You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@subversion.apache.org by asdf <fi...@yahoo.com> on 2007/01/24 16:57:21 UTC
svn 1.4.2 tests fail on Solaris 9
I have built svn 1.4.2 on Solaris 9 using gcc-3.4.6 with the following
configure line:
./configure --enable-shared=yes --without-berkeley-db --without-ssl
--with-devrandom=/dev/urandom
When I run the test suite (make check), a number of the tests fail:
Running all tests in commit_tests.py...FAILURE
Running all tests in prop_tests.py...FAILURE
Running all tests in log_tests.py...FAILURE
Running all tests in diff_tests.py...FAILURE
Running all tests in trans_tests.py...FAILURE
Running all tests in svnsync_tests.py...FAILURE
all other tests passed. A closer look at test suite log file,
subversion-1.4.2/tests.log, reveals that the tests are failing because
they are all using hook scripts that call Python and the Python binary
failed to load/run:
EXPECTED STDOUT:
Sending svn-test-work/working_copies/commit_tests-34/iota
Transmitting file data .
Committed revision 2.
Warning: 'post-commit' hook failed with error output:
Post-commit hook failed
ACTUAL STDOUT:
Sending svn-test-work/working_copies/commit_tests-34/iota
Transmitting file data .
Committed revision 2.
Warning: 'post-commit' hook failed with error output:
ld.so.1: /usr/local/bin/python: fatal: libstdc++.so.6: open failed: No
such file or directory
EXCEPTION: SVNLineUnequal
FAIL: commit_tests.py 34: post commit hook failure case testing
And this is the same error that occurs for all the other failing tests.
What I find strange is that I am able to run Python without any
problems otherwise since /usr/local/bin is already in my PATH and
/usr/local/lib is also already in LD_LIBRARY_PATH.
I have Python 2.4.3 installed to /usr/local via the standard
"./configure && make all && make install" and this installation of
Python is also used for Subversion 1.3.2 which is currently installed
on this Solaris 9 system and which passed all of its tests without a
problem.
Can anyone explain why the Python binary would fail to load when called
from the hook scripts in the specific tests failing above?
There was one other user who reported a similar (identical perhaps?)
problem in Nov. 2006 when running the test suite on Solaris 9 but there
were no resolutions to it.
Thanks.
PS: It is worth mentioning that I built svn-1.4.2 on an AIX 5.3 machine
in a similar manner where Python is also installed to /usr/local/bin
and the test suite ran to completion without any failures.
____________________________________________________________________________________
Never miss an email again!
Yahoo! Toolbar alerts you the instant new Mail arrives.
http://tools.search.yahoo.com/toolbar/features/mail/
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: svn 1.4.2 tests fail on Solaris 9
Posted by asdf <fi...@yahoo.com>.
Ed,
Sorry for the very late reply to your reply below...
As per your suggestion below, once I had set LDFLAGS with the location
of my gcc installation I was able to run the subversion test suite on
Solaris 9. And I finally got rid of having to set LD_LIBRARY_PATH just
to run the subversion binaries.
Thanks for help!
Regards,
--- Ed Hillmann <ed...@gmail.com> wrote:
> I try to avoid using the LD_LIBRARY_PATH in Solaris. I found that to
> avoid the svn users from defining directories in their
> LD_LIBRARY_PATH
> values, I needed to compile svn with the following setting:
>
> export LDFLAGS="-L$GCC_HOME/lib -R$GCC_HOME/lib"
>
> Where GCC_HOME is the home directory of my gcc installation. The -R
> flag links the binary with the required libs at linking time instead
> of evaluation time. The svn make file tells you this during a "make
> install" (which I repeatedly ignored until the penny dropped for me).
>
> Once I did this during a build process, I never needed to set
> LD_LIBRARY_PATH again (at least for my Subversion executables).
>
> Hope this helps,
> Ed
>
____________________________________________________________________________________
Bored stiff? Loosen up...
Download and play hundreds of games for free on Yahoo! Games.
http://games.yahoo.com/games/front
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: svn 1.4.2 tests fail on Solaris 9
Posted by Ed Hillmann <ed...@gmail.com>.
I try to avoid using the LD_LIBRARY_PATH in Solaris. I found that to
avoid the svn users from defining directories in their LD_LIBRARY_PATH
values, I needed to compile svn with the following setting:
export LDFLAGS="-L$GCC_HOME/lib -R$GCC_HOME/lib"
Where GCC_HOME is the home directory of my gcc installation. The -R
flag links the binary with the required libs at linking time instead
of evaluation time. The svn make file tells you this during a "make
install" (which I repeatedly ignored until the penny dropped for me).
Once I did this during a build process, I never needed to set
LD_LIBRARY_PATH again (at least for my Subversion executables).
Hope this helps,
Ed
On 1/26/07, asdf <fi...@yahoo.com> wrote:
> --- Ryan Schmidt <su...@ryandesign.com> wrote:
>
> > [snip]
> >
> > Hook scripts run with an empty environment (on non-Windows operating
> >
> > systems) so the variables PATH and LD_LIBRARY_PATH are *not* set
> > when /usr/local/bin/python is trying to locate libstdc++.so.6.
> >
>
> That explains it - I also checked the FAQ and I tested one of the
> failing hook scripts as suggested in the FAQ and I got the same error
> about being unable to load libstdc++.so.6 when trying to run the python
> binary.
>
> The question is: how can one run the test suite when both gcc and
> python are both installed to /usr/local? Surely one ought to be able
> to run the test suite irrespective of where the python binary (and its
> dependencies) are installed to, no?
