tests: we can assume GNU make: related small fixes
* t/self-check-shell-no-trail-bslash.sh: Do not explicitly
require GNU make. Issue revealed by the maintainer check
'sc_tests_no_gmake_requirement'.
* t/ax/shell-no-trail-bslash.in: Remove workaround for BSD
make. Improve comments.
* master: (30 commits)
news: we'll remove AM_PROG_MKDIR in 1.14, not 1.13
tests: give few vala tests more significant names
vala tests: don't use the 'posix' profile, it's no longer supported
news: fixup: bug#8847 is fixed in 1.12.6, not in 1.12.5
news: document the fix for automake bug#10227
tests: fix a spurious testsuite failure on Solaris
tests: give few aclocal tests more significant names
tests: remove a redundant workaround for a testsuite weakness
depcomp: remove 'icc' depmode, add 'tcc' one
news: update w.r.t. recent depcomp changes
tests: fix spurious failures with tcc
depcomp: improve comments about tcc support
depcomp: remove last relics of libtool 1.4 support
depcomp: remove an unused variable
depcomp: rewrite a sed invocation for less quoting
depcomp: whitespace fixes
depcomp: shell code style changes
depcomp: make some code more self-documenting
depcomp: less duplication between AIX and Tru64 modes
depcomp: avoid an extra fork when possible
...
* depcomp-work:
depcomp: remove 'icc' depmode, add 'tcc' one
news: update w.r.t. recent depcomp changes
tests: fix spurious failures with tcc
depcomp: improve comments about tcc support
depcomp: remove last relics of libtool 1.4 support
depcomp: remove an unused variable
depcomp: rewrite a sed invocation for less quoting
depcomp: whitespace fixes
depcomp: shell code style changes
depcomp: make some code more self-documenting
depcomp: less duplication between AIX and Tru64 modes
depcomp: avoid an extra fork when possible
depcomp: reduce code duplication
depcomp: don't rely on character ranges working as in C locale
depcomp: cosmetic fixlets to a comment
depcomp: safer quoting in variable expansion
depcomp: correctly propagate exit status in exit trap
depcomp: use "rmdir", not "rm -rf", to remove the empty lock directory
depcomp: remove useless quoting in variable definitions
depcomp: be more idiomatic in checking for zero exit statuses
* maint:
news: we'll remove AM_PROG_MKDIR in 1.14, not 1.13
tests: give few vala tests more significant names
vala tests: don't use the 'posix' profile, it's no longer supported
news: fixup: bug#8847 is fixed in 1.12.6, not in 1.12.5
news: document the fix for automake bug#10227
tests: fix a spurious testsuite failure on Solaris
python: make installed modules find by default on Debian and Ubuntu
vala tests: don't use the 'posix' profile, it's no longer supported
Fixes automake bug#12934 a.k.a. bug#12522.
Some of automake's Vala tests used to rely on "valac --profile posix"
(which makes it not use glib). However, the posix profile was removed
in August 2012, and is no longer present in the GNOME 3.6.2 version
of valac (see commit ca020bf0 in the vala Git repository).
Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
* t/vala-libs.sh: Adjust to avoid using the posix profile.
* t/vala-mix.sh: Likewise.
* t/vala-mix2.sh: Likewise.
* t/vala-parallel.sh: Likewise.
* t/vala-vapi.sh: Likewise.
* t/vala-vpath.sh: Likewise.
* NEWS: Update and adjust.
According to the comments in the now-removed test 'depcomp5.test' (removed
by commit 'v1.11-1829-g5066c1b' of 2012-02-08, "tests: improve and rework
tests on dependency tracking") "icc 8.0 and greater understand gcc options,
so depmode is set to gcc". According to Wikipedia, the 8.0 version of the
Intel C compiler has been released in December 2003, that is, almost nine
yeas ago. So we can remove explicit support for icc 7.x from the depcomp
script.
