'make bootstrap' failure on m88k-dg-dgux5.4R3.10

'make bootstrap' failure on m88k-dg-dgux5.4R3.10

[jepler]

Hi.

I'm not sure if m88k-dg-dgux5.4R3.10 is supposed to be a target supported
by the latest egcs snapshot (or any other egcs snapshot for that matter),
but I just tried to boostrap gcc and g++ with the 971105 snapshot and got
the following problem right at the beginning of the second stage compile:

make CC="stage1/xgcc -Bstage1/" CFLAGS="-O2 -g -O2" LDFLAGS=""
libdir=/usr/local/lib STAGE_PREFIX=stage1/ LANGUAGES="c proto gcov  c++"
make[2]: Entering directory `/work/egcs-971105/gcc'
stage1/xgcc -Bstage1/  -DIN_GCC  -O -mstandard -mlegend  -O2 -g -O2
-DHAVE_CONFIG_H     -I. -I. -I./config \
  -DGCC_INCLUDE_DIR=\"/usr/local/lib/gcc-lib/m88k-dg-dgux5.4R3.10/egcs-2.90.16/include\"
\
  -DGPLUSPLUS_INCLUDE_DIR=\"/usr/local/include/g++\" \
  -DOLD_GPLUSPLUS_INCLUDE_DIR=\"/usr/local/lib/g++-include\" \
  -DLOCAL_INCLUDE_DIR=\"/usr/local/include\" \
  -DCROSS_INCLUDE_DIR=\"/usr/local/lib/gcc-lib/m88k-dg-dgux5.4R3.10/egcs-2.90.16/sys-include\"
\
  -DTOOL_INCLUDE_DIR=\"/usr/local/m88k-dg-dgux5.4R3.10/include\" \
  -c `echo ./cccp.c | sed 's,^\./,,'`
as: Multiple input files specified, ',-lansi-c' and ',-keep-std'
make[2]: *** [cccp.o] Error 1
make[2]: Leaving directory `/work/egcs-971105/gcc'
make[1]: *** [bootstrap] Error 2
make[1]: Leaving directory `/work/egcs-971105/gcc'
make: *** [bootstrap] Error 2

The stage1 compiler was produced by "gcc version 2.7.2", and "./configure"
was given no extra flags.

I have made the assumption that gas/gnu ld aren't required to build this
port, since they aren't required in 2.7.2 and I was unable to build some
2.8ish binutils a few months ago when I tried.

I modified this commandline to use -v, and this is the output:

Reading specs from stage1/specs
gcc version egcs-2.90.16 971105 (gcc2-970802 experimental)
 stage1/cpp -lang-c -v -I. -I. -I./config -isystem include -isystem stage1/include -undef -D__GNUC__=2 -D__GNUC_MINOR__=90 -Dm88000 -Dm88k -Dunix -DDGUX -D__CLASSIFY_TYPE__=2 -D__svr4__ -D__m88000__ -D__m88k__ -D__unix__ -D__DGUX__ -D__CLASSIFY_TYPE__=2 -D__svr4__ -D__m88000 -D__m88k -D__unix -D__DGUX -Asystem(unix) -Acpu(m88k) -Amachine(m88k) -D__OPTIMIZE__ -g -D__OPEN_NAMESPACE__ -D_DGUX_TARGET -DIN_GCC -DHAVE_CONFIG_H -DGCC_INCLUDE_DIR="/usr/local/lib/gcc-lib/m88k-dg-dgux5.4R3.10/egcs-2.90.16/include" -DGPLUSPLUS_INCLUDE_DIR="/usr/local/include/g++" -DOLD_GPLUSPLUS_INCLUDE_DIR="/usr/local/lib/g++-include" -DLOCAL_INCLUDE_DIR="/usr/local/include" -DCROSS_INCLUDE_DIR="/usr/local/lib/gcc-lib/m88k-dg-dgux5.4R3.10/egcs-2.90.16/sys-include" -DTOOL_INCLUDE_DIR="/usr/local/m88k-dg-dgux5.4R3.10/include" cccp.c /tmp/cca000m-.i
GNU CPP version egcs-2.90.16 971105 (gcc2-970802 experimental) (m88k, $Revision: 1.1.1.1 $)
#include "..." search starts here:
#include <...> search starts here:
 .
 config
 include
 /usr/include
End of search list.
 stage1/cc1 /tmp/cca000m-.i -quiet -dumpbase cccp.c -mstandard -mlegend -g -O -O2 -O2 -version -o /tmp/cca000m-.s
GNU C version egcs-2.90.16 971105 (gcc2-970802 experimental) (m88k-dg-dgux5.4R3.10) compiled by GNU C version 2.7.2.
 as -V -Wc,-fix-bb,-s"cccp.c" ,-lansi-c ,-keep-std -o cccp.o /tmp/cca000m-.s
as: Multiple input files specified, ',-lansi-c' and ',-keep-std'

