } else {
> > +		lappend all_registers_lines $expect_out(0,string)
> > +	    }
> > +	    exp_continue
> > +	}
> >  	-re "^\[^ \t\]+\[ \t\]+\[^\r\n\]+\r\n" {
> >  	    lappend all_registers_lines $expect_out(0,string)
> >  	    exp_continue
> 
> I think we only want to check the X and F registers here -- essentially
> every CSR is a special register where you can't really rely on the value not
> being changed somewhere by hardware.  For example:
> 
> * The interrupt pending bits could flip at any point, even if interrupts are
> disabled.
> * The floating-point dirty and exception state bits could change if a
>  floating-point instruction executes.
> * The various trap CSRs (epc, badaddr, cause, etc) get set whenever a trap
> is  executed.

I'm not sure about the second case.  If GDB is stopped and we perform
an inferior call, then ideally the entire floating point state,
including contents of things like fcsr would be reset, otherwise, when
we continue the behaviour might not be as we expect.

I do agree with you that the two registers I've filtered so far
probably aren't enough, but I'm really reluctant to _only_ check X and
F registers.  I think a better selection would be X, F, and read/write
user CSRs.  Which means I need to build the list of CSRs to filter, I
was hoping to put that off for another day for now...

Let me know how you'd feel about leaving this as it is for now, and
extending the filter list at a later date.

thanks,
Andrew

Re: [PATCH 2/3] gdb/testsuite: Filter out some registers for riscv

[Palmer Dabbelt]

On Mon, 09 Apr 2018 15:26:07 PDT (-0700), andrew.burgess@embecosm.com wrote:
> * Palmer Dabbelt <palmer@sifive.com> [2018-04-09 14:28:33 -0700]:
>
>> On Mon, 09 Apr 2018 08:15:28 PDT (-0700), andrew.burgess@embecosm.com wrote:
>> > On riscv the cycle counter, and instructions retired counter CSRs are
>> > read only, this causes problems in the gdb.base/callfuncs.exp test, as
>> > the values in these CSRs change after an inferior call, the check that
>> > no target registers have been modified then fails.
>> >
>> > Luckily the test already has a mechanism in place for filtering out
>> > registers that are modified (and can't be restored) by an inferior call,
>> > so this commit adds the problem registers into this list for riscv.
>> >
>> > In the future we may end up needing to filter out more CSRs, but right
>> > now, for the targets I have access too, these are the only ones causing
>> > problems.
>> >
>> > gdb/testsuite/ChangeLog:
>> >
>> > 	* gdb.base/callfuncs.exp (fetch_all_registers): Add riscv register
>> > 	filter pattern.
>> > ---
>> >  gdb/testsuite/ChangeLog              |  5 +++++
>> >  gdb/testsuite/gdb.base/callfuncs.exp | 10 ++++++++++
>> >  2 files changed, 15 insertions(+)
>> >
>> > diff --git a/gdb/testsuite/gdb.base/callfuncs.exp b/gdb/testsuite/gdb.base/callfuncs.exp
>> > index 94636938752..c5e39918c2a 100644
>> > --- a/gdb/testsuite/gdb.base/callfuncs.exp
>> > +++ b/gdb/testsuite/gdb.base/callfuncs.exp
>> > @@ -285,6 +285,16 @@ proc fetch_all_registers {test} {
>> >  	    }
>> >  	    exp_continue
>> >  	}
>> > +	-re "^\(?:cycle\|instret\)\[ \t\]+\[^\r\n\]+\r\n" {
>> > +	    if [istarget "riscv*-*-*"] {
>> > +		# Filter out the cycle counter and instructions
>> > +		# retired counter CSRs which are read-only, giving
>> > +		# spurious differences.
>> > +	    } else {
>> > +		lappend all_registers_lines $expect_out(0,string)
>> > +	    }
>> > +	    exp_continue
>> > +	}
>> >  	-re "^\[^ \t\]+\[ \t\]+\[^\r\n\]+\r\n" {
>> >  	    lappend all_registers_lines $expect_out(0,string)
>> >  	    exp_continue
>>
>> I think we only want to check the X and F registers here -- essentially
>> every CSR is a special register where you can't really rely on the value not
>> being changed somewhere by hardware.  For example:
>>
>> * The interrupt pending bits could flip at any point, even if interrupts are
>> disabled.
>> * The floating-point dirty and exception state bits could change if a
>>  floating-point instruction executes.
>> * The various trap CSRs (epc, badaddr, cause, etc) get set whenever a trap
>> is  executed.
>
> I'm not sure about the second case.  If GDB is stopped and we perform
> an inferior call, then ideally the entire floating point state,
> including contents of things like fcsr would be reset, otherwise, when
> we continue the behaviour might not be as we expect.
>
> I do agree with you that the two registers I've filtered so far
> probably aren't enough, but I'm really reluctant to _only_ check X and
> F registers.  I think a better selection would be X, F, and read/write
> user CSRs.  Which means I need to build the list of CSRs to filter, I
> was hoping to put that off for another day for now...
>
> Let me know how you'd feel about leaving this as it is for now, and
> extending the filter list at a later date.

I think it's fine for now, we can fix it when another test fails :)

Re: [PATCH 2/3] gdb/testsuite: Filter out some registers for riscv

[Pedro Alves]

On 04/10/2018 09:25 PM, Palmer Dabbelt wrote:
> On Mon, 09 Apr 2018 15:26:07 PDT (-0700), andrew.burgess@embecosm.com wrote:

>> Let me know how you'd feel about leaving this as it is for now, and
>> extending the filter list at a later date.
> 
> I think it's fine for now, we can fix it when another test fails :)

This is OK, then.

For the register lists, I wonder whether a better option would be to
instead let gdb determine which registers should be compared, instead
of hard coding in the testcase.  Like, look at the save/restore
reggroups instead of all registers.

Thanks,
Pedro Alves