* Makefile.am: Add versioning_script.at to testsuite.
* doc/libtool.texi: Update to describe new test file.
* tests/versioning_script.at: Autotest file for testing the output of
the libtool-next-version script.
The Microsoft Macro Assembler can be used to assemble and link one
or more assembly-language source files with libtool.
Reported: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=48993
* Makefile.am: Pass ML64 flags to tests.
* configure.ac: Add language support for the Microsoft Macro Assembler.
* doc/libtool.texi: Update documentation to include ML64.
* m4/libtool.m4: Enable macro support for ML64.
* tests/flags.at: Add test for ML64 flags.
* tests/infer-tag.at: Add test for ML64 tag.
* NEWS: Update.
Test 'deplibs without file command' fails on MSVC with a 'hard failure'.
Since this test was created for MinGW, skipping for now.
Test 'Libtool stress test.' fails on MSVC when the cl compiler cannot
open an expsym file, which appears to exist but is empty.
* tests/deplibs-mingw.at: Add check for MSVC build to skip test.
* tests/stresstest.at: Add check for MSVC build to skip test.
If mklink is used for symlinking files, the expout will not match, so
stdout is used instead.
* tests/libtoolize.at: Use stdout for checks instead of expout on MSVC.
When comparing expected output in LT_AT_CHECK_EXECUTE, several tests
fail with mismatched line endings. This patch should fix Windows
environments that have CRLF text line endings disabled and only use
the LF character for line endings.
* tests/testsuite.at: Removed '\r' from awk's printf command.
With MSVC, linker flags were not properly being linked in during a
compile command. A check has been added to verify if there are linker
flags set and then use a different command when linking.
Fixes applied for bug#74175.
* build-aux/ltmain.in: Fix command parser for '-Wl' option for MSVC.
* m4/libtool.m4: Fix MSVC compile commands.
* tests/flags.at: Use '$GREP --' instead of '$FGREP' when comparing with
stdout.
The test for 'darwin fat compile' is being skipped on 64-bit machines,
since the architecture option was specifying i386 (deprecated on macOS).
Reported: https://savannah.gnu.org/support/?111158
* tests/darwin.at: Update to use architecture arm64 instead of i386.
Current, age, and revision values for a library seem to have been chosen
arbitrarily, which is limiting users to a 5 digit maximum. Now, users
can set revision to at most 19 digits, which is referencing Unix epoch
time in nanoseconds.
Reported: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=75880
* build-aux/ltmain.in: Update revision check to allow up to 19 digits.
* tests/versioning.at: Add more version checks to test.
* NEWS: Update.
Reported: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=67754
* Makefile.am: Set environment variables for tests.
* configure.ac: Add language support for Objective-C and Objective-C++.
* build-aux/ltmain.in: Update help message for tags available.
* m4/libtool.m4: Add tagging support for OBJC and OBJCXX.
* tests/flags.at: Add tests for passing OBJC and OBJCXX flags.
* tests/infer-tag.at: Add tests for inferring OBJC and OBJCXX tags.
* doc/libtool.texi: Update documentation.
* NEWS: Update.
The test 'deplibs_check_method' was designed to verify pass_all
works as expected. Many warnings occur for systems that do not
have deplibs_check_method set to pass_all, so this test is skipped
on those.
* tests/demo.at: Add check for deplibs_check_method=pass_all.
* build-aux/ltmain.in: Make error message more explicit for users,
and add tag name information to help menu.
* tests/infer-tag.at: Add testcase for unsupported compiler use.
Passing --no-warnings to libtool would not suppress warning messages.
* build-aux/ltmain.in: Add 'opt_warning' check before printing out
warning messages.
* tests/libtool.at: Add simple test for '--no-warnings'.
* NEWS: Update.
Add option to reorder the shared library cache in OpenBSD so that user
preferred directories for shared libraries can be used when linking
before directories previously listed in the shared library cache.
This allows for users in OpenBSD to easily switch between versions of
libraries with the same name during testing.
* NEWS: Update for new (OpenBSD) option.
* build-aux/ltmain.in: Add option --reorder-cache=DIRS.
* doc/libtool.texi: Update documentation for new option.
* test/bug_71489.at: Alter test for OpenBSD to utilize new option.
The following errors can be seen for the 'execute mode' test in OpenBSD:
build/libtool: test: <: missing second argument
build/libtool: test: >: missing second argument
The cwrapper in OpenBSD cannot parse these symbols as arguments in
execute mode if there are no other characters with them.
* tests/execute-mode.at: Escape '<' and '>' in OpenBSD.
Update tests to skip if required autotools are not installed for
executing AUTORECONF and to have stdout saved for checks.
* tests/bug_42313.at: Use LT_AT_AUTORECONF instead of AT_CHECK.
* tests/testsuite.at: Save stdout for LT_AT_AUTORECONF.