Does someone have a clue for me?

Jeff
-- 
Jeff Epler (just a student -- ignore my opinion, please)

Re: 'make bootstrap' failure on m88k-dg-dgux5.4R3.10

[Manfred Hollstein]

On Thu, 6 November 1997, 14:14:59, jepler@inetnebr.com wrote:

    Hi.
    
    I'm not sure if m88k-dg-dgux5.4R3.10 is supposed to be a target supported
    by the latest egcs snapshot (or any other egcs snapshot for that matter),
    but I just tried to boostrap gcc and g++ with the 971105 snapshot and got
    the following problem right at the beginning of the second stage compile:
    
    [snip]

    I modified this commandline to use -v, and this is the output:
    
    Reading specs from stage1/specs
    gcc version egcs-2.90.16 971105 (gcc2-970802 experimental)
     stage1/cpp -lang-c -v -I. -I. -I./config -isystem include -isystem stage1/include -undef -D__GNUC__=2 -D__GNUC_MINOR__=90 -Dm88000 -Dm88k -Dunix -DDGUX -D__CLASSIFY_TYPE__=2 -D__svr4__ -D__m88000__ -D__m88k__ -D__unix__ -D__DGUX__ -D__CLASSIFY_TYPE__=2 -D__svr4__ -D__m88000 -D__m88k -D__unix -D__DGUX -Asystem(unix) -Acpu(m88k) -Amachine(m88k) -D__OPTIMIZE__ -g -D__OPEN_NAMESPACE__ -D_DGUX_TARGET -DIN_GCC -DHAVE_CONFIG_H -DGCC_INCLUDE_DIR="/usr/local/lib/gcc-lib/m88k-dg-dgux5.4R3.10/egcs-2.90.16/include" -DGPLUSPLUS_INCLUDE_DIR="/usr/local/include/g++" -DOLD_GPLUSPLUS_INCLUDE_DIR="/usr/local/lib/g++-include" -DLOCAL_INCLUDE_DIR="/usr/local/include" -DCROSS_INCLUDE_DIR="/usr/local/lib/gcc-lib/m88k-dg-dgux5.4R3.10/egcs-2.90.16/sys-include" -DTOOL_INCLUDE_DIR="/usr/local/m88k-dg-dgux5.4R3.10/include" cccp.c /tmp/cca000m-.i
    GNU CPP version egcs-2.90.16 971105 (gcc2-970802 experimental) (m88k, $Revision: 1.1.1.1 $)
    #include "..." search starts here:
    #include <...> search starts here:
     .
     config
     include
     /usr/include
    End of search list.
     stage1/cc1 /tmp/cca000m-.i -quiet -dumpbase cccp.c -mstandard -mlegend -g -O -O2 -O2 -version -o /tmp/cca000m-.s
    GNU C version egcs-2.90.16 971105 (gcc2-970802 experimental) (m88k-dg-dgux5.4R3.10) compiled by GNU C version 2.7.2.
     as -V -Wc,-fix-bb,-s"cccp.c" ,-lansi-c ,-keep-std -o cccp.o /tmp/cca000m-.s
    as: Multiple input files specified, ',-lansi-c' and ',-keep-std'
    
    Does someone have a clue for me?

