x86: Do not dump DS/CS segment overrides for branch hints

[Borislav Petkov]

On Wed, Dec 02, 2020 at 05:49:55PM +1030, Alan Modra wrote:
> i386-bsd  +FAIL: i386 branch

Ok, found the first one - that's a --target=i386-bsd build AFAICT.

The first failure I can trigger here is:

FAIL: gas/i386/align-branch-9

Debug output from the test run says:

...
regexp "^  1e:  39 c5                   cmp    %eax,%ebp$"
line   "  1e:   39 c5                   cmp    %eax,%ebp"
regexp "^  20:  70 62                   jo     84 <foo\+0x84>$"
line   "  20:   70 62                   jo     84 <.L_2>"
regexp_diff match failure

because the test has:

  cmp  %eax, %ebp
  jo  .L_2

it disassembles to

  1e:   39 c5                   cmp    %eax,%ebp
  20:   70 62                   jo     84 <.L_2>

but the .d file has

  1e:	39 c5                	cmp    %eax,%ebp
  20:	70 62                	jo     84 <foo\+0x84>

i.e., it has that local label .L_2 instead of the offset foo+0x84.

Now I've checked out master before my patch:

9f1212394fa2 (HEAD -> refs/heads/master, refs/remotes/origin/master, refs/remotes/origin/HEAD) Fix Value.format_string docu for static members argument
96fb90861a4e Don't delete the locator win info
632ee6fd822f x86: Do not dump DS/CS segment overrides for branch hints

<--- HERE

568cbddc710f gold: Convert x86-64 GOTPCRELX only if addend == -4

and it fails the same way.

So I'm thinking for this particular case, I should simply fix
align-branch-9.d to match the local label

	[0-9]+ <.L_2>

because we show local labels now in .o disassembly.

Yes?

> i386-darwin  +FAIL: i386 branch
> i386-darwin  +FAIL: x86-64 branch
> i386-msdos  +FAIL: i386 branch
> x86_64-w64-mingw32  +FAIL: x86-64 branch

I'll analyze the rest later.

Thx.

-- 
Regards/Gruss,
    Boris.

SUSE Software Solutions Germany GmbH, GF: Felix Imendörffer, HRB 36809, AG Nürnberg