On 27 December 2010 22:07, Oliver Kullmann wrote: > Just as your e-mail came in, I've discovered that myself > (reading the make-output :-(). > > There are always so many unknowns in these (unfortunately fully > unspecified) builds, one always runs around like a randomised chicken. > > It runs now much longer as before, so there's hope. > > Will it actually solve the linking problems by itself (apparently > the only possibility is static linking), or does one have to add > run-time linker-instructions when running gcc? I hope not (that > was the major motivation to do it this way). It's entirely possible to link dynamically to libgmp.so and libmpfr.so, by installing gmp and mpfr in system library diretories or by using --with-gmp and --with-mpfr, but if you do that then you need to ensure that those libs will be found at runtime (via ldconfig or LD_LIBRARY_PATH or equivalent mechanisms) Which is why I always recommend that people either build gmp and mpfr with --disable-shared (so they only get static libs) or they put the source trees in the gcc tree, which automatically does static linking.