Can you please check this patch?

Sat Nov  8 11:37:49  Manfred Hollstein  (manfred@s-direktnet.de)

	* dgux.h (ASM_CPU_SPEC): Reformatted to suppress wrong whitespace in
	generated `specs' file.
    
*** egcs-971105.orig/gcc/config/m88k/dgux.h	Mon Aug 11 17:57:19 1997
--- egcs-971105/gcc/config/m88k/dgux.h	Sat Nov  8 11:32:55 1997
*************** Boston, MA 02111-1307, USA.  */
*** 137,149 ****
  #define ASM_CPU_SPEC "\
  		  %{v:-V}\
  		  %{g:\
! 		  %{mno-legend:-Wc,off}\
! 		  %{!mno-legend:-Wc,-fix-bb,-s\"%i\"\
! 		  %{traditional:,-lc}\
! 		  %{!traditional:,-lansi-c}\
! 		  %{mstandard:,-keep-std}\
! 		  %{mexternal-legend:,-external}\
! 		  %{mocs-frame-position:,-ocs}}}"
  
  #define CPP_CPU_SPEC "\
                    %{!m88000:%{!m88100:%{m88110:-D__m88110__}}} \
--- 137,149 ----
  #define ASM_CPU_SPEC "\
  		  %{v:-V}\
  		  %{g:\
! %{mno-legend:-Wc,off}\
! %{!mno-legend:-Wc,-fix-bb,-s\"%i\"\
! %{traditional:,-lc}\
! %{!traditional:,-lansi-c}\
! %{mstandard:,-keep-std}\
! %{mexternal-legend:,-external}\
! %{mocs-frame-position:,-ocs}}}"
  
  #define CPP_CPU_SPEC "\
                    %{!m88000:%{!m88100:%{m88110:-D__m88110__}}} \

-- 
 Manfred Hollstein       If you have any questions about GNU software:
 Hindenburgstr. 13/1                   < mailto:manfred@s-direktnet.de >
 75446 Wiernsheim, FRG  < http://www.s-direktnet.de/HomePages/manfred/ >
 PGP key:    < http://www.s-direktnet.de/HomePages/manfred/manfred.asc >

Re: 'make bootstrap' failure on m88k-dg-dgux5.4R3.10

[Jeffrey A Law]

  In message < 199711081040.LAA08343@saturn.s-direktnet.de >you write:
  > Can you please check this patch?
  > 
  > Sat Nov  8 11:37:49  Manfred Hollstein  (manfred@s-direktnet.de)
  > 
  > 	* dgux.h (ASM_CPU_SPEC): Reformatted to suppress wrong whitespace in
  > 	generated `specs' file.
Installed.  Thanks.

jeff

Re: 'make bootstrap' failure on m88k-dg-dgux5.4R3.10

[Jeff Epler]

[-- Attachment #1: Type: text/plain, Size: 2412 bytes --]

> On Thu, 6 November 1997, 14:14:59, jepler@inetnebr.com wrote:
> 
>     Hi.
>     
>     I'm not sure if m88k-dg-dgux5.4R3.10 is supposed to be a target supported
>     by the latest egcs snapshot (or any other egcs snapshot for that matter),
>     but I just tried to boostrap gcc and g++ with the 971105 snapshot and got
>     the following problem right at the beginning of the second stage compile:
>     
>     [snip]
> 
On Sat, Nov 08, 1997 at 11:40:52AM +0100, Manfred Hollstein wrote:
> Can you please check this patch?
> 
> Sat Nov  8 11:37:49  Manfred Hollstein  (manfred@s-direktnet.de)
> 
> 	* dgux.h (ASM_CPU_SPEC): Reformatted to suppress wrong whitespace in
> 	generated `specs' file.
[...]

