[Bug c/33396] New: add --enable-intermodule

[Bug c/33396] New: add --enable-intermodule

[aldot at gcc dot gnu dot org]

Add --enable-intermodule for the compilation of various parts of gcc.


-- 
           Summary: add --enable-intermodule
           Product: gcc
           Version: 4.3.0
            Status: UNCONFIRMED
          Severity: enhancement
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: aldot at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33396

[Bug c/33396] add --enable-intermodule

[aldot at gcc dot gnu dot org]

------- Comment #1 from aldot at gcc dot gnu dot org  2007-09-11 16:47 -------
Created an attachment (id=14190)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14190&action=view)
libgcc patch#1

libgcc/ChangeLog:
2007-09-11  Bernhard Fischer  <>

        * configure.ac: Add option --enable-intermodule for onestep compile.
        * Makefile.in: Use onestep.

gcc/ChangeLog:
2007-09-11  Bernhard Fischer  <>

        * gbl-ctors.h: Add header guard.
        * libgcc2.c: Avoid redefinition of FUNC, use FUNCU for __floatundi?f


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33396

[Bug c/33396] add --enable-intermodule

[aldot at gcc dot gnu dot org]

------- Comment #2 from aldot at gcc dot gnu dot org  2007-09-11 16:54 -------
With trunk r127829, this currently gives:

/scratch/obj.i686/gcc-4.3/./prev-gcc/libgcc.a(libgcc_onestep.o): In function
`isinfd128':
../../../../src/gcc-4.3/libgcc/config/libbid/_isinfd128.c:38: undefined
reference to `__bid128_isInf'
collect2: ld returned 1 exit status
make[3]: *** [cc1-dummy] Error 1

The very same thing happens with pristine trunk, where linking doesn't fail,
though.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33396

[Bug c/33396] add --enable-intermodule

[aldot at gcc dot gnu dot org]

------- Comment #3 from aldot at gcc dot gnu dot org  2007-09-11 17:58 -------
(In reply to comment #2)
> With trunk r127829, this currently gives:
> 
> /scratch/obj.i686/gcc-4.3/./prev-gcc/libgcc.a(libgcc_onestep.o): In function
> `isinfd128':
> ../../../../src/gcc-4.3/libgcc/config/libbid/_isinfd128.c:38: undefined
> reference to `__bid128_isInf'
> collect2: ld returned 1 exit status
> make[3]: *** [cc1-dummy] Error 1
> 
> The very same thing happens with pristine trunk, where linking doesn't fail,
> though.
> 

Janis will fix this.

@@ -478,7 +512,7 @@
 ifneq ($(D32PBIT)$(D64PBIT)$(D128PBIT),)
 ifeq ($(enable_decimal_float),bid)
 dfp-filenames += decimal_globals decimal_data binarydecimal \
-                _isinfd32 _isinfd64 _isinfd128 bid64_noncomp \
+                _isinfd32 _isinfd64 _isinfd128 bid64_noncomp bid128_noncomp \
                 bid128_fma bid_round bid_from_int convert_data \
                 bid64_add bid128_add bid64_div bid128_div \
                 bid64_mul bid128_mul bid64_compare bid128_compare \


libgcc/ChangeLog:
2007-09-11  Bernhard Fischer  <>

        * Makefile.in: Compile bid128_noncomp.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33396

[Bug c/33396] add --enable-intermodule

[aldot at gcc dot gnu dot org]

------- Comment #4 from aldot at gcc dot gnu dot org  2007-09-26 18:43 -------

For the fortran frontend, It seems that i have this patch (must be relatively
old; undetermined status, ATM).

