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]
Other format: [Raw text]

Re: [patch] Allow build to specify what sed mkheaders uses


On Thu, 2009-08-27 at 23:46 +0200, Ralf Wildenhues wrote:

> > Ralf, is this expected?
> 
> Is what expected?  That AC_PROG_SED ignores seds that don't cope with
> long command lines?  Yes, that was the primary reason Libtool came up
> with this macro originally.  Is it expected that it discards a $SED that
> was passed in by the user?  I can't yet see where it would do that, but
> that would seem to be a bug.  IIRC, the versions of the macro in Libtool
> did that, but fixincludes/configure isn't using that, and maybe the
> Autoconf 2.59 one, so if you didn't test this with a tree from since
> Sunday, you might want to retry.
> 
> Cheers,
> Ralf

I did a bit more checking, it looks like if I run configure directly in
the fixincludes directory with 'SED=/usr/bin/sed' as an argument on the
configure line then I do use that SED command.  Previously what I was
doing was putting 'SED=/usr/bin/sed' on the top-level GCC configure
command.  It looks like that does not get passed down to the fixincludes
configure script and so fixincludes then does not use /usr/bin/sed.  If
I do 'export SED=/usr/bin/sed' before calling the top-level configure
script, then fixincludes will use /usr/bin/sed.  So the issue may not be
AC_PROG_SED but in how the top-level GCC configure calls
sub-configure's.  Maybe we need to add SED to the list of variables
exported from the top-level GCC Makefile.  I will test that.

Steve Ellcey
sje@cup.hp.com


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