libtool-patches
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: libtool cvs: AC_PROG_NM -B when build!=host


From: Kevin Ryde
Subject: Re: libtool cvs: AC_PROG_NM -B when build!=host
Date: Sun, 03 Feb 2002 08:47:42 +1000
User-agent: Gnus/5.090006 (Oort Gnus v0.06) Emacs/21.1 (i386-debian-linux-gnu)

I posted this previously, so apologies for the duplication, but I
think the problem still exists.

When doing a cross compile, AC_PROG_NM will fall back on plain "nm" if
it can't find an "${ac_tool_prefix}nm".  However in this case it
doesn't try to add -B or -p to select BSD format.  This is not good
for instance on OSF when doing a "cross compile" from say alpha ev5 to
ev6 and using the ordinary system nm.

The problem can be seen with a minimal configure.in

    AC_PREREQ(2.50)
    AC_INIT(configure.in)
    AC_PROG_CC
    AC_PROG_LIBTOOL

and the usual

    libtoolize
    aclocal
    autoconf

Running on OSF system

    ./configure --host=alphaev5-dec-osf5.1 --host=alphaev6-dec-osf5.1

gives

    ...
    checking for BSD-compatible nm... nm

whereas I hoped it might give "/usr/bin/nm -B" as it does for a
"native" build.

I have to confess to not actually spotting any breakage in libtool
itself from a bad nm, though presumably $global_symbol_pipe etc
wouldn't work too well.

I hoped to have it in the right mode for the benefit of GMP, which
greps nm output for some assembler feature tests, like whether .align
is logarithmic, and a good local label prefix.

I'd like to suggest the change below to add an extra loop testing
first $ac_tool_prefix then an unprefixed nm.  I've left the
indentation unchanged to keep the diff to the pertinent part.

        * libtool.m4 (AC_PROG_NM): Try plain "nm" if "${ac_tool_prefix}nm" is
        not found, for the benefit of cross compiles using the native "nm".

Attachment: libtool.m4.prefixed-nm.diff
Description: Source code patch


reply via email to

[Prev in Thread] Current Thread [Next in Thread]