Re: [PATCH] gdb.base/sigstep.exp: Mark xfail on unsupported scenarios on riscv64-linux-gnu

[Lancelot SIX <lsix@lancelotsix.com>]

On Mon, Jul 05, 2021 at 09:39:08AM +0100, Andrew Burgess wrote:
> 
> I wonder if, rather than hard coding the xfail to check for RISC-V, we
> could add a helper function to lib/gdb.exp like:
> 
>   proc uses_software_single_step_p {} {
>     ...
>   }
> 
> which would return true/false indicating if the current target uses
> software single stepping?
> 
> We might want to use this test in other places later, and if there are
> other targets that also have this issue it is then super easy for them
> to add their architecture into this new proc.
> 
> However, rather than have the new proc check a list of architectures,
> we could possible make it do something like this:
> 
>   (gdb) pipe maintenance print architecture | grep gdbarch_software_single_step_p
>   gdbarch_dump: gdbarch_software_single_step_p() = 0
> 
> The above was run on x86-64 which doesn't use software single step,
> but I believe on RISC-V/Linux the above should look like this:
> 
>   (gdb) pipe maintenance print architecture | grep gdbarch_software_single_step_p
>   gdbarch_dump: gdbarch_software_single_step_p() = 1

Hi

Yes, it does!

> 
> This would then automatically catch any target that use software
> single stepping.
> 
> What do you think?
> 
> Thanks,
> Andrew
> 

Your option is nice, but as Pedro pointed out in a later email there is
already a proc can_single_step_to_signal_handler that I somehow missed.
For the context of this particular patch, I think I’ll go with it.

That being said, I kinda prefer your approach that should be equivalent
(I guess) and avoid the hard written list of unsupported host targets.
I’ll add this to my TODO list.

Lancelot.