[gdb/python] Make gdb.UnwindInfo.add_saved_register more robust

  On arm-linux, until commit bbb12eb9c84 ("gdb/arm: Remove tpidruro register
from non-FreeBSD target descriptions") I ran into:
...
FAIL: gdb.base/inline-frame-cycle-unwind.exp: cycle at level 5: \
  backtrace when the unwind is broken at frame 5
...

What happens is the following:
- the TestUnwinder from inline-frame-cycle-unwind.py calls
  gdb.UnwindInfo.add_saved_register with reg == tpidruro and value
  "<unavailable>",
- pyuw_sniffer calls value->contents ().data () to access the value of the
  register, which throws an UNAVAILABLE_ERROR,
- this causes the TestUnwinder unwinder to fail, after which another unwinder
  succeeds and returns the correct frame, and
- the test-case fails because it's counting on the TestUnwinder to succeed and
  return an incorrect frame.

Fix this by checking for !value::entirely_available as well as
valued::optimized_out in unwind_infopy_add_saved_register.

Tested on x86_64-linux and arm-linux.

PR python/31437
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=31437
---
 gdb/python/py-unwind.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)


base-commit: a6a3b67fa9052bba81ed91a38569c11ecb95baf1
  

On 3/2/24 13:36, Tom de Vries wrote:
> On arm-linux, until commit bbb12eb9c84 ("gdb/arm: Remove tpidruro register
> from non-FreeBSD target descriptions") I ran into:
> ...
> FAIL: gdb.base/inline-frame-cycle-unwind.exp: cycle at level 5: \
>    backtrace when the unwind is broken at frame 5
> ...
> 
> What happens is the following:
> - the TestUnwinder from inline-frame-cycle-unwind.py calls
>    gdb.UnwindInfo.add_saved_register with reg == tpidruro and value
>    "<unavailable>",
> - pyuw_sniffer calls value->contents ().data () to access the value of the
>    register, which throws an UNAVAILABLE_ERROR,
> - this causes the TestUnwinder unwinder to fail, after which another unwinder
>    succeeds and returns the correct frame, and
> - the test-case fails because it's counting on the TestUnwinder to succeed and
>    return an incorrect frame.
> 
> Fix this by checking for !value::entirely_available as well as
> valued::optimized_out in unwind_infopy_add_saved_register.
> 
> Tested on x86_64-linux and arm-linux.
> 

Ping.

Thanks,
- Tom

> PR python/31437
> Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=31437
> ---
>   gdb/python/py-unwind.c | 12 ++++++++++++
>   1 file changed, 12 insertions(+)
> 
> diff --git a/gdb/python/py-unwind.c b/gdb/python/py-unwind.c
> index 56f925bc57f..1c1289f7e7d 100644
> --- a/gdb/python/py-unwind.c
> +++ b/gdb/python/py-unwind.c
> @@ -362,6 +362,18 @@ unwind_infopy_add_saved_register (PyObject *self, PyObject *args, PyObject *kw)
>         return nullptr;
>       }
>   
> +  if (value->optimized_out () || !value->entirely_available ())
> +    {
> +      /* If we allow this value to be registered here, pyuw_sniffer is going
> +	 to run into an exception when trying to access its contents.
> +	 Throwing an exception here just puts a burden on the user to
> +	 implement the same checks on the user side.  We could return False
> +	 here and True otherwise, but again that might require changes in user
> +	 code.  So, handle this with minimal impact for the user, while
> +	 improving robustness: silently ignore the register/value pair.  */
> +      Py_RETURN_NONE;
> +    }
> +
>     gdbpy_ref<> new_value = gdbpy_ref<>::new_reference (pyo_reg_value);
>     bool found = false;
>     for (saved_reg &reg : *unwind_info->saved_regs)
> 
> base-commit: a6a3b67fa9052bba81ed91a38569c11ecb95baf1
  

On 3/19/24 10:34, Tom de Vries wrote:
> On 3/2/24 13:36, Tom de Vries wrote:
>> On arm-linux, until commit bbb12eb9c84 ("gdb/arm: Remove tpidruro 
>> register
>> from non-FreeBSD target descriptions") I ran into:
>> ...
>> FAIL: gdb.base/inline-frame-cycle-unwind.exp: cycle at level 5: \
>>    backtrace when the unwind is broken at frame 5
>> ...
>>
>> What happens is the following:
>> - the TestUnwinder from inline-frame-cycle-unwind.py calls
>>    gdb.UnwindInfo.add_saved_register with reg == tpidruro and value
>>    "<unavailable>",
>> - pyuw_sniffer calls value->contents ().data () to access the value of 
>> the
>>    register, which throws an UNAVAILABLE_ERROR,
>> - this causes the TestUnwinder unwinder to fail, after which another 
>> unwinder
>>    succeeds and returns the correct frame, and
>> - the test-case fails because it's counting on the TestUnwinder to 
>> succeed and
>>    return an incorrect frame.
>>
>> Fix this by checking for !value::entirely_available as well as
>> valued::optimized_out in unwind_infopy_add_saved_register.
>>
>> Tested on x86_64-linux and arm-linux.
>>
> 
> Ping.
> 

Ping once more.

