Re: [PATCH v2 0/6] gas: Add support for LLVM addrsig and addrsig_sym.

[Tatsuyuki Ishi <ishitatsuyuki@gmail.com>]

On Tue, Jun 28, 2022 at 6:20 PM Alan Modra <amodra@gmail.com> wrote:
>
> On Fri, Jun 24, 2022 at 12:13:48AM +0900, Tatsuyuki Ishi wrote:
> > v2 includes patch to objcopy and ld -r to discard the .llvm_addrsig
> > section, in order to prevent linkers to get corrupted input now that
> > the SH_LINK heuristic they use is supplied with the proper value.
>
> This patchset wasn't tested, was it?

I did not run the testsuite, and I'm deeply sorry for that. I'll make
sure to run the full test suite on all platforms for any following
revisions as well as any patches I send from now on.

> It's not hard to see why.  Targets add their own symbols, some emit
> section symbols, some don't.

Let me know if there is a way to test this that works across architectures.
I suppose I can provide the expected output for each target by
duplicating this test in each target's folder?
Otherwise, the only thing I can think of right now is to teach readelf
to parse addrsig, but that adds new code which also needs to be tested
somehow.

> Even more embarrassing is that the expected output you put in
> addrsig.d is not correct for x86_64.  You ought to be generating
> a .llvm_addrsig section of 01020305 on x86_64, I think.

You're right. It seems that the way I numbered the symbols wasn't
quite right. I'll get it fixed by the next revision.

---
Tatsuyuki