Re: [PATCH v2] libgcc: m68k: avoid TEXTRELs in shared library (PR 86224)

[Sergei Trofimovich via gcc-patches]

On Mon, 17 Sep 2018 15:29:08 -0600
Jeff Law <l...@redhat.com> wrote:

> On 9/17/18 3:18 PM, Sergei Trofimovich wrote:
> > On Sat, 28 Jul 2018 20:42:02 +0100
> > "slyfox.inbox.ru via gcc-patches" <gcc-patches@gcc.gnu.org> wrote:
> >   
> >> From: Sergei Trofimovich <sly...@gentoo.org>
> >>
> >> Cc: Ian Lance Taylor <i...@airs.com>
> >> Cc: Jeff Law <l...@redhat.com>
> >> Cc: Andreas Schwab <sch...@linux-m68k.org>
> >> Signed-off-by: Sergei Trofimovich <sly...@gentoo.org>
> >> ---
> >>  libgcc/config/m68k/lb1sf68.S | 3 +++
> >>  1 file changed, 3 insertions(+)
> >>
> >> diff --git a/libgcc/config/m68k/lb1sf68.S b/libgcc/config/m68k/lb1sf68.S
> >> index 325a7c17d9b..d5240d4aa55 100644
> >> --- a/libgcc/config/m68k/lb1sf68.S
> >> +++ b/libgcc/config/m68k/lb1sf68.S
> >> @@ -435,6 +435,7 @@ $_exception_handler:
> >>    .text
> >>    FUNC(__mulsi3)
> >>    .globl  SYM (__mulsi3)
> >> +  .hidden SYM (__mulsi3)
> >>  SYM (__mulsi3):
> >>    movew   sp@(4), d0      /* x0 -> d0 */
> >>    muluw   sp@(10), d0     /* x0*y1 */
> >> @@ -458,6 +459,7 @@ SYM (__mulsi3):
> >>    .text
> >>    FUNC(__udivsi3)
> >>    .globl  SYM (__udivsi3)
> >> +  .hidden SYM (__udivsi3)
> >>  SYM (__udivsi3):
> >>  #ifndef __mcoldfire__
> >>    movel   d2, sp@-
> >> @@ -534,6 +536,7 @@ L2:    subql   IMM (1),d4
> >>    .text
> >>    FUNC(__divsi3)
> >>    .globl  SYM (__divsi3)
> >> +  .hidden SYM (__divsi3)
> >>  SYM (__divsi3):
> >>    movel   d2, sp@-
> >>  
> >> -- 
> >> 2.18.0
> >>  
> > 
> > Patch ping. Not sure which is preferable v1 (as other targets do) or v2.  
> Hmm, I thought Andreas NAK'd V1.  And I think Rich's comments
> essentially NAK'd V2.

Aha. I'm trying to clarify the desired state then to try to tweak
patch properly:

Current:
  - static library:
    * global public __divsi3
  - shared library: 
    * global public __divsi3

Desired:
  - static library:
    * global hidden __divsi3 [ABI change:public->hidden]
  - shared library: 
    * global public __divsi3 [no change]
    * global hidden __divsi3_internal [change:new symbol for internal 
references]

Does that sound reasonable?
Should all targets follow the same pattern? I think they don't today.

v1 was a step in desired direction but without user-visible ABI change.

-- 

  Sergei

pgppmyTjSoYAk.pgp
Description: Цифровая подпись OpenPGP