Problems building libstdc++-v3

Mack Lobell
Mon Sep 23 13:42:00 GMT 2002


you were absolutely right when you said that i was missing a target. Now I 
get a little bit further. But i have stumbled a cross another problem. When 
i try to configure powerpc-eabi/und/libiberty the build fails with the same 
output. From the config.log i got the following:

configure:1445: /home/lobell/build/gcc/gcc/xgcc 
-B/home/lobell/build/gcc/gcc/ -B/usr/local/powerpc/powerpc-eabi/bin/ 
-B/usr/local/powerpc/powerpc-eabi/lib/ -isystem 
/usr/local/powerpc/powerpc-eabi/include  -fleading-underscore 
-mrelocatable-lib -mno-eabi -mstrict-align -E  conftest.c >/dev/null 
configure:1536: checking for uintptr_t
configure:2196: checking whether the C compiler 
(/home/lobell/build/gcc/gcc/xgcc -B/home/lobell/build/gcc/gcc/ 
-B/usr/local/powerpc/powerpc-eabi/lib/ -isystem 
/usr/local/powerpc/powerpc-eabi/include  -fleading-underscore 
-mrelocatable-lib -mno-eabi -mstrict-align -myellowknife ) works
configure:2212: /home/lobell/build/gcc/gcc/xgcc 
-B/home/lobell/build/gcc/gcc/ -B/usr/local/powerpc/powerpc-eabi/bin/ 
-B/usr/local/powerpc/powerpc-eabi/lib/ -isystem 
/usr/local/powerpc/powerpc-eabi/include  -fleading-underscore 
-mrelocatable-lib -mno-eabi -mstrict-align -o conftest -myellowknife   
conftest.c  1>&5
/usr/local/powerpc/powerpc-eabi/lib/und/libyk.a(sbrk.o): In function `sbrk':
undefined reference to `__end'
collect2: ld returned 1 exit status
configure: failed program was:

In yellowknife.ld (my target), only _end is defined. Is there a reason for 
compiling with -fleading-underscore in powerpc-eabi/und/libiberty?

ca, le and nof are all configured without the flag.


>From: "Kai Ruottu" <>
>Subject: Re: Problems building libstdc++-v3
>Date: Mon, 23 Sep 2002 17:25:42 +0300
>"Mack Lobell" <> wrote:
> > I'm trying to build a crosscompiler, target=powerpc-eabi and host is 
> > Everything works and installs fine (i.e. binutils, newlib and bootstrap 
> > to the final compiler build. It always fail during configuration of
> > libstdc++-v3, see trace below.
> >
> > checking for object suffix... configure: error: installation or
> > configuration problem; compiler does not work
>  If the error message is this, it means that your "compiler does not
>work"... What else it could mean?
>  Of course the error message is stupid but aimed for newbies who most
>probably think the whole toolchain being the "compiler", not only the
>'gcc + cpp + cc1 + as'-chain for creating objects for libraries, as a
>normal (and more experienced) GCC-user may think...
>  But you wrote: "Everything works and installs fine (i.e. binutils,
>newlib and bootstrap gcc)", so you must have checked that everything
>works and the "compiler" really can create executables, which I think
>the "compiler" here meaning...
>  But as "" wrote:
>------------------- clip --------------------------------------
>  i have finished the build of cross compiler ,when i compiler a
>program using the powerpc-eabi-gcc, there are an error as following:
>  powerpc-eabi-gcc \
>  -I/usr/local/src/gnu/newlib-1.8.2/newlib/libc/include \
>  -o hello hello.c
>  warning: cannot find entry symbol _start; defaulting to 01800074
>/tmp/ccD5t0Jf.o: In function `main':
>/tmp/ccD5t0Jf.o(.text+0x24): undefined reference to `printf'
>------------------- clip --------------------------------------
>a 'powerpc-eabi' targeted GCC doesn't create executables for the
>magical 'eabi' target-board-hardware/monitor-firmware combination,
>or for some operating system called 'eabi', because... believe me
>or not... this 'system' doesn't not exist at all!!!
>  Mack, if would have tries a 'hello.c' program with your new GCC,
>and it would have been compiled and linked ok, then your "Everything
>works" would have had a reason... I would claim that this is not the
>case now...
>  Ok, opening the GCC-manual and seeing the "Invoking GCC / Submodel
>Options / IBM RS6000 and PowerPC options" normally tells what is the
>"problem" : one must select some target from the real world for the
>produced executables: the 'psim'-simulator, one of the supported
>Motorola boards, or something the user has thought to be used as the
>"target system"...
>  The fact seems to be that nowadays one cannot build libstdc++-v3
>without having a working toolchain first, the C-library probably
>will be investigated just as happened with libiberty earlier (if
>not using the '--with-newlib').
>  So the default 'specs'-file must be fixed to have a default target
>for the produced executables, or an extra flag must be added to the
>CFLAGS, for instance the '-mads' (suggested for which kind of
>builders ;-).  I normally suggest using the '-msim', thinking that
>when one has built GDB, the "Hello World" can be seen on the
>screen on the host-platform using the 'powerpc-eabi-run hello'...
>Cheers, Kai

Send and receive Hotmail on your mobile device:

Want more information?  See the CrossGCC FAQ,
Want to unsubscribe? Send a note to

More information about the crossgcc mailing list