This test is added to ensure path comparisons pass when
installing to a destination, specifically when there is a
trailing slash in the destination.
* tests/destdir.at: Add test for trailing slash in destination.
Five tests use aclocal, but do not have a check for whether aclocal is
available to use. This causes reported tests failures in Linux From
Scratch and Darwin builds.
* tests/standalone.at: Add AT_XFAIL_IF check for aclocal for all tests.
* tests/subproject.at: Add AT_XFAIL_IF check for aclocal for last test.
The macro _LT_CMD_OLD_ARCHIVE inlines the definition of AC_PROG_RANLIB,
which seems to be outdated functionality. Running autoscan in a
directory with ltmain.sh produces conflicting warnings with libtoolize
about AC_PROG_RANLIB. Autoconf warns that AC_PROG_RANLIB is missing,
but libtoolize warns that it is obsolete. The outdated warning should
no longer appear after replacing AC_CHECK_TOOL(RANLIB, ranlib, :) with
AC_REQUIRE([AC_PROG_RANLIB]).
* m4/libtool.m4: Use AC_REQUIRE([AC_PROG_RANLIB]).
* Makefile.am: Added in tests/bug_42313.at.
* tests/bug_42313.at: Contains test cases for bug 42313.
Reported at <https://savannah.gnu.org/support/index.php?110674>.
* doc/libtool.texi (LT_INIT): Talk about --enable-pic instead of
--with-pic, and about --disable-pic instead of --without-pic.
(Test descriptions): Likewise.
* m4/ltoptions.m4 (_LT_WITH_PIC): Use AC_ARG_ENABLE instead of
AC_ARG_WITH. In the default case, use _AC_ENABLE_IF, in order to still
recognize the old options --with-pic and --without-pic.
* tests/demo.at: Add a test for --enable-pic, keeping the test for
--with-pic. Test '--disable-pic' instead of '--with-pic=no'.
* tests/with-pic.at: Test --enable-pic instead of --with-pic.
* NEWS: Mention the change.
Test 170 in bug_62343.at is failing to link object files with a library
file. This is fixed by choosing the correct architecture version of a
compiler by using the 'CXX' macro.
tests/bug_62343.at: Replace 'g++' with 'CXX' macro
It has been discovered that under certain conditions libtool creates
wrappers that prefer installed versions of the shared objects over
those built in the source tree. As a result, any tests run in the
source tree produce unreliable results.
* build-aux/ltmain.in: Alter to use shared objects in source tree
instead of installed.
* Makefile.am: Added in tests/bug_71489.at.
* tests/bug_71489.at: Contains test case for bug 71489.
Append '-no_fixup_chains' flag to disable chained fixups since it
is not compatible with '-undefined dynamic_lookup'.
* m4/libtool.m4: AC_VAR_APPEND will handle appending the option to
the variable, which allows the '+=' extension to be used by shells
that provide this capability for more efficient scaling. Also, bump
minimum required version of autoconf from 2.62 to 2.64.
* tests/no-executables.at: Set cache variable for link test.
* NO-THANKS: Add thanks for Carlo Cabrera and Dave Allured.
When linking on NetBSD, the test fails to locate a shared version of
gcc, so a static version of the module is generated, which does not pass
the defined checks. A FIXME note has been added for future debugging.
* tests/flags.at: Skip passing CXX flags test on NetBSD.
Without the no-undefined flag, the test will fail on Mingw and
Cygwin, so the flag will be appended for windows compilation. With
the flag, the test will fail on NetBSD, so the flag is removed.
* tests/bug_62343.at: Add no-undefined flag for windows-based
compilation and remove otherwise.
The compiler fails with the -no-canonical-prefixes flag in certain
environments. Check if the flag is supported in the current environment
and skip the test if it is not supported.
* tests/bug_62343.at: Skip test if -no-canonical-prefixes is unsupported
in the current environment.
* testsuite.at: $as_echo was deprecated in Autoconf commit 2b59b6f8a.
AS_ECHO uses printf which has been well supported for 15 to 20
years. Use of $as_echo was resulting in a warning message in the
testsuite logs, and replacing the deprecated macro removes the
warning messages.
* tests/libtoolize.at: When -program-prefix is used, stdout for the
program changes from the expected output. As a result, the test
fails. Replaced the transformed name to account for program-prefix.
Using clang with fsanitize results in an incompatible ASan runtime.
It results in the warning message "linked against incompatible ASan
runtimes". This only occurs with the mixed Fortran/C demo.
* tests/f77demo.at: Skip tests for clang compiling with fsanitize.
* tests/fcdemo.at: Skip tests for clang compiling with fsanitize.
A function used in a test was not returning a value, while an int was
expected so the test would fail to compile.
tests/resident.at: Return the result of atexit().