Thanks for the patch.  This gets things further, but I still get an
error on tree.o:

%make CC='stage1/xgcc' CFLAGS='-Bstage1/ -g -save-temps -O' tree.o
stage1/xgcc -c  -DIN_GCC  -O -mstandard -mlegend  -Bstage1/ -g -save-temps -O  -DHAVE_CONFIG_H     -I. -I. -I./config tree.c
tree.c: In function `build':
tree.c:2887: parse error before `tree'
tree.c:2887: parse error before `*'
tree.c:2888: parse error before `tree'
tree.c:2888: parse error before `*'
tree.c:2899: parse error before `tree'
tree.c:2899: parse error before `*'
tree.c:2913: parse error before `tree'
tree.c:2913: parse error before `*'
tree.c: In function `build_nt':
tree.c:3006: parse error before `tree'
tree.c:3006: parse error before `tree'
tree.c:3006: parse error before `tree'
tree.c:3006: parse error before `tree'
tree.c: In function `build_parse_node':
tree.c:3039: parse error before `tree'
tree.c:3039: parse error before `tree'
tree.c:3039: parse error before `tree'
tree.c:3039: parse error before `tree'
make: *** [tree.o] Error 1

It looks like there's braindamage in the varargs stuff, but again I'm
lost without much knowledge.

Line 2887 has expanded into the following:
      register tree arg0 = ( ( p ).__va_arg = ((( p ).__va_arg + (1 << (__alignof__(  tree ) >> 3)) - 1) & ~((1 << (__alignof__(  tree ) >> 3)) - 1)) + __va_size(  tree ), *((  tree  *) ((__va_reg_p(  tree ) && ( p ).__va_arg < 8 + __va_size(  tree ) ? ( p ).__va_reg : ( p).__va_stk) + (( p ).__va_arg - __va_size(  tree ))))) ;

and the rest are similar, see the attached (compressed) tree.i file.

Jeff
-- 
\/ jepler@inetnebr.com http://incolor.inetnebr.com/jepler/ (0|1(01*0)*1)+

Two is not equal to three, even for large values of two.

[-- Attachment #2: tree.i.gz --]
[-- Type: application/x-gzip, Size: 29000 bytes --]

Re: 'make bootstrap' failure on m88k-dg-dgux5.4R3.10

[Manfred Hollstein]

On Fri, 14 November 1997, 07:49:07, jepler@inetnebr.com wrote:

    [snip]

    Thanks for the patch.  This gets things further, but I still get an
    error on tree.o:
    
    %make CC='stage1/xgcc' CFLAGS='-Bstage1/ -g -save-temps -O' tree.o
    stage1/xgcc -c  -DIN_GCC  -O -mstandard -mlegend  -Bstage1/ -g -save-temps -O  -DHAVE_CONFIG_H     -I. -I. -I./config tree.c
    tree.c: In function `build':
    tree.c:2887: parse error before `tree'
    [snip]
    
    It looks like there's braindamage in the varargs stuff, but again I'm
    lost without much knowledge.
    
    Line 2887 has expanded into the following:
          register tree arg0 = ( ( p ).__va_arg = ((( p ).__va_arg + (1 << (__alignof__(  tree ) >> 3)) - 1) & ~((1 << (__alignof__(  tree ) >> 3)) - 1)) + __va_size(  tree ), *((  tree  *) ((__va_reg_p(  tree ) && ( p ).__va_arg < 8 + __va_size(  tree ) ? ( p ).__va_reg : ( p).__va_stk) + (( p ).__va_arg - __va_size(  tree ))))) ;
    
    and the rest are similar, see the attached (compressed) tree.i file.
    
Looking at your tree.i  attachment, there must be something  seriously
wrong with your   configuration: #include <stdarg.h> from  tree.c uses
/usr/include/stdarg.h instead of the fixed  one from gcc/include!  How
did you build egcs upto this error?

