[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Configuration bug in GMP 4.1.2
From: |
Torbjorn Granlund |
Subject: |
Re: Configuration bug in GMP 4.1.2 |
Date: |
03 Jan 2003 11:22:23 +0100 |
User-agent: |
Gnus/5.09 (Gnus v5.9.0) Emacs/21.2 |
"Krzysztof Kozminski" <address@hidden> writes:
Yes, this will work; but the real problem is that the effect
(failure when running nm) is so far from the cause (selection of cc
over gcc) than it may waste a lot of time for many people. Maybe
just add a hint to be printed when this particular test fails, so
that the output looks like this:
[snip]
Better error messages from configure would be nice, but it
is much easier said than done.
There is another problem. I added /usr/ccs/bin to my PATH
to get a version of 'nm' that understood the 64-bit code.
Now the configuration went OK, but g++ croaked, quite
expectedly, when trying to build .libs/libgmpxx.so.3.0.4:
./.libs/libgmp.so: file not recognized: File format not recognized
collect2: ld returned 1 exit status
Now this failure at the very end of the build might be
REALLY hard to figure out.
You need to use a 64-bit capable C++ compiler.
Perhaps this whole issue should be added to the
Known-Build-Problems page until the configuration script
tests the compilers' consistency?
There is some discussion about this sort of problems under
ISA and ABI.
You might get some joy by using Sun's C++ compiler. But if
I remember correctly, that compiler is not up to the task.
If you have a set of obsolete tools first in your PATH that
are unaware of the 64-bit ABI, don't be surprised that GMP
build stumbles over them. After all, `nm' is supposed to
work in a standard way, just like `as', `ar', etc.
--
Torbjörn