[llvm] r229352 - IR: Properly return nullptr when getAggregateElement is out-of-bounds

[Tom Stellard]

On Mon, Mar 16, 2015 at 07:40:19AM -0700, David Majnemer wrote:
> I'm pretty sure these two functions aren't virtual.
> 

OK, does that mean adding them anywhere will break the ABI or does it mean they
are safe to add?

-Tom

> On Monday, March 16, 2015, Tom Stellard <tom at stellard.net> wrote:
> 
> > On Sat, Mar 14, 2015 at 02:52:25PM +0100, Björn Steinbrink wrote:
> > > Hi Tom, Hi David,
> > >
> > > On 2015.02.16 04:02:10 -0000, David Majnemer wrote:
> > > > Author: majnemer
> > > > Date: Sun Feb 15 22:02:09 2015
> > > > New Revision: 229352
> > > >
> > > > URL: http://llvm.org/viewvc/llvm-project?rev=229352&view=rev
> > > > Log:
> > > > IR: Properly return nullptr when getAggregateElement is out-of-bounds
> > > >
> > > > We didn't properly handle the out-of-bounds case for
> > > > ConstantAggregateZero and UndefValue.  This would manifest as a crash
> > > > when the constant folder was asked to fold a load of a constant global
> > > > whose struct type has no operands.
> > > >
> > > > This fixes PR22595.
> > >
> > > Could this be added to 3.5.2?
> > >
> >
> > This adds a new function to two difference classes, which can
> > break the ABI in some cases.  According to abi-compliance-checker:
> >
> >
> > "You can add virtual functions at end of leaf classes with exported
> >  constructors."
> >
> > I have a buildbot setup to check abi compliance, so if you move these
> > functions to the end of the classes, I will approve merging it, but if
> > this is not enough to ensure ABI compliance, I will revert it before the
> > release.
> >
> > -Tom
> >
> > > Cheers
> > > Björn
> >