[PATCH v2] gold: Always resolve hidden weak undefined in PIE to 0

[Alan Modra]

On Sat, Aug 17, 2024 at 07:14:53AM -0700, H.J. Lu wrote:
> Hidden weak undefined symbol in PIE should be resolved to 0.

I don't think this patch is correct.  Why single out hidden visibility
rather than testing for non-default visibility?  Also, what is special
about weak symbols?  ie. if the patch is correct for STB_WEAK
undefined, why not STB_GLOBAL undefined?

Also, despite the comment of "Return true if the final value of this
symbol is known at link time", final_value_is_known appears to
actually be "Return true if the value at runtime of this symbol is
known".  That makes me wonder about changes here affecting targets
other than x86.  How did you test the change?

-- 
Alan Modra