Re: [PATCH] Allow direct access relocations referencing a protected function symbol

[Fangrui Song <i@maskray.me>]

On 2021-06-16, Alan Modra wrote:
>On Mon, Jun 14, 2021 at 08:19:32PM -0700, Fangrui Song wrote:
>> On 2021-06-15, Alan Modra wrote:
>> > On Mon, Jun 14, 2021 at 10:43:36AM -0700, Fangrui Song wrote:
>> > > I shall add a note that this patch will restore the behavior before late
>> > > 2015 or early 2016.
>> >
>> > Also add a note that the patch you posted is incorrect.
>>
>> Only the description needs an update. The code is correct.
>
>How would you respond if I submitted a patch to lld that you could see
>was wrong, but I kept arrogantly claiming was correct?
>
>-- 
>Alan Modra
>Australia Development Lab, IBM

If you have found a mistake, please point out the mistake. No euphemism or
sarcasm is needed.

I said "Treat protected function symbols local to the component like hidden/internal visibilities."
That is certainly not my invention of the semantics of STV_PROTECTED.
Multiple people agreed on this. https://groups.google.com/g/generic-abi/c/waK1dGiUGvM/m/7Uid2HH4IQAJ
(which I have linked to previously)

Solaris ld agrees with this (from the dump of the generic-abi discussion).
gold agrees with this: https://sourceware.org/bugzilla/show_bug.cgi?id=19823#c1


I have heard multiple folks said "protected symbols are broken."
Why?  As-is,

__attribute__((visibility("protected"))) void *foo () {
   return (void *)foo;
}

gcc -fpic -shared b.c -fuse-ld=bfd b.c  is broken.

The code generation does have inferior optimization in certain places but they
are not correctness issues.

My intention was to fix the ld problem. I have answered that there is no
compatibility issue.