Since our support for tcc has so far been relied on the icc depmode, we
must write a new explicit 'tcc' depmode. That is easy to do, since we can
rely on the simpler and better-specified format of tcc-specified makefiles;
for more info, see:
<http://lists.gnu.org/archive/html/automake/2012-10/msg00043.html>
* lib/depcomp (icc): Here. See also tcc commit 0394caf7 "Emit spaces
for -MD" and automake commit v1.11-768-gb96263 "depcomp: support tcc
(Tiny C Compiler)".
* lib/depcomp: Here, when we have to get the directory and basename
components of objects or source files; do so with the help of ...
(set_dir_from, set_base_from): ... these new shell functions.
depcomp: don't rely on character ranges working as in C locale
* lib/depcomp (pgcc): Here.
Related reorganization, with the unconditional definition of ...
($upper, $lower, $digits, $alnum, $alpha): ... these shell variables.
* lib/depcomp (pgcc): Here, as that's unportable according to the
autoconf manual. In fact, there's no need to re-trap the signals
just before exiting, since at that point a signal would cause at
most a useless but harmless attempt to remove the
Hmph, I did a merge of 'master' into 'ng/master' without pulling first,
so I missed earlier changes to 'ng/master' that were in the central
repository but not in my local repository. Since that merge required
some non-trivial conflicts resolutions, I don't want to redo it; I'm
thus merging the 'ng/master' from the central repository to the local
one "after the fact".
* master: (38 commits)
tests: fix a spurious failure with older flex versions
tests: avoid a failure due to m4 builtin 'sinclude' in package name
tests: fix a spurious failure when $PYTHON is in the environment
python tests: support PEP-3147 installation layout
python: uninstall cater to PEP-3147
tests: more explicative names for few aclocal tests
tests: improve coverage on Texinfo + silent rules
build: fix rebuild rules for Makefile.in and aclocal.m4
hacking: release procedure: fix order of some steps
maint: post-release minor version bump
release: stable release 1.12.5
maintcheck: avoid failures, sync with 'ng/master' more
maintcheck: remove 'sc_perl_syntax'
sync: update files from upstream with "make fetch"
maint: get rid of $perllibdir environment variable hack
configure: respect the '-q' option better
cosmetics: de-tabify configure.ac
fixup: Automake installation works again
tests: AC_CONFIG_MACRO_DIRS: ignore inevitable failures with old autoconf
recursion: remove _AM_EXTRA_RECURSIVE_TARGETS indirection
...
python: make installed modules find by default on Debian and Ubuntu
This change fixes automake bug#10227.
The code used to get the python package directory was wrong for Python 3,
at least on Debian and Ubuntu distributions. In the case the installation
was using the default prefix "/usr/local", python modules were incorrectly
installed in the directory
/usr/local/lib/python3/dist-packages
(which is *not* searched by default), rather than in a directory like
/usr/local/lib/python3.x/dist-packages
which is searched by default.
* m4/python.m4 (AM_PATH_PYTHON): Try to use the 'sysconfig' module if
possible, for better interactions with python 3.x.
tests: fix a spurious failure with older flex versions
See automake bug#11524 and bug#12836.
* t/lex-header.sh: Here, we use flex option '--header-file',
but that option is not supported by some older flex versions.
Simply skip this test if such an old flex version is detected.
* maint:
tests: fix a spurious failure when $PYTHON is in the environment
python tests: support PEP-3147 installation layout
python: uninstall cater to PEP-3147
tests: improve a comment
tests: honour $PYTHON override
tests: typofix in message
news: document fix for bug#8847 (PEP-3147, __pycache__)
python: improve support for modern python (CPython 3.2 and PyPy)
* python-pep-3147:
tests: fix a spurious failure when $PYTHON is in the environment
python tests: support PEP-3147 installation layout
python: uninstall cater to PEP-3147
tests: improve a comment
tests: honour $PYTHON override
tests: typofix in message
news: document fix for bug#8847 (PEP-3147, __pycache__)
python: improve support for modern python (CPython 3.2 and PyPy)
tests: fix a spurious failure when $PYTHON is in the environment
* t/python11.sh: This test doesn't interact well with user-overrides of
$PYTHON; and, given its particular nature, neither should it be expected
to honour that override. Just unset that $PYTHON variable and live
happy.
python tests: support PEP-3147 installation layout
This fixes several spurious testsuite failures with python >= 3.2,
introduced by recent commit v1.12.4-43-ge0e99ed, "python: improve
support for modern python (CPython 3.2 and PyPy)".