Also found that this fixes the gdb.base/inline-frame-cycle-unwind.exp 
failure on ppc64le reported in PR30548.

Thanks,
- Tom

>> PR python/31437
>> Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=31437
>> ---
>>   gdb/python/py-unwind.c | 12 ++++++++++++
>>   1 file changed, 12 insertions(+)
>>
>> diff --git a/gdb/python/py-unwind.c b/gdb/python/py-unwind.c
>> index 56f925bc57f..1c1289f7e7d 100644
>> --- a/gdb/python/py-unwind.c
>> +++ b/gdb/python/py-unwind.c
>> @@ -362,6 +362,18 @@ unwind_infopy_add_saved_register (PyObject *self, 
>> PyObject *args, PyObject *kw)
>>         return nullptr;
>>       }
>> +  if (value->optimized_out () || !value->entirely_available ())
>> +    {
>> +      /* If we allow this value to be registered here, pyuw_sniffer 
>> is going
>> +     to run into an exception when trying to access its contents.
>> +     Throwing an exception here just puts a burden on the user to
>> +     implement the same checks on the user side.  We could return False
>> +     here and True otherwise, but again that might require changes in 
>> user
>> +     code.  So, handle this with minimal impact for the user, while
>> +     improving robustness: silently ignore the register/value pair.  */
>> +      Py_RETURN_NONE;
>> +    }
>> +
>>     gdbpy_ref<> new_value = gdbpy_ref<>::new_reference (pyo_reg_value);
>>     bool found = false;
>>     for (saved_reg &reg : *unwind_info->saved_regs)
>>
>> base-commit: a6a3b67fa9052bba81ed91a38569c11ecb95baf1
>
  

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

> On arm-linux, until commit bbb12eb9c84 ("gdb/arm: Remove tpidruro register
> from non-FreeBSD target descriptions") I ran into:
> ...
> FAIL: gdb.base/inline-frame-cycle-unwind.exp: cycle at level 5: \
>   backtrace when the unwind is broken at frame 5
> ...
>
> What happens is the following:
> - the TestUnwinder from inline-frame-cycle-unwind.py calls
>   gdb.UnwindInfo.add_saved_register with reg == tpidruro and value
>   "<unavailable>",
> - pyuw_sniffer calls value->contents ().data () to access the value of the
>   register, which throws an UNAVAILABLE_ERROR,
> - this causes the TestUnwinder unwinder to fail, after which another unwinder
>   succeeds and returns the correct frame, and
> - the test-case fails because it's counting on the TestUnwinder to succeed and
>   return an incorrect frame.
>
> Fix this by checking for !value::entirely_available as well as
> valued::optimized_out in unwind_infopy_add_saved_register.
>
> Tested on x86_64-linux and arm-linux.
>
> PR python/31437
> Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=31437
> ---
>  gdb/python/py-unwind.c | 12 ++++++++++++
>  1 file changed, 12 insertions(+)
>
> diff --git a/gdb/python/py-unwind.c b/gdb/python/py-unwind.c
> index 56f925bc57f..1c1289f7e7d 100644
> --- a/gdb/python/py-unwind.c
> +++ b/gdb/python/py-unwind.c
> @@ -362,6 +362,18 @@ unwind_infopy_add_saved_register (PyObject *self, PyObject *args, PyObject *kw)
>        return nullptr;
>      }
>  
> +  if (value->optimized_out () || !value->entirely_available ())
> +    {
> +      /* If we allow this value to be registered here, pyuw_sniffer is going
> +	 to run into an exception when trying to access its contents.
> +	 Throwing an exception here just puts a burden on the user to
> +	 implement the same checks on the user side.  We could return False
> +	 here and True otherwise, but again that might require changes in user
> +	 code.  So, handle this with minimal impact for the user, while
> +	 improving robustness: silently ignore the register/value pair.  */
> +      Py_RETURN_NONE;
> +    }

Thanks for fixing this.

I agree with this approach.  Registers that aren't saved will be
reported as optimized out anyway, which seems good enough -- though
having them report as unavailable might be better, but that's not a
problem for this patch.

I think returning True/False would be OK,  I don't think users are
likely to be depending on this function returning None.  But that said,
I don't see any immediate need to change the return type; if a user
really wants to know then they can check for unavailable/optimized-out
values prior to saving the register value.

Approved-By: Andrew Burgess <aburgess@redhat.com>

Thanks,
Andrew


> +
>    gdbpy_ref<> new_value = gdbpy_ref<>::new_reference (pyo_reg_value);
>    bool found = false;
>    for (saved_reg &reg : *unwind_info->saved_regs)
>
> base-commit: a6a3b67fa9052bba81ed91a38569c11ecb95baf1
> -- 
> 2.35.3