Re: [PATCH 1/4] *-fbsd-nat: Handle null inferior in read_description.

[John Baldwin <jhb@FreeBSD.org>]

On 7/6/23 9:18 AM, Simon Marchi wrote:
> On 5/26/23 13:57, John Baldwin wrote:
>> Don't invoke ptrace in the target read_description method if there is
>> not an active inferior to query via ptrace.  Instead, use the default
>> register set for the architecture.
>>
>> Previously the native target could report an error from a failed
>> ptrace operation when fetching a tdesc without an attached process.
>> For example on FreeBSD/amd64:
>>
>> (gdb) target native
>> Done.  Use the "run" command to start a process.
>> (gdb) unset tdesc filename
>> Couldn't get registers: Operation not permitted.
>> ---
>>   gdb/aarch64-fbsd-nat.c | 3 +++
>>   gdb/amd64-fbsd-nat.c   | 3 +++
>>   gdb/arm-fbsd-nat.c     | 3 +++
>>   gdb/i386-fbsd-nat.c    | 3 +++
>>   4 files changed, 12 insertions(+)
>>
>> diff --git a/gdb/aarch64-fbsd-nat.c b/gdb/aarch64-fbsd-nat.c
>> index 709f5162ce0..29b58e5ff4a 100644
>> --- a/gdb/aarch64-fbsd-nat.c
>> +++ b/gdb/aarch64-fbsd-nat.c
>> @@ -120,6 +120,9 @@ aarch64_fbsd_nat_target::store_registers (struct regcache *regcache,
>>   const struct target_desc *
>>   aarch64_fbsd_nat_target::read_description ()
>>   {
>> +  if (inferior_ptid == null_ptid)
>> +    return nullptr;
>> +
>>     aarch64_features features;
>>     features.tls = have_regset (inferior_ptid, NT_ARM_TLS)? 1 : 0;
>>     return aarch64_read_description (features);
>> diff --git a/gdb/amd64-fbsd-nat.c b/gdb/amd64-fbsd-nat.c
>> index bae267f230f..adbd85571a5 100644
>> --- a/gdb/amd64-fbsd-nat.c
>> +++ b/gdb/amd64-fbsd-nat.c
>> @@ -310,6 +310,9 @@ amd64_fbsd_nat_target::read_description ()
>>     struct reg regs;
>>     int is64;
>>   
>> +  if (inferior_ptid == null_ptid)
>> +    return nullptr;
> 
> Why do you return nullptr here, instead of calling the beneath target,
> as done in arm-fbsd-nat.c?  Calling the beneath target seems like the
> right thing to do.  In practice, it will likely reach
> dummy_target::read_description, which returns nullptr.

I think you are right, and I will change both this and patch 2 for the
Linux targets to call the beneath one.  My thinking was that in general
returning nullptr causes the caller code to pick a "default" tdesc, but
that for arm-fbsd-nat.c I was trying to match the existing code
(copied from arm-linux-nat.c) which calls the beneath target for its fallback
instead of returning nullptr.

-- 
John Baldwin