tgl@sss.pgh.pa.us wrote:
> Yeah, we have seen this before on various platforms. On some,
> LD_LIBRARY_PATH isn't actually anything the dynamic linker pays
> attention to, and on others, it is trumped by the installation rpath
> built into the executables. Not putting an rpath into the executables
> would make life better for "make check" at the cost of making the
> executables much more fragile in actual use. I don't see a good
> solution :-( but if you do, step right up to the plate ...
Well, I'm not going to be the one who suggests that adding:
export LD_PRELOAD=$libdir/libpq.so
to pg_regress.sh before the initdb command will fix it (on Linux).
Because the more I think about it, the less I like the idea of "make check"
as a way to validate PostgreSQL. Not only because it looks like some of us
may have actually been testing a hybrid "new programs/old libraries"
combination, but because "make check" uses a temporary installation. For
acceptance testing, I want to know that it was built and installed
properly, so I'll have to use "make installcheck" after installation. (Say,
why isn't this listed in INSTALL anymore?)