As I just explained in another e-mail, it is very likely
the too long line which breaks the build: the same very long
line occurred on the other machine, but it can cope with it.
It would likely be better to avoid these extreme lines. I can't
say more about it than what I've copied; I could send the
build-log (gzipped 130 kb).
On Sun, Dec 26, 2010 at 01:32:38PM -0500, Raymond Toy wrote:
> On 12/26/10 12:54 PM, Oliver Kullmann wrote:
> > Hi,
> >
> > the build failed on our system (64-bit, AMD, gcc 4.1.2, Linux);
> > previously we built and used successfully 5.21.1, and built successfully
> > 5.22.1 (but didn't use it, due to floating-point precision problems).
>
> You mentioned something about FP precision problems some time ago, and I
> asked for more details then, but you never responded, so they're
> probably still there.
>
I guess so :-(. Was very busy.
> FWIW, I just tried the 5.23. build on a Linux systems using 64-bit ecl
> 10.4.1 (with gcc 4.4.1. Maxima builds just fine. Without more
> information, I'm going to blame gcc 4.1.2. :-)
>
> Are the really some 37000 characters in that one line? What file is that?
>
It is a call to gcc, which is already very long, and then long paths.
The main problem seems now the failure to load anything, as described
in my other e-mail.
Otherwise over the christmas-break I should have time to create test-examples regarding
the precison problems.
Oliver