Deploy PIC32MX gcc toolchain on Solaris [tio-wiki]

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
====== Deploy PIC32MX gcc toolchain on Solaris ======
  * The toolchain **binutils-2.36**, **gcc-10.3.0**, **newlib-1.4** no longer works after upgrade from **SRU62** to **SRU63**, since **libisl** was upgraded from version **15** to version **23**. Re-compilation with libisl 23 fixed the problem.
  * The toolchain no longer works on s11u4sru37 due to missing /usr/lib/libmpc.so.2 and /usr/lib/libmpfr.so.1 libraries - /opt/mipsel/libexec/gcc/mipsel/4.5.2/cc1 depends on them. Reason being the libraries were upgraded to /usr/lib/limpc.so.3 and /usr/lib/libmpfr.so.4 respectively.
To workaround the issue, missing libraries were copied from s11u4sru30 where both old as well as new versions of those libraries are available.
Also, new tool chain has been built based on binutils 2.36, gcc 10.3.0, newlib 4.1.0.
===== binutils 2.36 =====
  * start building new one
<code>
$ export PATH=/usr/gnu/bin:$PATH
$ mkdir ~/projects/gcc-mips-10.3.0/
$ cd ~/projects/gcc-mips-10.3.0/
$ wget http://ftp.gnu.org/gnu/binutils/binutils-2.36.tar.gz -O /tmp/binutils-2.36.tar.gz
$ tar xvpf /tmp/binutils-2.36.tar.gz
$ mkdir build-binutils-el-2.36
$ cd build-binutils-el-2.36
$ ../binutils-2.36/configure --target=mipsel --prefix=/opt/mipsel
$ make
$ sudo make install
</code>

===== binutils 2.37 =====
  * man pages were not built - instead use binutils 2.36

  * save existing toolchain and prepare for building new one
<code>
$ mv ~/projects/gcc-mips/ ~/projects/gcc-mips-ok/
# mv /data/apps-local/mipsel /data/apps-local/mipsel-ok
# mkdir /data/apps-local/mipsel
</code>
  * start building new one
<code>
$ export PATH=/usr/gnu/bin:$PATH
$ mkdir ~/projects/gcc-mips/
$ cd ~/projects/gcc-mips/
$ wget http://ftp.gnu.org/gnu/binutils/binutils-2.37.tar.gz -O /tmp/binutils-2.37.tar.gz
$ tar xvpf /tmp/binutils-2.37.tar.gz
$ mkdir build-binutils-el
$ cd build-binutils-el
$ ../binutils-2.37/configure --target=mipsel --prefix=/opt/mipsel
$ make
$ sudo make install
</code>
===== gcc 10.3.0 with newlib 4.1.0 =====
  * prepare working area
<code>
$ cd ~/projects/gcc-mips-10.3.0/
$ mkdir build-gcc-el-10.3.0
$ mkdir build-newlib-el-4.1.0
</code>
  * download and unpack
<code>
$ wget http://ftp.gnu.org/gnu/gcc/gcc-10.3.0/gcc-10.3.0.tar.gz -O /tmp/gcc-10.3.0.tar.gz
$ tar xvpf /tmp/gcc-10.3.0.tar.gz
$ wget ftp://sourceware.org/pub/newlib/newlib-4.1.0.tar.gz -O /tmp/newlib-4.1.0.tar.gz
$ tar xvpf /tmp/newlib-4.1.0.tar.gz
</code>
  * build and install (tools will be installed into /opt/mipsel/ directory). Building gcc with newlib is done in three steps.
===== Stage 1 - build bootstrap gcc =====
<code>
$ cd build-gcc-el-10.3.0
$ ../gcc-10.3.0/configure --target=mipsel --prefix=/opt/mipsel-10.3.0 --without-headers --with-gnu-as --with-gnu-ld --with-newlib --enable-languages=c --with-local-prefix=/opt/mipsel --with-native-system-header-dir=/opt/mipsel/include/ --disable-nls --disable-shared  --enable-multilib --disable-decimal-float -disable-threads --disable-libmudflap --disable-libssp --disable-libgomp --disable-libquadmath
$ make
$ sudo make install
</code>
  * **--enable-multilib** - builds several variants of libraries (big-endian, little-endian hard-float and soft-float). Default is little-endian hard-float.
