[Bug target/66609] New: [sh] Relative address expressions bind at as-time, even if symbol is weak

[Bug target/66609] New: [sh] Relative address expressions bind at as-time, even if symbol is weak

[bugdal at aerifal dot cx]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66609

            Bug ID: 66609
           Summary: [sh] Relative address expressions bind at as-time,
                    even if symbol is weak
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: bugdal at aerifal dot cx
  Target Milestone: ---

Minimal test case (compile with -fPIC and -Os or higher, for sh2 or later):

__attribute__((__weak__,__visibility__("hidden"))) void foo()
{
}

void bar()
{
    foo();
}

For the call from bar to foo, gcc generates:

    mov.l .L3,r1
    braf r1
.LPCS0:
    nop
    .align 2
.L3:
    .long foo-(.LPCS0+2)

Gas in turn ignores the fact that foo is weak and assembles the literal to a
constant with no relocation. This is the following bug I just reported in
binutils:

https://sourceware.org/bugzilla/show_bug.cgi?id=18561

However, GCC can and probably should work around it, by instead generating the
equivalent form that gas assembles correctly:

    .long foo@PCREL-(.LPCS0+2-.)

This is easily changed at the following location:

https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=gcc/config/sh/sh.c;h=e5fcd7683ef7e06d56a10427d792ba018d1af884;hb=HEAD#l1662

and changing it seems to have solved the problem for me, but I did not yet test
extensively. I'm not sure if the subsequent case for UNSPEC_PCREL_SYMOFF also
needs changes.

[Bug target/66609] [sh] Relative address expressions bind at as-time, even if symbol is weak

[kkojima at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66609

Kazumoto Kojima <kkojima at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |kkojima at gcc dot gnu.org

--- Comment #1 from Kazumoto Kojima <kkojima at gcc dot gnu.org> ---
Created attachment 36246
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=36246&action=edit
patch

>    .long foo@PCREL-(.LPCS0+2-.)

looks to be reasonable.  It seems that sh backend shouldn't
use UNSPEC_SYMOFF for that case in the first place.  Does
the patch attached work for you?

[Bug target/66609] [sh] Relative address expressions bind at as-time, even if symbol is weak

[bugdal at aerifal dot cx]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66609

--- Comment #2 from Rich Felker <bugdal at aerifal dot cx> ---
The patch in comment 1 applies successfully to GCC 5.2.0 and fixes both the
test case and the real-world code I was experiencing problems with.
Unfortunately it doesn't apply to some of the older GCC versions I need to
support, but I think I can backport it. Thanks!

[Bug target/66609] [sh] Relative address expressions bind at as-time, even if symbol is weak

[kkojima at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66609

--- Comment #3 from Kazumoto Kojima <kkojima at gcc dot gnu.org> ---
(In reply to Rich Felker from comment #2)

Thanks for the confirmation.  I'll commit the patch after the undergoing
additional test done.

[Bug target/66609] [sh] Relative address expressions bind at as-time, even if symbol is weak

[kkojima at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66609

--- Comment #4 from Kazumoto Kojima <kkojima at gcc dot gnu.org> ---
Author: kkojima
Date: Mon Aug 24 23:23:00 2015
New Revision: 227155

URL: https://gcc.gnu.org/viewcvs?rev=227155&root=gcc&view=rev
Log:
PR target/66609
* [SH] Take into account weak symbols for pc relative calls/sibcalls.


Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config/sh/sh-protos.h
    trunk/gcc/config/sh/sh.c
    trunk/gcc/config/sh/sh.md

[Bug target/66609] [sh] Relative address expressions bind at as-time, even if symbol is weak

[olegendo at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66609

Oleg Endo <olegendo at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |olegendo at gcc dot gnu.org

--- Comment #5 from Oleg Endo <olegendo at gcc dot gnu.org> ---
Kaz, do you think it's OK to backport this to GCC 5?  It looks like the patch
is not so intrusive...

[Bug target/66609] [sh] Relative address expressions bind at as-time, even if symbol is weak

[kkojima at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66609

--- Comment #6 from Kazumoto Kojima <kkojima at gcc dot gnu.org> ---
(In reply to Oleg Endo from comment #5)
> Kaz, do you think it's OK to backport this to GCC 5?  It looks like the
> patch is not so intrusive...

Changing relocation is always intrusive, I think.  I won't
object to backport it, though.  The patch does the right thing
and will affect only very limited cases.

[Bug target/66609] [sh] Relative address expressions bind at as-time, even if symbol is weak

[bugdal at aerifal dot cx]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66609

--- Comment #7 from Rich Felker <bugdal at aerifal dot cx> ---
Created attachment 36359
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=36359&action=edit
preprocessed source still affected by the bug

Oddly I'm still experiencing this bug for some functions but not others, even
with the patch applied. I've attached preprocessed output of a file which is
being miscompiled: the reference to __do_cleanup_push wrongly binds locally,
but the reference to __do_cleanup_pop is overridable like it should be.

[Bug target/66609] [sh] Relative address expressions bind at as-time, even if symbol is weak

[bugdal at aerifal dot cx]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66609

--- Comment #8 from Rich Felker <bugdal at aerifal dot cx> ---
Perhaps hold off on worrying about this; it's only happening with -mfdpic (with
my forward-port of the fdpic patch applied) so it's possible that the bug is on
my end in code that's not in upstream gcc. I'll follow up on this more later.

[Bug target/66609] [sh] Relative address expressions bind at as-time, even if symbol is weak

[bugdal at aerifal dot cx]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66609

--- Comment #9 from Rich Felker <bugdal at aerifal dot cx> ---
Indeed, the fdpic patch I forward-ported introduced new duplicates of some of
the fragments that were changed in sh.md by the above patch. Once I fixed
those, the problem went away. Sorry for the noise.

[Bug target/66609] [sh] Relative address expressions bind at as-time, even if symbol is weak

[olegendo at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66609

Oleg Endo <olegendo at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |FIXED

--- Comment #10 from Oleg Endo <olegendo at gcc dot gnu.org> ---
I think we can close this as fixed.