Index: gcc/fortran/Make-lang.in
===================================================================
--- gcc/fortran/Make-lang.in    (revision 128688)
+++ gcc/fortran/Make-lang.in    (working copy)
@@ -67,6 +67,8 @@ F95_OBJS = $(F95_PARSER_OBJS) \
     fortran/trans-intrinsic.o fortran/trans-io.o fortran/trans-openmp.o \
     fortran/trans-stmt.o fortran/trans-types.o

+F95_OBJS-onestep = f951_frontend.o
+
 #^L
 # Define the names for selecting gfortran in LANGUAGES.
 fortran: f951$(exeext)
@@ -90,11 +92,15 @@ gfortran-cross$(exeext): gfortran$(exeex
        -rm -f gfortran-cross$(exeext)
        cp gfortran$(exeext) gfortran-cross$(exeext)

+f951_frontend.o: $(F95_OBJS:.o=.c)
+        $(CC) $(ALL_CFLAGS) $(LDFLAGS) -c $(filter %.c,$^) -o $@ -combine
+
 # The compiler itself is called f951.
-f951$(exeext): $(F95_OBJS) \
+# Make-lang.in is not generated via configure, so we cannot use @onestep@
+f951$(exeext): $(F95_OBJS$(onestep)) \
                $(BACKEND) $(LIBDEPS) attribs.o
        $(CC) $(ALL_CFLAGS) $(LDFLAGS) -o $@ \
-               $(F95_OBJS) $(BACKEND) $(LIBS) attribs.o
+               $(F95_OBJS$(onestep)) $(BACKEND) $(LIBS) attribs.o

 gt-fortran-trans.h    : s-gtype; @true
 #^L


-- 

aldot at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|bootstrap                   |c


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33396

[Bug bootstrap/33396] add --enable-intermodule

[aldot at gcc dot gnu dot org]

------- Comment #5 from aldot at gcc dot gnu dot org  2007-09-26 18:53 -------
(In reply to comment #4)
> For the fortran frontend, It seems that i have this patch (must be relatively
> old; undetermined status, ATM).

Scratch that. Testing a working version that i will attach when it passes the
check.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33396

[Bug bootstrap/33396] add --enable-intermodule

[patchapp at dberlin dot org]

------- Comment #6 from patchapp at dberlin dot org  2007-11-20 05:05 -------
Subject: Bug number PR33396

A patch for this bug has been added to the patch tracker.
The mailing list url for the patch is
http://gcc.gnu.org/ml/gcc-patches/2007-11/msg01024.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33396

[Bug bootstrap/33396] add --enable-intermodule

[aldot at gcc dot gnu dot org]

------- Comment #7 from aldot at gcc dot gnu dot org  2008-01-23 14:41 -------
Created an attachment (id=15008)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=15008&action=view)
libgcc patch #2

Updated patch


-- 

aldot at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #14190|0                           |1
        is obsolete|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33396

[Bug bootstrap/33396] add --enable-intermodule

[aldot at gcc dot gnu dot org]

------- Comment #8 from aldot at gcc dot gnu dot org  2008-01-24 13:43 -------
Created an attachment (id=15016)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=15016&action=view)
updated patch

tiny fix: libgcc-s-objects-onestep is only needed for enable_shared=yes.

Two other buglets fixed that were exposed by building libgcc with
--enable-intermodule, see second changelog below.

libgcc/ChangeLog:
2007-11-19  Bernhard Fischer  <>

        * configure.ac: Add option --enable-intermodule for onestep
        compile.
        * Makefile.in: Use onestep.

gcc/ChangeLog:
2007-11-19  Bernhard Fischer  <>

        * gbl-ctors.h: Add header guard.
        * libgcc2.c: Avoid redefinition of FUNC, use FUNCU for
        __floatundi?f
        * config/libbid/bid64_div.c: Extern decls of
        __bid_{convert_table,factors,packed_10000_zeros} have to be const.
        * libgcc/config/libbid/bid128_div.c: Likewise.


-- 

aldot at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #15008|0                           |1
        is obsolete|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33396

[Bug bootstrap/33396] add --enable-intermodule

[aldot at gcc dot gnu dot org]

------- Comment #9 from aldot at gcc dot gnu dot org  2008-01-24 16:15 -------
Stats:

-rw-r--r-- 1 1000 1000 3853392 Jan 24 13:28
./build_i686/gcc-4.3.0-target.IMA/i686-linux-uclibc/libgcc/libgcc.a
2826503   34784       8 2861295  2ba8ef (TOTALS)


-rw-r--r-- 1 1000 1000 4569970 Jan 24 14:00
./build_i686/gcc-4.3.0-target/i686-linux-uclibc/libgcc/libgcc.a
2853491   34800     310 2888601  2c1399 (TOTALS)