<code>
$ /opt/mipsel-10.3.0/bin/mipsel-gcc -print-multi-lib
.;
soft-float;@msoft-float
eb;@EB
soft-float/eb;@msoft-float@EB
$
</code>
===== Stage 2 - deploy newlib =====
  * newlib requires mips-cc
<code>
# cd /opt/mipsel-10.3.0/bin
# ln -s mipsel-gcc mipsel-cc
</code>
  * build and install
<code>$ export PATH=/opt/mipsel-10.3.0/bin:$PATH
$ cd ~/projects/gcc-mips-10.3.0/build-newlib-el-4.1.0
$ ../newlib-4.1.0/configure --target=mipsel --prefix=/opt/mipsel-10.3.0
$ make ; echo $?
...
mipsel-cc -B/home/dambi/projects/gcc-mips/build-newlib-el/mipsel/newlib/ -isystem /home/dambi/projects/gcc-mips/build[27/1919]
l/mipsel/newlib/targ-include -isystem /home/dambi/projects/gcc-mips/newlib-4.1.0/newlib/libc/include -B/home/dambi/projects/gcc-mips/build-newlib-el/mipsel/libgloss/mips -L/home/dambi/projects/gcc-mips/build-newlib-el/mipsel/libgloss/libnosys -L/home/d
ambi/projects/gcc-mips/newlib-4.1.0/libgloss/mips    -DPACKAGE_NAME=\"newlib\" -DPACKAGE_TARNAME=\"newlib\" -DPACKAGE_VERSION=
\"4.1.0\" -DPACKAGE_STRING=\"newlib\ 4.1.0\" -DPACKAGE_BUGREPORT=\"\" -DPACKAGE_URL=\"\" -I. -I../../../../../newlib-4.1.0/new
lib/libc/stdlib -fno-builtin -DMISSING_SYSCALL_NAMES -DHAVE_INIT_FINI      -g -O2 -c -o lib_a-dtoa.o `test -f 'dtoa.c' || echo '../../../../../newlib-4.1.0/newlib/libc/stdlib/'`dtoa.c
In file included from ../../../../../newlib-4.1.0/newlib/libc/stdlib/mprec.h:29,
                 from ../../../../../newlib-4.1.0/newlib/libc/stdlib/dtoa.c:33:
/home/dambi/projects/gcc-mips/newlib-4.1.0/newlib/libc/include/ieeefp.h:152:2: error: #error "double and long double are the s
ame size but LDBL_EQ_DBL is not defined"
  152 | #error "double and long double are the same size but LDBL_EQ_DBL is not defined"
      |  ^~~~~
/home/dambi/projects/gcc-mips/newlib-4.1.0/newlib/libc/include/ieeefp.h:195:3: error: unknown type name '__ieee_ext_field_type
'
  195 |   __ieee_ext_field_type  ext_fracl : EXT_FRACLBITS;
      |   ^~~~~~~~~~~~~~~~~~~~~
/home/dambi/projects/gcc-mips/newlib-4.1.0/newlib/libc/include/ieeefp.h:195:38: error: 'EXT_FRACLBITS' undeclared here (not in
 a function); did you mean 'EXT_FRACBITS'?
  195 |   __ieee_ext_field_type  ext_fracl : EXT_FRACLBITS;
      |                                      ^~~~~~~~~~~~~
      |                                      EXT_FRACBITS
/home/dambi/projects/gcc-mips/newlib-4.1.0/newlib/libc/include/ieeefp.h:195:26: error: bit-field 'ext_fracl' width not an inte
ger constant
  195 |   __ieee_ext_field_type  ext_fracl : EXT_FRACLBITS;
      |                          ^~~~~~~~~
/home/dambi/projects/gcc-mips/newlib-4.1.0/newlib/libc/include/ieeefp.h:196:3: error: unknown type name '__ieee_ext_field_type
'
  196 |   __ieee_ext_field_type  ext_frach : EXT_FRACHBITS;
      |   ^~~~~~~~~~~~~~~~~~~~~
/home/dambi/projects/gcc-mips/newlib-4.1.0/newlib/libc/include/ieeefp.h:196:38: error: 'EXT_FRACHBITS' undeclared here (not in
 a function); did you mean 'EXT_FRACBITS'?