Jeff, can you please send the following:

- directory listing of the fixed include dir (i.e. ls -lR gcc/include)
- contents of:
  1. gcc/config.log
  2. gcc/config.h
  3. gcc/config2.h

Another question, what  actually is `m88k-dg-dgux5.4R3.10'? Is it UNIX
Sys V R3 or R4?

Having that information, we can look further.
-- 
 Manfred Hollstein       If you have any questions about GNU software:
 Hindenburgstr. 13/1                   < mailto:manfred@s-direktnet.de >
 75446 Wiernsheim, FRG  < http://www.s-direktnet.de/HomePages/manfred/ >
 PGP key:    < http://www.s-direktnet.de/HomePages/manfred/manfred.asc >

Re: 'make bootstrap' failure on m88k-dg-dgux5.4R3.10

[Jeff Epler]

[-- Attachment #1: Type: text/plain, Size: 2380 bytes --]

> On Fri, 14 November 1997, 07:49:07, jepler@inetnebr.com wrote:
>     Thanks for the patch.  This gets things further, but I still get an
>     error on tree.o:
[snip]
>     and the rest are similar, see the attached (compressed) tree.i file.
>     
On Sat, Nov 15, 1997 at 02:01:02PM +0100, Manfred Hollstein wrote:
> Looking at your tree.i  attachment, there must be something  seriously
> wrong with your   configuration: #include <stdarg.h> from  tree.c uses
> /usr/include/stdarg.h instead of the fixed  one from gcc/include!  How
> did you build egcs upto this error?

Starting from egcs-{core,g++}-971105, I applied the earlier dgux patch
and did the following (I'm doing it again just to make sure the same
things happen):
	% cd egcs-971105
	% make distclean
	% ./configure
	% make bootstrap
[wait wait wait -- same error, eventually]
	
> 
> Jeff, can you please send the following:
> 
> - directory listing of the fixed include dir (i.e. ls -lR gcc/include)
> - contents of:
>   1. gcc/config.log
>   2. gcc/config.h
>   3. gcc/config2.h

These files should be attached to the message.

The directory looks like this:
	% ls -lR gcc/include
	total 31
	-rw-r--r--   1 jepler   general      495 Nov 15 14:58 README
	-rw-r--r--   1 jepler   general      548 Nov 15 14:58 _int_varargs.h
	-rw-r--r--   1 jepler   general      772 Nov 15 14:58 exception
	-rw-r--r--   1 jepler   general     3304 Nov 15 15:02 float.h
	-rw-r--r--   1 jepler   general     3207 Nov 15 14:58 limits.h
	-rw-r--r--   1 jepler   general     1168 Nov 15 14:58 new
	-rw-r--r--   1 jepler   general      174 Nov 15 14:58 new.h
	-rw-r--r--   1 jepler   general      330 Nov 15 14:58 syslimits.h
	-rw-r--r--   1 jepler   general     1244 Nov 15 14:58 typeinfo

> Another question, what  actually is `m88k-dg-dgux5.4R3.10'? Is it UNIX
> Sys V R3 or R4?

It's supposed to be "posix"(ish), and variors #defines seem to be able to
select R3 or R4 behavior.  The default seems to be R4ish behavior,
looking at /usr/include/sys/_int_features.h.

> Having that information, we can look further.

Well, looks like no stdarg.h is built ... For grins, I'll include a
copy of the /usr/include/stdarg.h from the current system.

Jeff
-- 
\/ jepler@inetnebr.com http://grail.cnri.reston.va.us/grail/ (0|1(01*0)*1)+

The primary requisite for any new tax law is for it to exempt enough
voters to win the next election.

[-- Attachment #2: gcc-dgux-files.tar.gz --]
[-- Type: application/x-gzip, Size: 1480 bytes --]

missing iostream, fstream iomanip, etc

[Tudor Hulubei]

Hi,

I've noticed that egcs doesn't come with the <iostream>, <fstream>,
and <iomanip> header files.  Instead, I have to include <iostream.h>,
<fstream.h>, etc.  Is this a known issue?

Thanks,
Tudor