>
> Judging by the (general) lack of responses to my original post, it
> doesn't appear that too many people have tripped up on this issue.
>
> >
> > I cannot explain why Subversion 1.3.2 would not have this problem.
> >
>
> Perhaps because none of the tests in the 1.3.2 test suite had custom
> hooks calling the Python binary?
>
> Thanks.
>
> PS: I am already subscribed to the list, so need to CC: me on replies. :)
>
>
>
> ____________________________________________________________________________________
> Have a burning question?
> Go to www.Answers.yahoo.com and get answers from real people who know.
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
> For additional commands, e-mail: users-help@subversion.tigris.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: svn 1.4.2 tests fail on Solaris 9
Posted by Ryan Schmidt <su...@ryandesign.com>.
On Jan 25, 2007, at 10:06, asdf wrote:
>> Hook scripts run with an empty environment (on non-Windows operating
>> systems) so the variables PATH and LD_LIBRARY_PATH are *not* set
>> when /usr/local/bin/python is trying to locate libstdc++.so.6.
>
> That explains it - I also checked the FAQ and I tested one of the
> failing hook scripts as suggested in the FAQ and I got the same error
> about being unable to load libstdc++.so.6 when trying to run the
> python
> binary.
>
> The question is: how can one run the test suite when both gcc and
> python are both installed to /usr/local? Surely one ought to be able
> to run the test suite irrespective of where the python binary (and its
> dependencies) are installed to, no?
>
> Judging by the (general) lack of responses to my original post, it
> doesn't appear that too many people have tripped up on this issue.
I've never used LD_LIBRARY_PATH; I'm not familiar with software that
tries to link to other libraries via relative paths, like your python
appears to. I'm used to using software that already knows the
absolute path to the libraries it requires. I was under the
impression that the latter was the usual way software is compiled,
but I don't have that much experience with compiling software, so I
may not know what I'm talking about.
>> I cannot explain why Subversion 1.3.2 would not have this problem.
>
> Perhaps because none of the tests in the 1.3.2 test suite had custom
> hooks calling the Python binary?
If that's so, then that would certainly explain it.
> PS: I am already subscribed to the list, so need to CC: me on
> replies. :)
I guess you meant "no need to CC: me." But of the hundreds of
subscribers to the list, I'm not going to be able to remember who
wants me to CC them and who doesn't. This list does not munge the
Reply-To header. When I press Reply to one of your messages, it goes
to you. When I press Reply To All, it goes to you and the list. If
you want replies to go only to the list, you need to set the Reply-To
header of your outgoing mails to the list's address.
--
To reply to the mailing list, please use your mailer's Reply To All
function
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: svn 1.4.2 tests fail on Solaris 9
Posted by asdf <fi...@yahoo.com>.
--- Ryan Schmidt <su...@ryandesign.com> wrote:
> [snip]
>
> Hook scripts run with an empty environment (on non-Windows operating
>
> systems) so the variables PATH and LD_LIBRARY_PATH are *not* set
> when /usr/local/bin/python is trying to locate libstdc++.so.6.
>
That explains it - I also checked the FAQ and I tested one of the
failing hook scripts as suggested in the FAQ and I got the same error
about being unable to load libstdc++.so.6 when trying to run the python
binary.
The question is: how can one run the test suite when both gcc and
python are both installed to /usr/local? Surely one ought to be able
to run the test suite irrespective of where the python binary (and its
dependencies) are installed to, no?
Judging by the (general) lack of responses to my original post, it
doesn't appear that too many people have tripped up on this issue.
>
> I cannot explain why Subversion 1.3.2 would not have this problem.
>
Perhaps because none of the tests in the 1.3.2 test suite had custom
hooks calling the Python binary?
Thanks.
PS: I am already subscribed to the list, so need to CC: me on replies. :)
____________________________________________________________________________________
Have a burning question?
Go to www.Answers.yahoo.com and get answers from real people who know.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: svn 1.4.2 tests fail on Solaris 9
Posted by Ryan Schmidt <su...@ryandesign.com>.
On Jan 24, 2007, at 10:57, asdf wrote:
> When I run the test suite (make check), a number of the tests fail:
>
> Running all tests in commit_tests.py...FAILURE
> Running all tests in prop_tests.py...FAILURE
> Running all tests in log_tests.py...FAILURE
> Running all tests in diff_tests.py...FAILURE
> Running all tests in trans_tests.py...FAILURE
> Running all tests in svnsync_tests.py...FAILURE
[snip]
> Warning: 'post-commit' hook failed with error output:
> ld.so.1: /usr/local/bin/python: fatal: libstdc++.so.6: open failed: No
> such file or directory
[snip]
> What I find strange is that I am able to run Python without any
> problems otherwise since /usr/local/bin is already in my PATH and
> /usr/local/lib is also already in LD_LIBRARY_PATH.
>
> I have Python 2.4.3 installed to /usr/local via the standard
> "./configure && make all && make install" and this installation of
> Python is also used for Subversion 1.3.2 which is currently installed
> on this Solaris 9 system and which passed all of its tests without a
> problem.
>
> Can anyone explain why the Python binary would fail to load when
> called
> from the hook scripts in the specific tests failing above?
[snip]
Hook scripts run with an empty environment (on non-Windows operating
systems) so the variables PATH and LD_LIBRARY_PATH are *not* set
when /usr/local/bin/python is trying to locate libstdc++.so.6.
I cannot explain why Subversion 1.3.2 would not have this problem.
--
To reply to the mailing list, please use your mailer's Reply To All
function
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org