[Bug driver/53883] New: GCC 4.7.1 doesn't build on Mac OS X 10.4.11 Tiger/PowerPC (32-bit)

[Bug driver/53883] New: GCC 4.7.1 doesn't build on Mac OS X 10.4.11 Tiger/PowerPC (32-bit)

[dwalker07 at yahoo dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53883

             Bug #: 53883
           Summary: GCC 4.7.1 doesn't build on Mac OS X 10.4.11
                    Tiger/PowerPC (32-bit)
    Classification: Unclassified
           Product: gcc
           Version: 4.7.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: driver
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: dwalker07@yahoo.com


I'm using an August-2002 Apple eMac, with a PowerPC-G4 chip.  GCC is one of the
software packages offered by MacPorts.  I'm on a Tiger, PowerPC, and 32-bit
system; any of those puts me in near-abandoned status, so the fact I'm all 3
meant that their team abandoned me to fix the problem on my own, possibly with
your help.

I'm frustrated since GCC 4.7.0 built fine with some tweaks provided by the
MacPorts team.  From a brief glance at the logs, something about a linking
error with your included version of libFFI came up.  Unlike some earlier GCC
versions on MacPorts, like GCC 4.4 and 4.5, I don't think GCC 4.7 uses any
patched files; issues with LibFFI on earlier versions may have been integrated
into the source, but now I'm not so sure.

Here's how I tried to build it:
1. Use "sudo port edit --editor nano gcc47" to edit the Portfile for GCC-4.7
1a. Comment-out the "supported_archs     i386 x86_64" line to:
    "#supported_archs     i386 x86_64"
    (They "dealt" with the build problems on PowerPC by summarily executing
     the PowerPC build.  Comment out the ban so you can try it.)
2. "sudo port upgrade gcc47 configure.compiler=apple-gcc-4.2"
   (This requires MacPorts' version of Apple-provided GCC 4.2 to be
    installed first.)
3. This ended up with an error.
4. Use "port log gcc47 > gcc47.log" to get the log so I can attach it here

(I don't know how to kick-start the process on a PowerPC system that never had
"gcc47" on it.  For all I know, the MacPorts program may read the
supported-architectures list and ban you from downloading it.)

The worst part is that I can't even go back.  The last version, which should be
GCC 4.7.0, is still installed.  It can print out the version information, but
attempting to compile a program results in a link error with LibPPI.  (But this
is probably more of a MacPorts problem than a GCC one, so don't worry about
it.)  I'm totally screwed for now, except for the copy of GCC 4.6 I have lying
around.

(Should I try building GCC 4.7.1 straight from you guys, without going through
MacPorts?)

[Bug driver/53883] GCC 4.7.1 doesn't build on Mac OS X 10.4.11 Tiger/PowerPC (32-bit)

[dwalker07 at yahoo dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53883

--- Comment #1 from Daryle Walker <dwalker07 at yahoo dot com> 2012-07-07 02:20:51 UTC ---
Created attachment 27756
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27756
The log MacPorts left when trying to build GCC 4.7.1.

[Bug driver/53883] GCC 4.7.1 doesn't build on Mac OS X 10.4.11 Tiger/PowerPC (32-bit)

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53883

--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-07-07 12:46:54 UTC ---
(In reply to comment #0)
> (Should I try building GCC 4.7.1 straight from you guys, without going through
> MacPorts?)

Yes, probably.

[Bug driver/53883] GCC 4.7.1 doesn't build on Mac OS X 10.4.11 Tiger/PowerPC (32-bit), at least with MacPorts

[dwalker07 at yahoo dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53883

--- Comment #3 from Daryle Walker <dwalker07 at yahoo dot com> 2012-07-08 20:45:06 UTC ---
(In reply to comment #2)
> (In reply to comment #0)
> > (Should I try building GCC 4.7.1 straight from you guys, without going through
> > MacPorts?)
> 
> Yes, probably.

I'm working on that now....

[Bug driver/53883] GCC 4.7.1 doesn't build on Mac OS X 10.4.11 Tiger/PowerPC (32-bit), at least with MacPorts

[dwalker07 at yahoo dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53883

--- Comment #4 from Daryle Walker <dwalker07 at yahoo dot com> 2012-07-10 01:32:12 UTC ---
(In reply to comment #3)
> (In reply to comment #2)
> > (In reply to comment #0)
> > > (Should I try building GCC 4.7.1 straight from you guys, without going through
> > > MacPorts?)
> > 
> > Yes, probably.
> 
> I'm working on that now....

...And here's what I tried:

//---
svn co svn://gcc.gnu.org/svn/gcc/tags/gcc_4_7_1_release gcc
cd gcc
nano README
nano contrib/gcc_update
./contrib/gcc_update --touch
cd ..
mkdir gcc-obj
cd gcc-obj
../gcc/configure --with-gmp=/opt/local --with-mpfr=/opt/local
--with-mpc=/opt/local
make
//---

(I'm using GMP, MPFR, and MPC from MacPorts.  Note that it used the system
compiler, Apple's GCC 4.0.1, and not anything from MacPorts for the first
compiler phase.)  It didn't build all the way through; here's the last few
lines:

