[2/2] Don't lose language determined from the "main" name (fix gdb.ada/minsyms.exp)
Commit Message
On 11/21/2017 04:42 PM, Pedro Alves wrote:
> On 11/21/2017 04:23 PM, Sergio Durigan Junior wrote:
>> Since this is guaranteed to be an stap probe, WDYT about moving this
>> scoped_restore_current_language to
>> stap-probe.c:stap_evaluate_probe_argument? This way we won't be bit by
>> this problem in other parts that also evaluate arguments of probes.
>>
>> Arguably, this should be set for every probe type IMHO, but it's fine if
>> we just do it for stap probes for now.
>
> That sounds like a good idea. But we could do it in
> evaluate_probe_argument then, which handles all probe types?
>
> [In your probe C++ification, that translates to evaluate_probe_argument
> becoming a non-virtual method of probe, which then calls into a
> protected virtual method that is overridden by the actual probe
> implementation (see e.g., the do_xxx methods of class ui_out).]
Hmm, maybe what we need instead is to make expression evaluation
never set the selected frame (and thus language as side effect)
if it wasn't selected/set already. Like below. This fixes
the testcase too. I'll run the full testsuite now. WDYT?
From c5fd954787e058abcc72aca329414b0edb40ac1b Mon Sep 17 00:00:00 2001
From: Pedro Alves <palves@redhat.com>
Date: Tue, 21 Nov 2017 16:50:32 +0000
Subject: [PATCH] alternative
---
gdb/eval.c | 12 +++++++++---
1 file changed, 9 insertions(+), 3 deletions(-)
Comments
On 11/21/2017 04:56 PM, Pedro Alves wrote:
> On 11/21/2017 04:42 PM, Pedro Alves wrote:
>> On 11/21/2017 04:23 PM, Sergio Durigan Junior wrote:
>
>>> Since this is guaranteed to be an stap probe, WDYT about moving this
>>> scoped_restore_current_language to
>>> stap-probe.c:stap_evaluate_probe_argument? This way we won't be bit by
>>> this problem in other parts that also evaluate arguments of probes.
>>>
>>> Arguably, this should be set for every probe type IMHO, but it's fine if
>>> we just do it for stap probes for now.
>>
>> That sounds like a good idea. But we could do it in
>> evaluate_probe_argument then, which handles all probe types?
>>
>> [In your probe C++ification, that translates to evaluate_probe_argument
>> becoming a non-virtual method of probe, which then calls into a
>> protected virtual method that is overridden by the actual probe
>> implementation (see e.g., the do_xxx methods of class ui_out).]
>
> Hmm, maybe what we need instead is to make expression evaluation
> never set the selected frame (and thus language as side effect)
> if it wasn't selected/set already. Like below. This fixes
> the testcase too. I'll run the full testsuite now. WDYT?
Full testsuite run complete; no regressions.
Thanks,
Pedro Alves
On Tuesday, November 21 2017, Pedro Alves wrote:
> On 11/21/2017 04:42 PM, Pedro Alves wrote:
>> On 11/21/2017 04:23 PM, Sergio Durigan Junior wrote:
>
>>> Since this is guaranteed to be an stap probe, WDYT about moving this
>>> scoped_restore_current_language to
>>> stap-probe.c:stap_evaluate_probe_argument? This way we won't be bit by
>>> this problem in other parts that also evaluate arguments of probes.
>>>
>>> Arguably, this should be set for every probe type IMHO, but it's fine if
>>> we just do it for stap probes for now.
>>
>> That sounds like a good idea. But we could do it in
>> evaluate_probe_argument then, which handles all probe types?
>>
>> [In your probe C++ification, that translates to evaluate_probe_argument
>> becoming a non-virtual method of probe, which then calls into a
>> protected virtual method that is overridden by the actual probe
>> implementation (see e.g., the do_xxx methods of class ui_out).]
>
> Hmm, maybe what we need instead is to make expression evaluation
> never set the selected frame (and thus language as side effect)
> if it wasn't selected/set already. Like below. This fixes
> the testcase too. I'll run the full testsuite now. WDYT?
That does look better, indeed. I was trying to think if we'd encounter
any situation where setting the language is on of the desired effects,
but couldn't think of any.
