[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[libtool 2.2.2] testsuite: ... 38 ... 56 ... failed

From: Michael Haubenwallner
Subject: [libtool 2.2.2] testsuite: ... 38 ... 56 ... failed
Date: Fri, 18 Apr 2008 16:22:03 +0200


Here, on AIX, I have some gcc (4.2.3) in non-standard prefix, providing
and there is some older gcc (3.3.6) in /usr/local, providing

These two checks definitely are related to that layout:
   38: C++ subdir-objects     FAILED (am-subdir.at:148)
   56: simple template test   FAILED (template.at:92)

Because these checks do not actually install something, they use
"/usr/local" as prefix, either from defaults or explicitly.

Because of this, in test 38, the binary for subdir/subdemo
(subdir/.libs/subdemo) gets "/usr/local/lib" encoded as runpath,
intended to find "libsub.a(libsub.so.0)" if it were installed.

But now, when executing subdir/subdemo, it finds "libstdc++.a" from
"/usr/local/lib", wich lacks libstdc++.so.6, causing this test to fail.

Test 56 has the very same problem, although it uses explicit
"-rpath /usr/local/lib".

IMO libtool testsuite should use build-environment-agnostic prefixes.

Attached is:

[lt-2.2.2-testsuite-38.log.bz2] testsuite.dir/38/testsuite.log
[lt-2.2.2-testsuite-56.log.bz2] testsuite.dir/56/testsuite.log

proposed patch to change any default or explicit occurrence of
"/usr/local/*" to "/tmp/nonexistent", although it actually hurts only in
template.at and am-subdir.at (inherits prefix from testsuite.at) here.



Attachment: lt-2.2.2-testsuite-38.log.bz2
Description: application/bzip

Attachment: lt-2.2.2-testsuite-56.log.bz2
Description: application/bzip

Attachment: lt-2.2.2-testsuite-usrlocal.patch
Description: Text Data

reply via email to

[Prev in Thread] Current Thread [Next in Thread]