Partial inlining

[Jan Hubicka]

> On 06/27/2010 10:14 PM, Paolo Carlini wrote:
> > On 06/27/2010 10:02 PM, Jan Hubicka wrote:
> >> Paolo: Hope that the issues are solved now, but if you will still see
> >> libstdc++
> >> breakage due to the partial inlining, you can disable it with
> >> -fno-partial-inlining (or by patching out the flag_partial_inlining set in
> >> opts.c).
> >>   
> > Thanks. Normally I regtest on x86_64-linux -m64 and I can confirm that
> > for that specific target things are back to normality. If I notice
> > something useful for -m32, I'll let you know immediately.
> 
> As reported already by others (see HJ's testresults) unfortunately I can
> also confirm that -m32 is still broken: abi_check fails immediately but
> the setup of the machine is fine, it's simply miscompiled. As for the
> glibc version, nothing special, a 2.10.1.

I just posted fix for NRV problem that at least has the property that the testcase
reproduces on 32 bit compilation only.
Does the problem reproduce for you with --test-board=unit,-m32 on 64bit build?

Honza
> 
> Paolo.