On Fri, 2019-09-27 at 16:29 +0100, Stuart Henderson wrote:
> On 2019/09/27 11:18, Kurt Miller wrote:
> >
> > On Fri, 2019-09-27 at 16:11 +0100, Stuart Henderson wrote:
> > >
> > > On 2019/09/27 10:35, Kurt Miller wrote:
> > > >
> > > >
> > > > On Fri, 2019-09-27 at 23:50 +1000, Jonathan Gray wrote:
> > > > >
> > > > >
> > > > > On Fri, Sep 27, 2019 at 06:44:24PM +1000, Jonathan Gray wrote:
> > > > > >
> > > > > >
> > > > > >
> > > > > > With U-Boot 2019.10-rc4 the am335x_evm SPL no longer fits when
> > > > > > building
> > > > > > with gcc 6.3 but does fit with 7.4.
> > > > > and with gcc 7.4 arm-trusted-firmware doesn't fit...
> > > > >
> > > > > aarch64-none-elf-ld.bfd: build/rk3399/debug/bl31/bl31.elf section
> > > > > `.pmusram' will not fit in region `PMUSRAM'
> > > > > aarch64-none-elf-ld.bfd: region `PMUSRAM' overflowed by 3928 bytes
> > > > >
> > > > This because CFLAGS passed by ports is overriding -Os
> > > > in the build.
> > > >
> > > > Does the following look ok?
> > > >
> > > > * Disable passing CFLAGS on to build as it overrides -Os and causes
> > > > the build to fail.
> > > > * Add rk3328
> > > >
> > > > Index: Makefile
> > > > ===================================================================
> > > > RCS file: /cvs/ports/sysutils/arm-trusted-firmware/Makefile,v
> > > > retrieving revision 1.11
> > > > diff -u -p -u -r1.11 Makefile
> > > > --- Makefile 12 Jul 2019 20:49:39 -0000 1.11
> > > > +++ Makefile 27 Sep 2019 14:31:37 -0000
> > > > @@ -9,6 +9,7 @@ GH_PROJECT= arm-trusted-firmware
> > > > GH_TAGNAME= v2.1
> > > >
> > > > EPOCH= 0
> > > > +REVISION= 0
> > > >
> > > > CATEGORIES= sysutils
> > > >
> > > > @@ -23,8 +24,10 @@ MAKE_ENV+= BUILD_STRING="${GH_TAGNAME:S/
> > > >
> > > > NO_TEST= Yes
> > > > USE_GMAKE= Yes
> > > > +CFLAGS=
> > > This is non-obvious enough that I'd like to have a comment please.
> > > Or would it be better to make it explicit with CFLAGS=-Os instead?
> > >
> > Checking the build log, I see various optimization levels
> > in use. I think we should allow it to use those levels
> > since this is building a firmware component.
> >
> > We could try rearrange the flags so that CFLAGS from the
> > env comes first if that is desired. I will a comment in
> > if we decide to keep the CFLAGS= in the port make file.
> Oh in that case I'm totally happy to avoid us overriding it,
> CFLAGS= and a comment sounds like the best option them.
>
Ok will do.
-Kurt