Thanks,
> From c5fd954787e058abcc72aca329414b0edb40ac1b Mon Sep 17 00:00:00 2001
> From: Pedro Alves <palves@redhat.com>
> Date: Tue, 21 Nov 2017 16:50:32 +0000
> Subject: [PATCH] alternative
>
> ---
> gdb/eval.c | 12 +++++++++---
> 1 file changed, 9 insertions(+), 3 deletions(-)
>
> diff --git a/gdb/eval.c b/gdb/eval.c
> index 14a3e05..086ac59 100644
> --- a/gdb/eval.c
> +++ b/gdb/eval.c
> @@ -1319,7 +1319,6 @@ evaluate_subexp_standard (struct type *expect_type,
>
> {
> struct symbol *sym = exp->elts[pc + 1].symbol;
> - struct frame_info *frame;
>
> if (noside == EVAL_AVOID_SIDE_EFFECTS)
> return value_zero (SYMBOL_TYPE (sym), not_lval);
> @@ -1329,7 +1328,9 @@ evaluate_subexp_standard (struct type *expect_type,
> error (_("Symbol \"%s\" does not have any specific entry value"),
> SYMBOL_PRINT_NAME (sym));
>
> - frame = get_selected_frame (NULL);
> + frame_info *frame = get_selected_frame_if_set ();
> + if (frame == NULL)
> + frame = get_current_frame ();
> return SYMBOL_COMPUTED_OPS (sym)->read_variable_at_entry (sym, frame);
> }
>
> @@ -1381,7 +1382,12 @@ evaluate_subexp_standard (struct type *expect_type,
> + gdbarch_num_pseudo_regs (exp->gdbarch))
> val = value_zero (register_type (exp->gdbarch, regno), not_lval);
> else
> - val = value_of_register (regno, get_selected_frame (NULL));
> + {
> + frame_info *frame = get_selected_frame_if_set ();
> + if (frame == NULL)
> + frame = get_current_frame ();
> + val = value_of_register (regno, frame);
> + }
> if (val == NULL)
> error (_("Value of register %s not available."), name);
> else
> --
> 2.5.5
n 11/21/2017 05:15 PM, Sergio Durigan Junior wrote:
> On Tuesday, November 21 2017, Pedro Alves wrote:
>
>> On 11/21/2017 04:42 PM, Pedro Alves wrote:
>>> On 11/21/2017 04:23 PM, Sergio Durigan Junior wrote:
>>
>>>> Since this is guaranteed to be an stap probe, WDYT about moving this
>>>> scoped_restore_current_language to
>>>> stap-probe.c:stap_evaluate_probe_argument? This way we won't be bit by
>>>> this problem in other parts that also evaluate arguments of probes.
>>>>
>>>> Arguably, this should be set for every probe type IMHO, but it's fine if
>>>> we just do it for stap probes for now.
>>>
>>> That sounds like a good idea. But we could do it in
>>> evaluate_probe_argument then, which handles all probe types?
>>>
>>> [In your probe C++ification, that translates to evaluate_probe_argument
>>> becoming a non-virtual method of probe, which then calls into a
>>> protected virtual method that is overridden by the actual probe
>>> implementation (see e.g., the do_xxx methods of class ui_out).]
>>
>> Hmm, maybe what we need instead is to make expression evaluation
>> never set the selected frame (and thus language as side effect)
>> if it wasn't selected/set already. Like below. This fixes
>> the testcase too. I'll run the full testsuite now. WDYT?
>
> That does look better, indeed. I was trying to think if we'd encounter
> any situation where setting the language is on of the desired effects,
> but couldn't think of any.
Alright, I've sent a v2 now.
Thanks,
Pedro Alves
@@ -1319,7 +1319,6 @@ evaluate_subexp_standard (struct type *expect_type,
{
struct symbol *sym = exp->elts[pc + 1].symbol;
- struct frame_info *frame;
if (noside == EVAL_AVOID_SIDE_EFFECTS)
return value_zero (SYMBOL_TYPE (sym), not_lval);
@@ -1329,7 +1328,9 @@ evaluate_subexp_standard (struct type *expect_type,
error (_("Symbol \"%s\" does not have any specific entry value"),
SYMBOL_PRINT_NAME (sym));
- frame = get_selected_frame (NULL);
+ frame_info *frame = get_selected_frame_if_set ();
+ if (frame == NULL)
+ frame = get_current_frame ();
return SYMBOL_COMPUTED_OPS (sym)->read_variable_at_entry (sym, frame);
}
@@ -1381,7 +1382,12 @@ evaluate_subexp_standard (struct type *expect_type,
+ gdbarch_num_pseudo_regs (exp->gdbarch))
val = value_zero (register_type (exp->gdbarch, regno), not_lval);
else
- val = value_of_register (regno, get_selected_frame (NULL));
+ {
+ frame_info *frame = get_selected_frame_if_set ();
+ if (frame == NULL)
+ frame = get_current_frame ();
+ val = value_of_register (regno, frame);
+ }
if (val == NULL)
error (_("Value of register %s not available."), name);
else