On Mon, 28 Mar 2011, Ian Lance Taylor wrote: > We have several bug reports for 4.6.0 about failures of the form > > checking dynamic linker characteristics... configure: error: Link tests are > not allowed after GCC_NO_EXECUTABLES. > > http://gcc.gnu.org/PR47836 > http://gcc.gnu.org/PR46586 > http://gcc.gnu.org/PR45174 > http://gcc.gnu.org/PR39107 > > The problem arising when building a cross-compiler for a GNU/Linux > target. Configure scripts like libiberty/configure.ac, > libquadmath/configure.ac, etc., use GCC_NO_EXECUTABLES. The > GCC_NO_EXECUTABLES macro checks whether a link succeeds. If it does > not, it sets gcc_no_link. If a later attempt to run a link test, the > above error is reported.
Most of the problem is that certain libraries are built when they shouldn't be built. In general link tests *are* required for GNU/Linux targets, because of e.g. uClibc configurations where some functions may be disabled and this can sometimes only be detected through a link test. Furthermore, when building shared libraries you need a shared libc to link them against, which means that in that case link tests will need to work anyway. For the initial bootstrap for such a target a static-only, C-only compiler is built and used for building libc headers and a dummy libc.so. http://www.eglibc.org/cgi-bin/viewcvs.cgi/*checkout*/trunk/libc/EGLIBC.cross-building?rev=2037 There certainly are issues with the list of libraries to disable in the course of such a bootstrap growing over time (with libquadmath, for example, being a recent addition, but I think the conclusion was that it *should* be built in general even if Fortran isn't being built), and maybe there should be special configure options for the initial bootstrap compiler (which could also replace the --with-newlib hack). But some of those bug reports refer to target libiberty - which I still maintain should never be built by GCC - and some to target zlib, which I think should only be relevant when building Java. -- Joseph S. Myers jos...@codesourcery.com
