Re: 3.1, Sun, sparc, solaris2.5.1 (fwd)

Re: 3.1, Sun, sparc, solaris2.5.1 (fwd)

[Hugh Sasse Staff Elec Eng]

On Tue, 18 Jun 2002, Jan Van Belle wrote:

> Hugh Sasse Staff Elec Eng wrote:
> >
> > On Thu, 13 Jun 2002, Jan Van Belle wrote:
> >
> > > Try to compile the new GCC version with the newest binutils (2.12 ??)
> > >
> > > When configuring GCC, use the options: --with-gnu-ld --with-gnu-as
        [...]
> > I did that, and it made a lot more progress,  However it still blew up
> > with :
> > /usr/ccs/bin/as: "/var/tmp//cctop87s.s", line 2296: error: unknown
> > opcode ".uleb128"
> > /usr/ccs/bin/as: "/var/tmp//cctop87s.s", line 2296: error: statement
> > syntax
> > gmake[3]: *** [java/lang/Class.lo] Error 1
> > gmake[3]: Leaving directory
> > `/apps/hgs/gcc-build/sparc-sun-solaris2.5.1/libjava'
        [...]
> >
> > So was I dreaming about using the correct options?
> >
> > brains hgs 185 %> more config.status
        [...]
> > --host=sparc-sun-solaris2.5.1 --with-gnu-ld --with-gnu-as --norecursion
        [...]
> > Apparently not.  So why is it ignoring the with-gnu-as?
> >
> >         Hugh
> >
>
> Maybe it isn't selecting the right tool. Do a # which as # to find out
> which program is selected. if it is: /usr/local/bin, than you have a
> problem. If it is: /usr/ccs/bin/as, change your PATH variable.

It is the latter.  However, the whole point of --with-gnu-as is to correct that
choice, isn't it?  It only blows up on the java, so I would guess it is only
the java that is wrong.  I now know how to work around this, by fiddling with
$PATH, but am I correct in thinking that I should not have to if the confiure
were working completely correctly?


While looking for evidence in the config files to support this
hypothesis (comparing gcc with java) I found this gem:

brains hgs 210 %> tail config.log
configure:7634: checking assembler dwarf2 debug_line support
configure:7690: checking assembler --gdwarf2 support
configure:7719: checking assembler --gstabs support
configure:7747: checking linker PT_GNU_EH_FRAME support
configure:7762: checking whether linker eh_frame optimizations work properly
  5 .eh_frame_hdr 00000024  000003b4  000003b4  000003b4  2**2
conftest.s: Assembler messages:
conftest.s:54: Internal error, aborting at write.c line 2927 in number_to_chars_bigendian
Please report this bug.
configure:7992: checking whether to enable maintainer-specific portions of Makefiles
brains hgs 211 %> pwd
/apps/hgs/gcc-build/gcc
brains hgs 212 %>

I get a different error on our Solaris2.7 box:


neelix hgs 87 %> tail config.log
configure:7537: checking for assembler offsetable %lo() support
configure:7634: checking assembler dwarf2 debug_line support
configure:7690: checking assembler --gdwarf2 support
configure:7719: checking assembler --gstabs support
configure:7747: checking linker PT_GNU_EH_FRAME support
configure:7762: checking whether linker eh_frame optimizations work properly
  5 .eh_frame_hdr 00000024  000003b4  000003b4  000003b4  2**2
conftest.s: Assembler messages:
conftest.s:54: Fatal error: conftest.s:31: bad return from bfd_install_relocation
configure:7992: checking whether to enable maintainer-specific portions of Makefiles

Both of these could be due to picking up the wrong as.

To save fiddling with $PATH (which may upset something else) does the
configuration procedure check for gas before checking for as -- i.e if I
symlink /usr/locla/bin/gas to /usr/local/bin/as will that fix things?

My path does have /usr/local/bin in it, but it is towards the back:
brains hgs 212 %> echo $PATH
/bin:/usr/bin:/progs/forte6/SUNWspro/bin:/apps/jdk1.2/bin:/usr/ucb:/usr/sbin:/usr/ccs/bin:/etc:/usr/local/bin:/usr/local/mysql/bin:/users/bin:  plus a bunch
of other stuff you're not interested in I expect!


        Thank you
        Hugh
>
> --
> Jan Van Belle
> Jan.Van_Belle@alcatel.be
>

Re: 3.1, Sun, sparc, solaris2.5.1 -- possible solution

[Hugh Sasse Staff Elec Eng]

Earlier today Wed, 19 Jun 2002, I wrote:

        [...]
> Both of these could be due to picking up the wrong as.
>
> To save fiddling with $PATH (which may upset something else) does the
> configuration procedure check for gas before checking for as -- i.e if I
> symlink /usr/locla/bin/gas to /usr/local/bin/as will that fix things?

OK, not generally recommended to followup your own posts, but I have
tested this now, to some extent.
This seems to have fixed things on our (faster) Solaris2.7 system.
I'm going to try this on the Solaris2.5.1 system.  Maybe I'll have the
results on Friday for you.

        Hugh