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

Problem building the ss20021007

Am Don, 2002-10-24 um 11.26 schrieb Paul Whitfield:
> Ralf Corsepius wrote:
> > Am Don, 2002-10-24 um 10.44 schrieb Paul Whitfield:
> > 
> >>Joe Samuel wrote:
> >>
> >>>Hi Paul,
> >>>    thank you for your valuable advice, but I'm still
> >>>stuck up with something else. Here's the relevant
> >>>build trace:
> >>>../cpukit/score/cpu/unix/cpu.c
> >>>rm -f o-optimize/libscorecpu.a
> >>>no ruv o-optimize/libscorecpu.a o-optimize/cpu.o
> >>>gmake[6]: no: Command not found
> >>
> > 
> >>Do you have a environment variable
> >>called AR set to NO?
> >>
> >>This could be the cause.
> > 
> > No, all this indicates that your PATH is set up wrong and that you are
> > trying to apply a non-functional or bogus toolchain.
> > 
> > Furthermore, you MUST NOT set CC or GCC nor AR, NM or whatever from your
> > environment before running configure - This definitely is wrong!
> Hi Ralf,
> Sorry if my advice was wrong, but certainly I have
> always had to (on older snap-shots, 4-5-0) set either create a
> link or set GCC=cc.
Hmm, the former is true for 4.5.0 under cygwin (It's a deficiency in
autoconf-2.13 causing problems), but I doubt the latter (GCC=cc),
because GCC is an interal autoconf variable used to denote if using GCC
or not. 

The symlink/link/copy gcc->cc however should not be necessary with
rtems-snapshots (Untested, I am not using cygwin.), which uses a newer
autoconf where this autoconf problem is supposed to be solved.

> Before I offered this advice I did confirm that it works!

Apologies if I may have sounded rude, but having to reiterate the same
answers twice a week is a bit annoying.

> As for the second problem I was suggesting that there was an
> INCORRECT environment variable AR that was set to NO!
Well, yes, this indicates a bug somewhere inside of the configuration,
however this actually points at two problems:

1. Why doesn't the configure script find the appropriate ar? 
The cause for this probably is lurking somewhere in Joe's

2. The configuration should abort or complain explicitly if "ar" is not