[Bug c/37865] gfortran build fails in stage 3 bootstrap with --enable-intermodule

[Bug c/37865] gfortran build fails in stage 3 bootstrap with --enable-intermodule

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37865

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |WONTFIX
   Target Milestone|---                         |4.7.0

--- Comment #8 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-01-28 20:49:46 UTC ---
-combine has been removed so closing as won't fix.  The new way is to use LTO.

[Bug c/37865] gfortran build fails in stage 3 bootstrap with --enable-intermodule

[bernard at brenda-arkle dot me dot uk]

------- Comment #7 from bernard at brenda-arkle dot me dot uk  2008-12-18 18:54 -------
Sorry about the bug-number error. 38528 was what I meant.
Andrew Pinski's reply there suggests that --enable-intermodule
is simply doomed anyway.

(a) Since submitting the original bug report I corrected
-march=athlon64-sse3 to -march=athlon64.  The abbreviated
report of the late failure failed to show this, and I forgot
to mention it.

(b) It still doesn't work, even without trying to specify
any particular target architecture with -march.  But that
belongs in 38528, if anywhere.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37865

[Bug c/37865] gfortran build fails in stage 3 bootstrap with --enable-intermodule

[mikael at gcc dot gnu dot org]

------- Comment #6 from mikael at gcc dot gnu dot org  2008-12-15 00:14 -------
(In reply to comment #4)
> Sounds a bit like the problem from comment #0 is a duplicate or related to
> PR31537 and/or PR35034 ?
> 
31537 seems definitely related. 
Is there a specific option/configuration to reproduce/bypass it?

(In reply to comment #5)
> Responding to comment #3: *Ahem*.  I tried with today's live sources
> (revision 142751).  I got as far as building libgfortran in stage 3, and
> *bang*.  Bug 35828, which see.  It's not obviously related to the
> problem here (though presumably there are some options which don't
> produce it, of which --enable-intermodule may perhaps be one).
> 
s/Bug 35828/Bug 38528/
It seems some configure options are not much tested. :-(
Something has just come to my mind. 
Since you are on i686-pc-linux-gnu, I'm not sure it is legal to specify
-march=athlon64-sse3 as it will use the 64 bits extensions.
>From the gcc manual:
    _k8, opteron, athlon64, athlon-fx_
          AMD K8 core based CPUs with x86-64 instruction set support.
          (This supersets MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and
          64-bit instruction set extensions.)

    _k8-sse3, opteron-sse3, athlon64-sse3_
          Improved versions of k8, opteron and athlon64 with SSE3
          instruction set support.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37865

[Bug c/37865] gfortran build fails in stage 3 bootstrap with --enable-intermodule

[bernard at brenda-arkle dot me dot uk]

------- Comment #5 from bernard at brenda-arkle dot me dot uk  2008-12-14 20:30 -------
Responding to comment #3: *Ahem*.  I tried with today's live sources
(revision 142751).  I got as far as building libgfortran in stage 3, and
*bang*.  Bug 35828, which see.  It's not obviously related to the
problem here (though presumably there are some options which don't
produce it, of which --enable-intermodule may perhaps be one).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37865

[Bug c/37865] gfortran build fails in stage 3 bootstrap with --enable-intermodule

[aldot at gcc dot gnu dot org]

------- Comment #4 from aldot at gcc dot gnu dot org  2008-12-14 15:30 -------
Sounds a bit like the problem from comment #0 is a duplicate or related to
PR31537 and/or PR35034 ?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37865

[Bug c/37865] gfortran build fails in stage 3 bootstrap with --enable-intermodule

[mikael at gcc dot gnu dot org]

------- Comment #3 from mikael at gcc dot gnu dot org  2008-12-14 14:46 -------
(In reply to comment #2)
> I'm not sure what 'now' means.  It's still true for release 4.3.2!
> I'm prepared to test with the live sources, but do you want me to?
Sure!

> Do you believe that the problem has been understood and addressed,
Not by me at least... ;-)
> or are you hoping that it's gone away (not for the first time in
> the history of gcc 4.x.x)?
Yes, I have seen some recent bootstrap fixing patches going in. 
So the bug could be fixed now. 
If it isn't, you will have to find the offending configure option(s) :-(. 


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37865

[Bug c/37865] gfortran build fails in stage 3 bootstrap with --enable-intermodule

[bernard at brenda-arkle dot me dot uk]

------- Comment #2 from bernard at brenda-arkle dot me dot uk  2008-12-13 23:53 -------
I'm not sure what 'now' means.  It's still true for release 4.3.2!
I'm prepared to test with the live sources, but do you want me to?
Do you believe that the problem has been understood and addressed,
or are you hoping that it's gone away (not for the first time in
the history of gcc 4.x.x)?

The following configuration options work fine with --disable-intermodule
substituted for --enable-intermodule.

'../gcc-4.3.2/configure' '--prefix=/usr' '--libexecdir=/usr/lib'
'--infodir=/usr/share/info' '--mandir=/usr/share/man' '--enable-libada'
'--enable-libssp' '--disable-werror' '--with-mpfr=/usr' '--with-gmp=/usr'
'--with-datarootdir=/usr/share' '--with-docdir=/usr/share/gcc-4.3.2/doc'
'--with-pdfdir=/usr/share/gcc-4.3.2/doc'
'--with-htmldir=/usr/share/gcc-4.3.2/doc/html' '--disable-coverage'
'--enable-nls' '--enable-__cxa_atexit' '--enable-decimal-float'
'--disable-fixed-point' '--enable-threads=posix' '--enable-clocale=gnu'
'--enable-shared' '--enable-intermodule'
'--enable-languages=ada,c++,fortran,java,objc,obj-c++,treelang,c'
'--with-local-prefix=/usr' '--with-gnu-ld' '--with-demangler-in-ld'
'--with-gnu-as' '--with-system-libunwind' '--with-system-zlib'
'--enable-bootstrap'

Configured exactly as given, the relevant bit of the build output is:

Comparing stages 2 and 3
warning: ./cc1objplus-checksum.o differs
warning: ./cc1-checksum.o differs
warning: ./cc1plus-checksum.o differs
warning: ./cc1obj-checksum.o differs
Bootstrap comparison failure!
./libbackend.o differs

That's with *almost* unmodified sources.  As it happens, as
part of working around a bug which I reported, I've adopted
two patches which have made their way into the live tree:

  (a) the patch for "my" bug 37870: the patch functionally
      changes gcc/expmed.c and adds a new test
  (b) the patch noted at
      http://gcc.gnu.org/ml/gcc-patches/2008-08/msg02201.html
      which I grabbed simply to confirm that it wasn't relevant
      (it also modifies gcc/expmed.c).

I don't believe that these changes are relevant (and in any case,
they make the 4.3.2 source more rather than less like the
current live tree).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37865

[Bug c/37865] gfortran build fails in stage 3 bootstrap with --enable-intermodule

[mikael at gcc dot gnu dot org]

------- Comment #1 from mikael at gcc dot gnu dot org  2008-12-08 22:32 -------
(Sorry for the delay)
Do you still see the problem? 


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37865