This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: tweaks for autoconf 2.50


On Thu, May 24, 2001 at 01:17:19AM -0700, Zack Weinberg wrote:
> On Thu, May 24, 2001 at 04:39:15AM -0300, Alexandre Oliva wrote:
> > Huh?  AC_PROG_CC tests the bootstrap compiler in the GCC directory.
> > The problem occurs when testing the compiler for the target.
> 
> Oh, now I get it.
> 
> Revised patch tomorrow morning, I need to go sleep now.

It's been almost four days.  Is that enough sleep?  :-)

For the record, for entertainment, and for the morbidly curious, here's what
happens with 2.50 when trying to regenerate the configury for libstdc++.
With an unmodified acinclude.m4 from current CVS (trunk or branch, same
in this respect):

    78% aclocal
    79% autoconf
    configure.in:47: error: defn: undefined: _AC_PROG_CC_GNU
    ./aclocal.m4:157: GLIBCPP_CONFIGURE is expanded from...
    configure.in:47: the top level
    80%

These statements beginning on line 24 probably have something to do with it:

  # Never versions of autoconf add an underscore to these functions.
  # Prevent future problems ...
  ifdef([AC_PROG_CC_G],[],[define([AC_PROG_CC_G],defn([_AC_PROG_CC_G]))])
  ifdef([AC_PROG_CC_GNU],[],[define([AC_PROG_CC_GNU],defn([_AC_PROG_CC_GNU]))])
  ifdef([AC_PROG_CXX_G],[],[define([AC_PROG_CXX_G],defn([_AC_PROG_CXX_G]))])
  ifdef([AC_PROG_CXX_GNU],[],[define([AC_PROG_CXX_GNU],defn([_AC_PROG_CXX_GNU]))

(I thought defn was supposed to define a name; why is it complaining about
that name being undefined?  Of course it's undefined, that's why we're
calling defn on it... agh need advil.)

Removing these four lines causes a different error message from autoconf.
Trying to make defn() look like what the manual says causes an error
message different from the first two error messages.  (Just because I love
m4 doesn't mean I don't also hate it.)

Applying Zack's initial patch for autoconf 2.50 (by hand to acinclude.m4
rather than aclocal.m4) causes a shell error ("unexpected token 'fi'")
due to the AC_EXEEXT trick a few lines further down from his change.
Commenting those three lines lets the configure script run, but can't
build anything since @gnugccCXX@ (or something like that) isn't AC_SUBST'd
anymore.  That bit get removed by Zack's patch, but is used elsewhere.
I stopped there.


Should I reinstall 2.13?  (Crap, that won't work... should I reinstall
whatever 2.14+changes interim version we had going?  if I can find the
tarball?)  I'm assuming yes, since these changes won't go in for 3.0.


Phil

-- 
pedwards at disaster dot jaj dot com  |  pme at sources dot redhat dot com
devphil at several other less interesting addresses in various dot domains
The gods do not protect fools.  Fools are protected by more capable fools.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]