This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
[RFC] LD_LIBRARY_PATH_64 for sparc solaris testing
- From: Andreas Tobler <toa at pop dot agri dot ch>
- To: GCC <gcc at gcc dot gnu dot org>, Eric Botcazou <ebotcazou at libertysurf dot fr>
- Date: Fri, 09 Jan 2004 14:40:01 +0100
- Subject: [RFC] LD_LIBRARY_PATH_64 for sparc solaris testing
Hi all,
during my tests on solaris 9 I failed to test with the -m64 on a 32 bit
compiler. I was digging for a solution and found the environment
variable LD_LIBRARY_PATH_64. Setting this one made the tests work when I
ran the testsuites with:
make RUNTESTFLAGS="--target_board 'unix{-m64,}'" check
I then thought this would be fine and added locally this path setting
and submitted a few patches to reflect this change.
Later I noticed that others of the solaris testers don't need this path
setting. Now I was confused completely and I started to investigate:
I found that I set this variable initially in my tcshrc.
On the sun documentation side I found the explanation:
http://docs.sun.com/db/doc/817-1983/6mhm6r4e8?q=Solaris+Linker&a=view
Section Evironment Variables.
So, when I initially set this variable (LD_LIBRARY_PATH_64) the
LD_LIBRARY_PATH gets overridden by it. And I need to set it everywhere
where I want to test 64bit libs in a 32 environment. Otherwise my test
suite doesn't find the libraries anymore.
Now my question: shall we introduce this path setting for solaris
everywhere in the testsuite ? Or do we have to document how we treat
LD_LIBRARY_PATH on solaris?
If we set it everywhere in the testsuite it lives as long as the test
suite runs. Then it is no longer needed.
In case one has to set this variable in his default shell due to
whatever reasons he will fail to run the 64 bit testsuites on solaris if
we don't decide to add this path.
I don't have a preference, I think it does not harm if we set it. It can
harm if we don't set it and don't document it.
Opinions?
Thanks i.a.
Andreas