Re: [gentoo-user] -flto set for simh-3.9.0-r1 let gcc fail on my Beaglebone Black

meino.cramer@×××.de <meino.cramer@×××.de> [13-08-09 18:32]:
> Adam Carter <adamcarter3@×××××.com> [13-08-09 17:27]:
> > > I already tried do set new CFLAGS via /etc/portage/env/nolto.conf:
> > >
> > >     beagleboneblack:portage/package.env>cat simh
> > >     app-emulation/simh no-lto.conf
> > >
> > > and and an according file in /etc/porttage/package.env:
> > >
> > >     beagleboneblack:portage/package.env>cat simh
> > >     app-emulation/simh no-lto.conf
> > >
> > > but this does not have any effect.
> > >
> >
> > What's the contents of no-lto.conf ? Was the missing - a typo?
> >
> > FWIW - i have a lot of build failures with lto and gcc-4.7.3, but many of
> > these work with lto and gcc-4.6.3
>
> Hi Adam,
>
> Here is the contents of both files:
>
>     beagleboneblack:portage/package.env>cat simh
>     app-emulation/simh no-lto.conf
>     beagleboneblack:portage/env>cat no-lto.conf
>     CFLAGS="${CFLAGS} -fno-lto -fno-use-linker-plugin"
>     CXXFLAGS="${CXXFLAGS} -fno-lto -fno-use-linker-plugin"
>     LDFLAGS="${LDFLAGS} -fno-lto -fno-use-linker-plugin"
>
> The (only) gcc installed on my beaglebone black is:
>
>     beagleboneblack:portage/env>gcc-config -l
>     [1] armv7a-hardfloat-linux-gnueabi-4.6.3 *
>
> The USE-flag configuration of the gcc is:
>     [I] sys-devel/gcc
>         Available versions:
>         (2.95)	[M]~*2.95.3-r10^s
>         (3.1)	[M]*3.1.1-r2
>         (3.2)	**3.2.2^s *3.2.3-r4
>         (3.3)	~*3.3.6-r1^s
>         (3.4)	3.4.6-r2^s
>         (4.0)	~*4.0.4^s
>         (4.1)	4.1.2^s
>         (4.2)	~4.2.4-r1^s
>         (4.3)	~4.3.3-r2^s 4.3.4^s ~4.3.5^s 4.3.6-r1^s
>         (4.4)	~4.4.2^s 4.4.3-r3^s 4.4.4-r2^s 4.4.5^s 4.4.6-r1^s 4.4.7^s
>         (4.5)	~4.5.1-r1^s ~4.5.2^s 4.5.3-r2^s 4.5.4^s
>         (4.6)	~4.6.0^s ~4.6.1-r1^s ~4.6.2^s 4.6.3^s ~4.6.4^s
>         (4.7)	~4.7.0^s ~4.7.1^s ~4.7.2-r1^s ~4.7.3^s
>         (4.8)	[M]**4.8.0^s [M]**4.8.1^s
>         {{altivec bootstrap boundschecking build cxx d doc fixed-point fortran gcj go graphite gtk hardened ip28 ip32r10k java libssp lto mudflap multilib multislot n32 n64 nls nopie nossp nptl objc objc++ objc-gc openmp regression-test static vanilla}}
>         Installed versions:  4.6.3(4.6)^s(05:38:07 08/03/13)(cxx fortran mudflap nls nptl openmp -altivec -doc -fixed-point -gcj -graphite -gtk -hardened -libssp -lto -multilib -multislot -nopie -nossp -objc -objc++ -objc-gc -regression-test -vanilla)
>         Homepage:            http://gcc.gnu.org/
>         Description:         The GNU Compiler Collection
>
> That is, there is no LTO-support compiled in the gcc (as mentioned in
> previous mail). As I think to have understood from searching and
> reading the web, LTO on the ARM platform is possible. But I am no
> gcc guru and therefore it is really possible that I have misunderstand
> what has been written... ;)
>
> Best regards,
> mcc
>
>
>
>


I have digged a little deeper...

The gcc which comes with the prepared stage3 image for the Gentoo on
Beaglebone/Black includes gcc 4.6.3.
As shown above the USE flag "lto" is disabled.

Nonetheless gcc reports "-flto" when called with

    gcc --help=optimizers


reports:
-flto-report                Report various link-time optimization statistics

This output i "grepped" by simh's own makefile to detect whether gcc
can be called with "-flto" or not.
It decides to use -flto while compiling and therefore the process
breaks with 

    cc1: error: LTO support has not been enabled in this configuration

The previous version of simh - which does not seem to use this feature at all --
compiels fine on the beaglebone.

For me I need to know, whether it is safe to recompile gcc natively on
the beaglebone with lto support enabled via USE flag.
This compilation will take about eight hours.
So it would be fine to decide on a little bit more than try and error. ;)

Or how can I circumvent the described problems?

Best regards,
mcc