On 8/8/2012 8:56 PM, Ben Laurie wrote: > On Wed, Aug 8, 2012 at 5:03 PM, Joe Orton <jor...@redhat.com> wrote: >> On Wed, Aug 08, 2012 at 08:00:25AM +0200, Kaspar Brand wrote: >>> My thinking was that people should explicitly tell configure that they >>> want to link with the libs in a build directory (so that they don't >>> "accidentally" use a directory which might only temporarily exist - >>> that's also the primary reason for preferring the static over the shared >>> libs from that dir). >> >> This all seems totally crazy to me. Why are we adding complexity to the >> httpd build system so openssl devs can skip typing the " install" part >> of running "make install" when testing against modified versions of >> OpenSSL? Joe > > IIRC, the issue is that it doesn't build correctly even if you do that > (to a non-standard location - and clearly installing to a standard > location is nuts).
I have no issues installing to a non-standard location, and with a very few hacks, can even eliminate the -R compiled-in search paths and create a build entirely managed by LD_LIBRARY_PATH. It would help if you explained better what bug you see in handling the --with-ssl arguement, and I think we could resolve this easily. If the issue is that openssl.pc is not processed correctly by httpd or not created correctly by openssl.org, then we can fix either defect.