When the configure option "--disable-shared" is used, this test will be
skipped since no libtool libs are built.
tests/bug_62343.at: Check if build_libtool_libs is set to no.
When using the -no-canonical-prefixes flag in libtool, the flag is
removed from the linking command, causing linking to fail. This adds
the flag to the list of linker flags to be passed through to the
linker.
* build-aux/ltmain.in: Pass -no-canonical-prefixes to the linker.
* Makefile.am: Added in tests/bug_62343.at.
* tests/bug_62343.at: Contains test case for bug 62343.
When running `libtoolize --ltdl`, a symlink to the source config-h.in
is used rather than a copy of the file. This causes the build system
to look at its old timestamp relative to the other generated files
that are copied in, and then try to regenerate things, and update the
config-h.in timestamp by touching the symlink (which will dereference
things). This breaks `make distcheck` because when a few tests run
that invoke `libtoolize --ltdl --install` w/out --copy, and then try
to build the project, automake's rules trigger autoheader regen, and
then update the timestamp, which updates the underlying file, and then
causes libtool's own files to get out of sync. So the next time make
is run in the tree, it tries to regenerate the source files, and fails
as the source tree is read-only.
Change libtoolize to always copying the header over during install.
It's not clear to me whether this is 100% the right answer, but it
shouldn't make things worse beyond missing a case to use a symlink.
We probably can't avoid this as long as aclocal.m4 itself is copied
in instead of symlinked. This changed between v2.4.2 and v2.4.3 in
commit 4e671d6ba69c889b1f9aee12de6c504d26dc6514 ("libtoolize:
simplify file-copying and -linking call-graph.").
Example failure log:
$ make distcheck
...
make[3]: Entering directory '.../libtool-2.4.7.62-3e11/_build/sub'
GEN ../../libltdl/Makefile.in
GEN ../../libltdl/configure
autom4te-2.72: error: cannot create autom4te.cache in .../libtool-2.4.7.62-3e11/libltdl: Permission denied
make[3]: *** [Makefile:2411: ../../libltdl/configure] Error 1
make[3]: *** Waiting for unfinished jobs....
autom4te-2.72: error: cannot create autom4te.cache in .../libtool-2.4.7.62-3e11/libltdl: Permission denied
automake-1.16: error: autoconf failed with exit status: 1
make[3]: *** [Makefile:2387: ../../libltdl/Makefile.in] Error 1
We can see this in old-m4-iface.at:AC_WITH_LTDL which doesn't use
--copy with libtoolize:
$ stat libltdl/config-h.in | grep Modify
Modify: 2024-01-18 01:18:52.998059970 -0500
$ make check-local TESTSUITEFLAGS=100
...
100: AC_WITH_LTDL ok
...
$ stat libltdl/config-h.in | grep Modify
Modify: 2024-01-18 01:19:19.254727020 -0500
And in the test log we can see:
100. old-m4-iface.at:113: testing AC_WITH_LTDL ...
...
libtoolize: linking file 'libltdl/config-h.in'
...
./old-m4-iface.at:161: : ${CONFIG_SHELL=/bin/sh}; export CONFIG_SHELL; $CONFIG_SHELL ./configure $configure_options --prefix="$prefix" --exec-prefix="$prefix" --bindir="$prefix/bin" --libdir="$prefix/lib" --with-included-ltdl
...
./old-m4-iface.at:161: $as_unset LIBTOOL; $as_unset LIBTOOLIZE; $MAKE $target
stderr:
stdout:
cd libltdl && make
make[1]: Entering directory '/usr/local/src/gnu/libtool/tests/testsuite.dir/100/libltdl'
CDPATH="${ZSH_VERSION+.}:" && cd . && aclocal -I m4
cd . && automake --foreign
CDPATH="${ZSH_VERSION+.}:" && cd . && autoconf
/bin/sh ./config.status --recheck
...
(CDPATH="${ZSH_VERSION+.}:" && cd . && autoheader)
rm -f stamp-h1
touch config-h.in
...
* libtoolize.in: Copy config-h.in.
* tests/libtoolize.at: Change "linking" to "copying".
The gnuconfig project recognizes windows* as a host OS to denote native
Windows environments. The commit message makes it sound like LLVM and
Crablang communities will use the 'windows' value, whereas GNU will
continue to use 'mingw'. But I think it's only a matter of time until
people start to pass the option --host=x86_64-pc-windows to configure
scripts. We should be prepared for that.
Url: https://savannah.gnu.org/patch/?10387
* build-aux/ltmain.in: Treat windows* as equivalent to mingw*.
* m4/libtool.m4: Likewise.
* m4/ltdl.m4: Likewise.
* m4/ltoptions.m4: Likewise.
* tests/bindir.at: Likewise.
* tests/deplibs-mingw.at: Likewise.
* tests/lt_dladvise.at: Likewise.
* tests/testsuite.at: Likewise.