//---
checking for gawk... awk
checking whether make sets $(MAKE)... yes
checking for --enable-version-specific-runtime-libs... no
checking for style of include used by make... GNU
checking for powerpc-apple-darwin8.11.0-gcc...
/Users/daryle/Documents/Projects/GCC/gcc-obj/./gcc/xgcc
-B/Users/daryle/Documents/Projects/GCC/gcc-obj/./gcc/
-B/usr/local/powerpc-apple-darwin8.11.0/bin/
-B/usr/local/powerpc-apple-darwin8.11.0/lib/ -isystem
/usr/local/powerpc-apple-darwin8.11.0/include -isystem
/usr/local/powerpc-apple-darwin8.11.0/sys-include  -m64
checking for C compiler default output file name... a.out
checking whether the C compiler works... configure: error: in
`/Users/daryle/Documents/Projects/GCC/gcc-obj/powerpc-apple-darwin8.11.0/ppc64/libquadmath':
configure: error: cannot run C compiled programs.
If you meant to cross compile, use `--host'.
See `config.log' for more details.
make[1]: *** [configure-target-libquadmath] Error 1
make: *** [all] Error 2
[daryle:gcc-obj]$ l
//---

It seems that it tried to build "libquadmath" thinking all PowerPC systems are
64-bit, while mine is 32-bit.  I didn't reach the error I reported with
MacPorts' build yet.  Is there any way to force libquadmath to compile in
32-bit mode?  (If that's impossible, is it skippable?)  I'll try to attach the
"config.log" file.

I'll retry make when you guys let me know how to fix up libquadmath.

[Bug driver/53883] GCC 4.7.1 doesn't build on Mac OS X 10.4.11 Tiger/PowerPC (32-bit), at least with MacPorts

[dwalker07 at yahoo dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53883

--- Comment #5 from Daryle Walker <dwalker07 at yahoo dot com> 2012-07-10 01:34:10 UTC ---
Created attachment 27767
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27767
The log from the failed make described in Comment 4

[Bug driver/53883] GCC 4.7.1 doesn't build on Mac OS X 10.4.11 Tiger/PowerPC (32-bit), at least with MacPorts

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53883

--- Comment #6 from Andreas Schwab <schwab@linux-m68k.org> 2012-07-10 06:36:03 UTC ---
*-darwin8 assumes ppc64 multilib, try --disable-multilib.

[Bug driver/53883] GCC 4.7.1 doesn't build on Mac OS X 10.4.11 Tiger/PowerPC (32-bit), at least with MacPorts

[dwalker07 at yahoo dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53883

--- Comment #7 from Daryle Walker <dwalker07 at yahoo dot com> 2012-07-14 18:52:42 UTC ---
(In reply to comment #6)
> *-darwin8 assumes ppc64 multilib, try --disable-multilib.

I added that for my second try:

//===
cd ..
gnutar -cjf gcc-obj.tar.bz2 gcc-obj
rm -r gcc-obj
mkdir gcc-obj
cd gcc-obj
../gcc/configure --with-gmp=/opt/local --with-mpfr=/opt/local
--with-mpc=/opt/local --disable-multilib
make | tee ~/making1.txt
make | tee -a ~/making1.txt
make -k check | tee -a ~/making1.txt
//===

I ran the main "make" twice because I thought it ended by timing out.  But I
think it just succeeded, since the second one was a bunch of no-ops.

The third "make," for testing, is still going on.  So far, I've been getting
all FAIL.  (I don't have the "autogen" program.)  The main compile took over a
day; how long will the testing take?  Should I just quit it and "make install"?
 Or install "autogen" and test again instead?

[Bug driver/53883] GCC 4.7.1 doesn't build on Mac OS X 10.4.11 Tiger/PowerPC (32-bit), at least with MacPorts

[dwalker07 at yahoo dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53883

--- Comment #8 from Daryle Walker <dwalker07 at yahoo dot com> 2012-08-05 16:51:57 UTC ---
I haven't check the results of the testing-make yet, but I tried sending the
results in.

//===
nano your_commentary.txt
../gcc/contrib/test_summary -p your_commentary.txt -m
gcc-testresults@gcc.gnu.org | sh
//===

The second command gave an error of "postdrop: warning: unable to look up
public/pickup: No such file or directory\n" early on.  I left it running for
several days, then Control-C it because I was tired of waiting.  Trying to
command again with a pipe to "more" or to the console had the script
immediately end; I guess the first call worked long enough to purge the test
files.

My "your_commentary.txt" was:

//===
This is my attempt to compile GCC 4.7 directly, instead of using MacPorts,
as part of my bug report 53883.  I'm not sure if it passed or failed.

<http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53883>
//===

I don't know if you guys received anything.

[Bug driver/53883] GCC 4.7.1 doesn't build on Mac OS X 10.4.11 Tiger/PowerPC (32-bit), at least with MacPorts

[dwalker07 at yahoo dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53883

--- Comment #9 from Daryle Walker <dwalker07 at yahoo dot com> 2012-08-05 17:09:00 UTC ---
Created attachment 27945
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27945
Output from running make, for gcc and tests, with tee

[Bug driver/53883] GCC 4.7.1 doesn't build on Mac OS X 10.4.11 Tiger/PowerPC (32-bit), at least with MacPorts

[dwalker07 at yahoo dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53883

--- Comment #10 from Daryle Walker <dwalker07 at yahoo dot com> 2012-08-05 17:13:08 UTC ---
I've added attachment 27945, a bzip2 archive of the "making1.txt" file.  That
file is the "tee" results of the three "make" runs given in comment 7
(http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53883#c7).  The third "make" run
is the testing run, and starts around line 21,938 of the file.