>>>>> "Samuel" == Samuel Hornus <Samuel.Hornus@crans.org> writes:

    Samuel> On Sun, 29 Jul 2001, Tsubai Masanari wrote:
    Samuel> So I'll try Markus solution : using previous {base,comp}.tgz to
    Samuel> build userland. Should 1.5.1 .tgz be OK ?

I took them from a more recent -current snapshot I've built myself...

    Samuel> But then I may get the broken gcc-2.95.3 ?

Without 'USE_NEW_TOOLCHAIN' (which didn't work anyway for me) I still
get

  $ gcc -v
  Using builtin specs.
  gcc version egcs-2.91.66 19990314 (egcs-1.1.2 release)

But:

    Aymeric> Although the reason why it dumps core is not the same,
    Aymeric> maybe you're bitten by the following problem: (this is
    Aymeric> from /usr/src/UPDATING)

    Aymeric> 20010718:

    Aymeric>         Enabled correct .init/.fini processing in crt0.
    Aymeric>         The way this was done was to change a -I
    Aymeric>         directive to cc(1), which means make(1) will have
    Aymeric>         a stale dependency (it will be checking the
    Aymeric>         timestamp on the wrong "dot_init.h").

That would be the perfect explanation for me. But if you have a broken
userland already resulting from that (as I remember a lot of userland
programs were broken) the only solution for me was to use an older
userland and re-building from scratch.

    Aymeric> However, you could try to "make dependall && make
    Aymeric> install" in lib/csu and libexec/ld.elf_so before
    Aymeric> attempting a new build.

... if this still works with your broken userland!

If not you could try {base,comp}.tgz from 1.5.1 (I don't know if this
works). Otherwise I could upload you my -current ones tomorrow (mail
me).

Markus.