This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: 971031 and Solaris: update
- To: egcs at cygnus dot com
- Subject: Re: 971031 and Solaris: update
- From: Raja R Harinath <harinath at cs dot umn dot edu>
- Date: 03 Nov 1997 20:43:21 -0600
- References: <199711032245.OAA09386@atrus.synopsys.com>
Joe Buck <jbuck@synopsys.com> writes:
> These tests are for sparc-sun-solaris2.5.1, 971031 plus the dwarf2out
> patch.
>
> I now know the reason for the large number of failures I was getting:
> with --enable-shared, runtest doesn't do anything about setting
> LD_LIBRARY_PATH correctly. At one point, I still had LD_LIBRARY_PATH
> pointing to a previously installed libstdc++.so with the null-termination
> bug; then I had no library at all and every test that used the library
> failed. Since the standard error for execute tests is suppressed, I
> didn't see the message that would tell me that this is the problem.
That would explain my results too. I did a `make check' without a
preceding `make install'.
> I suggest modifying the tests to support shared libraries better (putting
> the directories with newly made libraries first on LD_LIBRARY_PATH or
> equivalent)
This is already done for `libio' and `libstdc++', but not for `gcc' and
`g++'.
I usually do `make install; make check'. But after noticing the
SAVE_LLPATH in my make log, I wrongly concluded that things would work
without a `make install'.
- Hari
--
Raja R Harinath ------------------------------ harinath@cs.umn.edu
"When all else fails, read the instructions." -- Cahn's Axiom
"Our policy is, when in doubt, do the right thing." -- Roy L Ash