On 8/11/06, Albert Chin <autoconf@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
On Fri, Aug 11, 2006 at 08:52:00AM -0400, Christopher Hulbert wrote: > I have a problem when compiling with intraprocedural analysis flags > (currently with the PGI compilers, I haven't checked other IPA/IPO > capable compilers) where after testing for a library with AC_TRY_LINK, > I test for another library. The first library was compiled with IPA > information, so the PGI compiler creates some updated IPA information > for the conftest. When testing for the next library compiled without > IPA information, it doesn't overwrite the original IPA file and gets > an undefined symbol resulting from the first library. > > So, to boil that confusion down, is there any way (other than issuing > an explicit rm -f *conftest* in my autoconf function) to make sure all > conftest files are deleted? Presumably, the IPA information is stored in the directory of the test. Why not create a shell script that wraps the compiler executable that removes these temporary files? Then, rebuild with: ./configure CC=<path to IPA-compiler wrapper> ...
I guess because the only place I see it useful is in the configure when testing libraries. It would also exhibit a problem to end-users that would require them to read a readme file and follow the instructions :)! Anyways, I put "rm -f *conftest*" at the end of my autoconf macro definition and it seems to work fine on linux. I guess I'll deal with other platforms as they exhibit problems. I was just wondering if there was a more "portable" way of making sure that all test files were removed.
-- albert chin (china@xxxxxxxxxxxxxxxxxx)
_______________________________________________ Autoconf mailing list Autoconf@xxxxxxx http://lists.gnu.org/mailman/listinfo/autoconf
