[Bug bootstrap/7548] gcc3.1.1 can't be bootstraped on HPUX 10.20 using HP compilers

[martin dot jost at icn dot siemens dot de]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=7548


martin dot jost at icn dot siemens dot de changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|3.1.1                       |3.3


------- Additional Comments From martin dot jost at icn dot siemens dot de  2003-07-23 13:34 -------
Hello,

I rechecked, using gcc 3.3. [I changed the version number in the
report accordingly]

These are my findings on HPUX 10.20, using HPs ANSI-C for bootstrapping.
I configured as advised in a separate directory using
export CFLAGS="+DAportable -Ae +O2"
../gcc-3.3/configure -prefix=/pd/gcc3.3 --without-gcc --with-gnu-as --with-
as=/pd/binutils/bin/as --without-gnu-ld --with-local-prefix=/pd/gcc3.3/include -
-enable-languages="c, c++"


First I get some errors from configure:

checking assembler --gstabs support... yes
checking linker read-only and read-write section mixing... conftest1.s: 
Assembler messages:
conftest1.s:1: Warning: rest of line ignored; first ignored character is `"'
conftest2.s: Assembler messages:
conftest2.s:1: Warning: rest of line ignored; first ignored character is `"'
conftest2.s:2: Warning: rest of line ignored; first ignored character is `1'
conftest3.s: Assembler messages:
conftest3.s:1: Warning: rest of line ignored; first ignored character is `"'
conftest3.s:2: Warning: rest of line ignored; first ignored character is `0'
/usr/ccs/bin/ld: (Warning) Can't hide symbol "ared"
/usr/ccs/bin/ld: No $UNWIND_END$ subspace has been defined (probably 
missing /usr/ccs/lib/crt0.o)
unknown
checking linker PT_GNU_EH_FRAME support... no
------------------------------------------
I didn't try to track those down. Suggestions ?

Then I do "make bootstrap"
This yields:
[...]
cc -Wp,-H256000 -c   -g  -DIN_GCC     -DHAVE_CONFIG_H    -I. -I. -I../../gcc-
3.3/gcc -I../../gcc-3.3/gcc/. -I../../gcc-3.3/gcc/config -I../../gcc-
3.3/gcc/../include ../../gcc-3.3/gcc/ggc-page.c -o ggc-page.o
cc: "../../gcc-3.3/gcc/ggc-page.c", line 191: error 1709: Long double is an 
ANSI feature.
gmake[2]: *** [ggc-page.o] Error 1

This  is caused  by '-Ae"  (In  the CFLAGS  env-var during  configure)
not showing up in the compiler-flags.
Adding -Ae to STAGE1_CFLAGS in gcc/Makefile solves this problem.

Funnily enough, if I look at some compiles sooner during the build, I find -Ae, 
e.g.:
cc -Wp,-H256000 -c -DHAVE_CONFIG_H +DAportable -Ae +O2 -I. -I../../gcc-
3.3/libiberty/../include   ../../gcc-3.3/libiberty/dyn-string.c -o dyn-string.o

But adding -Ae to the CFLAGS in Top-level-Makefile will "make
bootstrap" fail for later stages !
(configure fails, because "compiler can't produce executable !")

So later on I have to remove it again and do a new 'make bootstrap'.
Is there a "correct" way to supply the flags specific to the stage-1
compiler to configure, so they get propagated correctly ?

Other than that the compile and link was successful.

I'm still fighting with tha java-part. (and thus fastjar). I will
update the report, as soon as I get reasonable results.

BTW:
I got some problems when trying to _use_ the resulting g++/stdlib. It
gets upset about missing ios_base::streamoff.
I found PR 7219, which suggest to rebuild with
CXXFLAGS='-D_GLIBCPP_DEPRECATED'. This solved the problem.
Could configure please add this if it finds gcc being compiled on and
for HPUX 10.20 ?!