  196 |   __ieee_ext_field_type  ext_frach : EXT_FRACHBITS;
      |                                      ^~~~~~~~~~~~~
      |                                      EXT_FRACBITS
/home/dambi/projects/gcc-mips/newlib-4.1.0/newlib/libc/include/ieeefp.h:196:26: error: bit-field 'ext_frach' width not an inte
ger constant
  196 |   __ieee_ext_field_type  ext_frach : EXT_FRACHBITS;
      |                          ^~~~~~~~~
/home/dambi/projects/gcc-mips/newlib-4.1.0/newlib/libc/include/ieeefp.h:197:3: error: unknown type name '__ieee_ext_field_type
'
  197 |   __ieee_ext_field_type  ext_exp   : EXT_EXPBITS;
      |   ^~~~~~~~~~~~~~~~~~~~~
/home/dambi/projects/gcc-mips/newlib-4.1.0/newlib/libc/include/ieeefp.h:197:38: error: 'EXT_EXPBITS' undeclared here (not in a
 function); did you mean 'EXT_EXP_BIAS'?
  197 |   __ieee_ext_field_type  ext_exp   : EXT_EXPBITS;
      |                                      ^~~~~~~~~~~
      |                                      EXT_EXP_BIAS
/home/dambi/projects/gcc-mips/newlib-4.1.0/newlib/libc/include/ieeefp.h:197:26: error: bit-field 'ext_exp' width not an integer constant
  197 |   __ieee_ext_field_type  ext_exp   : EXT_EXPBITS;
      |                          ^~~~~~~
/home/dambi/projects/gcc-mips/newlib-4.1.0/newlib/libc/include/ieeefp.h:198:3: error: unknown type name '__ieee_ext_field_type'
  198 |   __ieee_ext_field_type  ext_sign  : 1;
      |   ^~~~~~~~~~~~~~~~~~~~~
make[5]: *** [Makefile:734: lib_a-dtoa.o] Error 1
make[5]: Leaving directory '/data/smb/users/dambi/projects/gcc-mips/build-newlib-el/mipsel/newlib/libc/stdlib'
make[4]: *** [Makefile:683: all-recursive] Error 1
make[4]: Leaving directory '/data/smb/users/dambi/projects/gcc-mips/build-newlib-el/mipsel/newlib/libc'
make[3]: *** [Makefile:641: all-recursive] Error 1
make[3]: Leaving directory '/data/smb/users/dambi/projects/gcc-mips/build-newlib-el/mipsel/newlib'
make[2]: *** [Makefile:452: all] Error 2
make[2]: Leaving directory '/data/smb/users/dambi/projects/gcc-mips/build-newlib-el/mipsel/newlib'
make[1]: *** [Makefile:8492: all-target-newlib] Error 2
make[1]: Leaving directory '/data/smb/users/dambi/projects/gcc-mips/build-newlib-el'
make: *** [Makefile:879: all] Error 2
2
</code>

* **LDBL_EQ_DBL** not defined in some header files
<code>
$ pwd
~/projects/gcc-mips-10.3.0/build-newlib-el-4.1.0
$ ggrep -Rl LDBL_EQ ./
./mipsel/newlib/targ-include/newlib.h
./mipsel/newlib/newlib.h
./mipsel/soft-float/newlib/targ-include/newlib.h
./mipsel/soft-float/newlib/config.log
./mipsel/soft-float/newlib/config.status
./mipsel/soft-float/newlib/newlib.h
./mipsel/soft-float/eb/newlib/newlib.h
./mipsel/soft-float/eb/newlib/targ-include/newlib.h
./mipsel/soft-float/eb/newlib/config.log
./mipsel/soft-float/eb/newlib/config.status
./mipsel/eb/newlib/config.log
./mipsel/eb/newlib/targ-include/newlib.h
./mipsel/eb/newlib/newlib.h
./mipsel/eb/newlib/config.status
</code>
* define **LDBL_EQ_DBL** in files where it is not defined
<code>
$ vim -p ./mipsel/newlib/targ-include/newlib.h ./mipsel/newlib/newlib.h
$ make ; echo $?