So roughly a 15% on-disk size-saving for the static libgcc.

libgcc_s.so:
-rwxr-xr-x 1 1000 1000 40608 Jan 24 15:58
./gcc-4.3.0-target.IMA.so/i686-linux-uclibc/libgcc/libgcc_s.so.1
-rwxr-xr-x 1 1000 1000 187940 Jan 24 15:14
./gcc-4.3.0-target.so/i686-linux-uclibc/libgcc/libgcc_s.so.1

   text    data     bss     dec     hex filename
  11210     284      32   11526    2d06
./gcc-4.3.0-target.IMA.so/i686-linux-uclibc/libgcc/libgcc_s.so.1
  42013     436     336   42785    a721
./gcc-4.3.0-target.so/i686-linux-uclibc/libgcc/libgcc_s.so.1

Approximately 73% size-saving and 78% on-disk savings.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33396

[Bug bootstrap/33396] add --enable-intermodule

[aldot at gcc dot gnu dot org]

------- Comment #10 from aldot at gcc dot gnu dot org  2008-01-29 12:27 -------
Created an attachment (id=15046)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=15046&action=view)
updated patch for libgcc

Fixes an error about {mul,div}{d,x,t}c3 which wants to be built for different
modes / different types;

output is
/scratch/obj.i686/gcc-4.3/i686-linux-gnu/./libgfortran/.libs/libgfortr
an.so: undefined reference to `__divdc3'
/scratch/obj.i686/gcc-4.3/i686-linux-gnu/./libgfortran/.libs/libgfortran.so:
und
efined reference to `__divxc3'
/scratch/obj.i686/gcc-4.3/i686-linux-gnu/./libgfortran/.libs/libgfortran.so:
und
efined reference to `__muldc3'
/scratch/obj.i686/gcc-4.3/i686-linux-gnu/./libgfortran/.libs/libgfortran.so:
und
efined reference to `__mulxc3'
collect2: ld returned 1 exit status

We need to compile for each of the dc, xc, tc variants separately, due to
implementation details (macros). lib2funcs_{d,x,t}c variables and corresponding
pbjects are used for this.


-- 

aldot at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #15016|0                           |1
        is obsolete|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33396

[Bug bootstrap/33396] add --enable-intermodule

[aldot at gcc dot gnu dot org]

------- Comment #11 from aldot at gcc dot gnu dot org  2008-06-11 12:30 -------
Created an attachment (id=15748)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=15748&action=view)
updated patch including unwind and eh and handling of static

includes untested draft for --disable-shared which contains:

# XXX: FIXME: for static libs vis_hide is forced off, so why does the normal
# build above have this superfluous occurence?
# Play safe and provide this wrong rule and retain the mention of vis_hide
# above.


-- 

aldot at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #15046|0                           |1
        is obsolete|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33396

[Bug bootstrap/33396] add --enable-intermodule

[aldot at gcc dot gnu dot org]

------- Comment #12 from aldot at gcc dot gnu dot org  2008-06-19 14:57 -------
Created an attachment (id=15789)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=15789&action=view)
handle libdecnumber

libdecnumber/ChangeLog

      * configure.ac: Add option --enable-intermodule and adjust bid_OBJS.
      * Makefile.in: Handle onestep build.
      (COMPILE-m): New variable to build with --combine.
      * configure: Regenerate.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33396

[Bug bootstrap/33396] add --enable-intermodule

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33396

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
         Resolution|                            |WONTFIX

--- Comment #14 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-09-29 11:13:44 UTC ---
WONTFIX (--enable-intermodule has been removed and is obsoleted by
--with-build-config=bootstrap-lto).

[Bug bootstrap/33396] add --enable-intermodule

[manu at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33396

Manuel López-Ibáñez <manu at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING
   Last reconfirmed|                            |2010.09.29 11:00:18
               date|                            |
                 CC|                            |manu at gcc dot gnu.org
     Ever Confirmed|0                           |1

--- Comment #13 from Manuel López-Ibáñez <manu at gcc dot gnu.org> 2010-09-29 11:00:18 UTC ---
I think we should we close this as WONTFIX. Or are these patches still useful?