* t/ax/am-test-lib.sh (python_has_pep3147, pyc_location, py_installed):
New functions.
($am_pep3147_tag): New variable.
* t/py-compile-basic2.sh: Deleted, too difficult to adapt for the
gain it would offer; move the still relevant parts ...
* t/py-compile-basic.sh: ... here. Adapt and adjust the rest of
the test as well.
* t/nobase-python.sh: Adapt and adjust.
* t/py-compile-basedir.sh: Likewise.
* t/py-compile-destdir.sh: Likewise.
* t/py-compile-option-terminate.sh: Likewise.
* t/python-pr10995.sh: Likewise.
* t/python-virtualenv.sh: Likewise.
* t/python10.sh: Likewise.
* t/python12.sh: Likewise.
* t/python3.sh: Likewise.
* t/list-of-tests.mk: Adjust list of tests.
After recent commit commit v1.12.4-43-ge0e99ed, "python: improve
support for modern python (CPython 3.2 and PyPy)", the python install
rules have been made smart enough to install "*.pyc" byte-compiled
files according to PEP-3147 with modern (post-3.2) pythons. However,
the uninstall rules hadn't been updated accordingly, causing leftover
files to remain around after "make uninstall", as well as failures in
"make distcheck".
* lib/am/python.am (am__pep3147_tweak): New internal macro, used
in ...
(uninstall-%DIR%PYTHON): ... the recipe of this target, which has
been adjusted to cater to PEP-3147
* t/silent-texi.sh: Here, by basic importing the version of the
test that is the 'ng/master' branch (that of Automake-NG), plus
a minor tweak to avoid a spurious failure with NetBSD make.
* maint:
build: fix rebuild rules for Makefile.in and aclocal.m4
hacking: release procedure: fix order of some steps
maint: post-release minor version bump
release: stable release 1.12.5
build: fix rebuild rules for Makefile.in and aclocal.m4
They have been broken after our removal of the 'perllibdir' hack
in yesterday's commit v1.12.4-45-g4872dfe.
* configure.ac: Rather then redefining AUTOMAKE and ACLOCAL to
explicitly override the perllibdir, the acdir and the libdir, use
the ready-made t/wrap/{aclocal,automake}-$APIVERSION scripts.
* NEWS: Document the fix.
hacking: release procedure: fix order of some steps
* HACKING (Release procedure): The manuals should be rebuilt and
uploaded at at www.gnu.org *before* bumping the version number to
the next alpha release.
maintcheck: avoid failures, sync with 'ng/master' more
* syntax-checks.mk (sc_diff_aclocal_in_aclocal,
sc_diff_automake_in_automake): Rename ...
(sc_diff_aclocal, sc_diff_automake): ... respectively to these, rewrite
to use static pattern rules and 'diff -u' instead of bare 'diff'. This
makes the rules more robust in the face of edits to 'automake.in' and
'aclocal.in', and get them in sync with the corresponding checks in the
'ng/master' branch.
(syntax_check_rules): Adjust.
It is quite useless, since a "make bootstrap" or "make check" would
immediately catch a breakage in the 'aclocal' or 'automake' scripts
anyway. In addition, the recent removal of the 'perllibdir' environment
variable hack has broken this check.
* syntax-checks.mk (sc_perl_syntax): Remove.
(syntax_check_rules): No longer list it.
maint: get rid of $perllibdir environment variable hack
That was needed when the automake-$APIVERSION and aclocal-$APIVERSION
wrapper scripts used by the testsuite were shell script, rather than
perl scripts acting as a thin layers.
