mbox series

[v7,0/7] x86: suffix handling changes

Message ID dc39887b-fc6e-2827-e0b1-e99949e08796@suse.com
Headers
Series x86: suffix handling changes |

Message

Jan Beulich Dec. 2, 2022, 10:13 a.m. UTC 
  ... accompanied by a few other improvements (or so I hope) found
along the way.

Apart from re-basing this is merely a re-submission, in an attempt to
revive the discussion: There are multiple bugs being fixed by this
series, which I'd like to ask to not put off simply with a subjective
qualification of them being minor. Plus even if minor, it's one thing
to not invest time into addressing a minor bug, but here a fix is
being supplied, ready for committing. I'd like to also stress again
that I've taken quite a bit of time to address concerns, and to hence
find a viable compromise.

1: constify parse_insn()'s input
2: re-work insn/suffix recognition
3: don't recognize/derive Q suffix in the common case
4: allow HLE store of accumulator to absolute 32-bit address
5: move bad-use-of-TLS-reloc check
6: drop (now) stray IsString
7: further re-work insn/suffix recognition to also cover MOVSX

Note that patches 3...5 were previously approved (albeit 3 has now
changed because of the re-ordering of the series in v5), but can't be
committed ahead of the earlier ones.

While I don't think there's any collision with the also pending "x86:
break gas dependency on libopcodes" series (apart from generated files),
just in case: The series here is presented on top of that one, but could
be easily re-based ahead.

Jan

Comments

Jan Beulich Dec. 9, 2022, 10:56 a.m. UTC | #1 
On 02.12.2022 11:13, Jan Beulich via Binutils wrote:
> ... accompanied by a few other improvements (or so I hope) found
> along the way.
> 
> Apart from re-basing this is merely a re-submission, in an attempt to
> revive the discussion: There are multiple bugs being fixed by this
> series, which I'd like to ask to not put off simply with a subjective
> qualification of them being minor. Plus even if minor, it's one thing
> to not invest time into addressing a minor bug, but here a fix is
> being supplied, ready for committing. I'd like to also stress again
> that I've taken quite a bit of time to address concerns, and to hence
> find a viable compromise.
> 
> 1: constify parse_insn()'s input
> 2: re-work insn/suffix recognition
> 3: don't recognize/derive Q suffix in the common case
> 4: allow HLE store of accumulator to absolute 32-bit address
> 5: move bad-use-of-TLS-reloc check
> 6: drop (now) stray IsString
> 7: further re-work insn/suffix recognition to also cover MOVSX

In the absence of alternative proposals on how to address the several
bugs addressed within this series I'm intending to commit this perhaps
early next week, ...

> Note that patches 3...5 were previously approved (albeit 3 has now
> changed because of the re-ordering of the series in v5), but can't be
> committed ahead of the earlier ones.
> 
> While I don't think there's any collision with the also pending "x86:
> break gas dependency on libopcodes" series (apart from generated files),
> just in case: The series here is presented on top of that one, but could
> be easily re-based ahead.

... on top of this one (and finally unblocking further work I have
pending on top).

Jan