From: "Ben Elliston" <b...@au1.ibm.com>
On Fri, 2009-04-10 at 23:56 -0400, Kaveh R. GHAZI wrote:
Ah, but cake is only easy when someone else bakes it. :-)
While you're baking, Kaveh :-) could you see if your patch could also
fix:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34818
Thanks, Ben
I don't think they're related. The original bug I investigated is only
exposed when combining in-tree builds of either gmp or mpfr, with specifying
out-of-tree location for the other of the two libraries.
This new PR appears to be caused by having an old mpfr in the same directory
as the gmp specified by --with-gmp and trying to specify a new mpfr location
via --with-mpfr as well. So you get two copies of mpfr in your -I search
path and in this particular case the old one gets picked by chance. Fixing
it would break the reverse case of having two gmp libraries.
I am inclined to declare this to be user error.
--Kaveh
