Re: [PATCH 1/2] Add linux_get_hwcap

[Simon Marchi <simark@simark.ca>]

On 2019-03-25 8:05 a.m., Alan Hayward wrote:
> Tidy up calls to read HWCAP (and HWCAP2) by adding common functions,
> removing the PPC and AArch64 specific versions.
> 
> The only function difference is in aarch64_linux_core_read_description - if
> the hwcap read fails it now return a valid description instead of nullptr.

ARM also seems to have changed behavior, both the native target and core 
target.  It's not necessarily, a problem, as long as it's a conscious 
change.

> diff --git a/gdb/linux-tdep.h b/gdb/linux-tdep.h
> index 824ba3afaf..ee9c2bcc90 100644
> --- a/gdb/linux-tdep.h
> +++ b/gdb/linux-tdep.h
> @@ -61,4 +61,10 @@ extern void linux_init_abi (struct gdbarch_info info, struct gdbarch *gdbarch);
>   
>   extern int linux_is_uclinux (void);
>   
> +/* Fetch the AT_HWCAP entry from the auxv vector for the given TARGET.  */
> +extern CORE_ADDR linux_get_hwcap (struct target_ops *target);
> +
> +/* Fetch the AT_HWCAP2 entry from the auxv vector for the given TARGET.  */
> +extern CORE_ADDR linux_get_hwcap2 (struct target_ops *target);

For these two functions, can you mention that 0 is returned if the 
search in the AUXV vector fails?

I was a bit surprised you didn't keep your version returning a bool to 
indicate whether the search succeeded or failed, but if this version is 
good enough, I am fine with it.  If we ever have a target that really 
needs to differentiate between a lookup failure and a lookup success 
that returns the value 0, we can change it back or add another overload.

This patch LGTM with the comment above updated.

Simon