[Bug gold/13245] PREVAILING_DEF reported too often.

[hubicka at ucw dot cz]

http://sourceware.org/bugzilla/show_bug.cgi?id=13245

--- Comment #6 from hubicka at ucw dot cz 2011-10-02 15:22:07 UTC ---
> http://sourceware.org/bugzilla/show_bug.cgi?id=13245
> 
> --- Comment #5 from Octoploid <cryptooctoploid at gmail dot com> 2011-10-02 
> 13:01:38 UTC ---
> What about the following naive patch?
> 
> diff --git a/gold/plugin.cc b/gold/plugin.cc
> index b5880a1..d999254 100644
> --- a/gold/plugin.cc
> +++ b/gold/plugin.cc
> @@ -889,10 +889,10 @@ Pluginobj::get_symbol_resolution_info(int nsyms,
>              res = LDPR_RESOLVED_EXEC;
>            else if (lsym->object()->pluginobj() == this)
>             {
> -             if (is_referenced_from_outside(lsym))
> -               res = LDPR_PREVAILING_DEF;
> -             else if (is_visible_from_outside(lsym))
> +             if (is_visible_from_outside(lsym))
>                 res = ldpr_prevailing_def_ironly_exp;
> +             else if (is_referenced_from_outside(lsym))
> +               res = LDPR_PREVAILING_DEF;

I think the condiitonals would need to be swapped at least. When the symbol is
used from .o file it has
to be LDPR_PREVAILING_DEF no matter if it is exported or not.
> 
> It also survived bootstrap-lto of gcc and an "-flto -fno-fat-lto-objects" 
> build
> of Firefox.

Cool!  If you happen to have a lot of memory (15GB), you may try if it solves
the -fprofile-generate -flto
problem.
In meantime I suceeded building FDO+LTO firefox with -fprofile-generate
-fno-lto train run (that makes more sense
anyway, I am trying to debug the other just to be sure that there are no more
problems in this area)

Honza

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.