$ sudo make install
</code>

===== Stage 3 - deploy gcc and g++ with newlib suport =====
<code>
$ cd ~/projects/gcc-mips-10.3.0/build-gcc-el-10.3.0
$ ../gcc-10.3.0/configure --target=mipsel --prefix=/opt/mipsel-10.3.0 --with-gnu-as --with-gnu-ld --with-newlib --enable-languages=c,c++ --with-local-prefix=/opt/mipsel --with-native-system-header-dir=/opt/mipsel/include --disable-nls --disable-shared  --enable-multilib --disable-decimal-float -disable-threads --disable-libmudflap --disable-libssp --disable-libgomp --disable-libquadmath
$ make -j 10 ; echo $?
$ sudo make install
</code>
===== binutils 2.22 =====
  * prepare working area
<code>
$ mkdir ~/projects/gcc-mips/
$ cd ~/projects/gcc-mips/
$ mkdir build-binutils-el
</code>
  * download and unpack
<code>
$ wget http://ftp.gnu.org/gnu/binutils/binutils-2.22.tar.gz -O /tmp/binutils-2.22.tar.gz
$ tar xvpf /tmp/binutils-2.22.tar.gz
</code>
  * build and install (all tools will be installed into /opt/mipsel/ directory)
<code>
$ cd build-binutils-el
$ ../binutils-2.22/configure --target=mipsel --prefix=/opt/mipsel
$ make
$ sudo make install
</code>
===== gcc 4.5.2 with newlib 1.20.0 =====
  * prepare working area
<code>
$ cd ~/projects/gcc-mips/
$ mkdir build-gcc-el
$ mkdir build-newlib-el
</code>
  * download and unpack
<code>
$ wget http://ftp.gnu.org/gnu/gcc/gcc-4.5.2/gcc-4.5.2.tar.gz -O /tmp/gcc-4.5.2.tar.gz
$ tar xvpf /tmp/gcc-4.5.2.tar.gz
$ wget ftp://sources.redhat.com/pub/newlib/newlib-1.20.0.tar.gz -O /tmp/newlib-1.20.0.tar.gz
$ tar xvpf /tmp/newlib-1.20.0.tar.gz
</code>
  * build and install (tools will be installed into /opt/mipsel/ directory). Building gcc with newlib is done in three steps.
===== Stage 1 - build bootstrap gcc =====
<code>
$ cd build-gcc-el
$ ../gcc-4.5.2/configure --target=mipsel --prefix=/opt/mipsel --with-gmp-include=/usr/include/gmp --with-mpfr-include=/usr/include/mpfr --without-headers --with-gnu-as --with-gnu-ld --with-newlib --enable-languages=c --with-local-prefix=/opt/mipsel --with-native-system-header-dir=/opt/mipsel/include/ --disable-nls --disable-shared  --enable-multilib --disable-decimal-float -disable-threads --disable-libmudflap --disable-libssp --disable-libgomp --disable-libquadmath
$ make
$ sudo make install
</code>
  * **--enable-multilib** - builds several variants of libraries (big-endian, little-endian hard-float and soft-float). Default is little-endian hard-float.
<code>
$ /opt/mipsel/bin/mipsel-gcc -print-multi-lib
.;
soft-float;@msoft-float
eb;@EB
soft-float/eb;@msoft-float@EB
$
</code>
===== Stage 2 - deploy newlib =====
  * build and install
<code>
$ cd ~/projects/gcc-mips/build-newlib-el
$ ../newlib-1.20.0/configure --target=mipsel --prefix=/opt/mipsel
$ make
$ sudo make install
</code>
===== Stage 3 - deploy gcc with newlib suport =====
<code>
$ cd ~/projects/gcc-mips/build-gcc-el
$ ../gcc-4.5.2/configure --target=mipsel --prefix=/opt/mipsel --with-gmp-include=/usr/include/gmp --with-mpfr-include=/usr/include/mpfr --with-gnu-as --with-gnu-ld --with-newlib --enable-languages=c --with-local-prefix=/opt/mipsel --with-native-system-header-dir=/opt/mipsel/include --disable-nls --disable-shared  --enable-multilib --disable-decimal-float -disable-threads --disable-libmudflap --disable-libssp --disable-libgomp --disable-libquadmath
$ make
$ sudo make install
</code>