configure.in patch for --enable-intermodule

[Kelley Cook]

On cygwin --enable-intermodule has never done anything.

Checking config.status I realized that onestep was never being set.
Probably a shell portability issue, since the current code does work
under Linux.

Anyway, here is a small patch which makes the configure.in intermodule
switch setting more like the other switches.

Verified by checking before and after patch of the generated
gcc/Makefile after a "configure [xxx];make configure-gcc" where [xxx] was

a) {blank}
b) --enable-intermodule
c) --enable-intermodule=yes
d) --enable-intermodule=no
e) --disable-intermodule

Variations of gcc/Makefile on Linux were identical before and after
patch ***

All variations on Cygwin never had onestep set before the patch; after
the patch onestep was correctly set for choices b and c.

For posterity sake bootstrapped with --enable-intermodule on a 4GB
2x2.4Ghz Xeon i686-pc-linux-gnu with make -j 4

In addition, it is in stage3 of bootstrap on a 256Meg 1.8Ghz P4
i686-pc-cygwin.  It must be noted that it is definately not a good idea
to --enable-intermodule on a 256Meg box.  It went swap crazy and took
4050 seconds of real time and over 1.25G of virtual memory to compile
libbackend.o and that was *after* disabling GC since the first attempt
lasted over 4 hours.

Anyway, OK to install?

Kelley Cook

*** Technically, they were identical after applying a soon to be posted
obvious followup patch.

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: intmod.diff
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20030807/176a332e/attachment.ksh>