[Bug ipa/61555] [4.9/4.10 Regression] LLVM build failure

[Bug ipa/61555] [4.9/4.10 Regression] LLVM build failure

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61555

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

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

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Are you sure this is not a LLVM source issue? 

llvm::cl::parser<llvm::PassInfo const*>::getOption(unsigned int) const
 is called by llvm::cl::C<llvm::D>::getExtraOptionNames(int&).

llvm::cl::C<llvm::D>::getExtraOptionNames(int&) is part of vtable for
llvm::cl::C<llvm::D>.


I don't see anything wrong here really as the type of Parser can found at
compile time and it is D inherits from cl::parser<const PassInfo *>.
So getExtraOptionNames calls Parser.getExtraOptionNames which in turns calls
this->getOption (0) which is a virtual function call.  Since we know the type
of *this here, we call directly llvm::cl::parser<llvm::PassInfo
const*>::getOption(unsigned int) const.

So the devirtuatization seems correct and it is just a LLVM source problem.

[Bug ipa/61555] [4.9/4.10 Regression] LLVM build failure

[trippels at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61555

--- Comment #2 from Markus Trippelsdorf <trippels at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #1)
> Are you sure this is not a LLVM source issue? 
> 
> llvm::cl::parser<llvm::PassInfo const*>::getOption(unsigned int) const
>  is called by llvm::cl::C<llvm::D>::getExtraOptionNames(int&).
> 
> llvm::cl::C<llvm::D>::getExtraOptionNames(int&) is part of vtable for
> llvm::cl::C<llvm::D>.
> 
> 
> I don't see anything wrong here really as the type of Parser can found at
> compile time and it is D inherits from cl::parser<const PassInfo *>.
> So getExtraOptionNames calls Parser.getExtraOptionNames which in turns calls
> this->getOption (0) which is a virtual function call.  Since we know the
> type of *this here, we call directly llvm::cl::parser<llvm::PassInfo
> const*>::getOption(unsigned int) const.
> 
> So the devirtuatization seems correct and it is just a LLVM source problem.

Thanks. I just wasn't 100% sure. LLVM bug:
http://llvm.org/bugs/show_bug.cgi?id=20067

[Bug ipa/61555] [4.9/4.10 Regression] LLVM build failure

[mustrumr97 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61555

Hristo Venev <mustrumr97 at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mustrumr97 at gmail dot com

--- Comment #3 from Hristo Venev <mustrumr97 at gmail dot com> ---
In the LLVM codebase getOption() is defined but still not written in the object
file. The reduced test case is wrong. Why can't I reopen this bug?

[Bug ipa/61555] [4.9/4.10 Regression] LLVM build failure

[rafael.espindola at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61555

Rafael Avila de Espindola <rafael.espindola at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rafael.espindola at gmail dot com

--- Comment #4 from Rafael Avila de Espindola <rafael.espindola at gmail dot com> ---
A valid testcase reduction is at
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61659