* aclocal.in (BEGIN): Update @INC based on the contents of the array
'@Aclocal::perl_libdirs', rather than of the environment variable
'perllibdir'.
* t/wrap/aclocal.in (BEGIN): Initialize '@Aclocal::perl_libdirs'
rather than $ENV{'perllibdir'}.
* automake.in (BEGIN): Update @INC based on the contents of the array
'@Automake::perl_libdirs', rather than of the environment variable
'perllibdir'.
* t/wrap/automake.in (BEGIN): Initialize '@Automake::perl_libdirs'
rather than $ENV{'perllibdir'}.
* NEWS: Update.
* master:
fixup: Automake installation works again
tests: AC_CONFIG_MACRO_DIRS: ignore inevitable failures with old autoconf
recursion: remove _AM_EXTRA_RECURSIVE_TARGETS indirection
aclocal: AC_CONFIG_MACRO_DIRS: work around autom4te option parsing bugs
aclocal: tracing AC_CONFIG_MACRO_DIRS can work with older autoconf as well
news: we trace AC_CONFIG_MACRO_DIRS
news: improve wording
aclocal: trace AC_CONFIG_MACRO_DIR_TRACE
aclocal: avoid spurious warnings from autom4te with AC_CONFIG_MACRO_DIRS
coverage: expose a bug in aclocal (spurious warnings)
aclocal: smash newlines in arguments of traced macros
tests: better coverage for AC_CONFIG_MACRO_DIRS
aclocal: diagnose non-existing directories in AC_CONFIG_MACRO_DIRS better
aclocal: multiple local m4 macro dirs with AC_CONFIG_MACRO_DIRS
By renaming 'dist_automake_ac_DATA' to 'nobase_dist_automake_ac_DATA' in
a recent patch, we have broken the Automake installation, since m4 files
that were expected to be installed in '${prefix}/aclocal-${APIVERSION}'
are now being installed in '${prefix}/aclocal-${APIVERSION}/m4' instead.
Fix this mess.
* Makefile.am (nobase_dist_automake_ac_DATA): Rename ...
(dist_automake_ac_DATA): ... like this. And move
'm4/internal/ac-config-macro-dirs.m4' ...
(dist_automake_internal_ac_DATA): ... here.
(automake_internal_acdir): New, simply defined to
'$(automake_acdir)/internal'.
* ac-config-macro-dirs:
tests: AC_CONFIG_MACRO_DIRS: ignore inevitable failures with old autoconf
aclocal: AC_CONFIG_MACRO_DIRS: work around autom4te option parsing bugs
aclocal: tracing AC_CONFIG_MACRO_DIRS can work with older autoconf as well
tests: AC_CONFIG_MACRO_DIRS: ignore inevitable failures with old autoconf
When "older" version of autoconf are used (that is, those before commit v2.69-44-g1ed0548), we have no sane way to prevent the autom4te invocation
issued from aclocal to possibly display warnings "MACRO m4_require'd but
not m4_defun'd". That's not a big deal, because that just means that
people using pre-2.70 autoconf with cutting-edge automake will see few
spurious warnings, but the actual semantics will remain correct.
However, this blemish was causing a couple of annoying testsuite failures.
Solve this by simply skipping the affected tests when older (pre-2.70)
autoconf versions are used.
* t/aclocl-macrodir.tap ("AC_CONFIG_MACRO_DIR interaction with
AC_REQUIRE"): Skip when older autoconf is in use.
* t/aclocl-macrodirs.tap ("AC_CONFIG_MACRO_DIR interaction with
AC_REQUIRE"): Likewise.
Since automake automatically smashes extra whitespace and newlines
in the arguments of the macros it traces, we can remove the extra
indirection with the _AM_EXTRA_RECURSIVE_TARGETS. This doesn't
truly simplify the existing code, but make it slightly easier to
understand anyway, since leave it with one less indirection.
* automake.in: Trace AM_EXTRA_RECURSIVE_TARGETS directly, rather
than tracing _AM_EXTRA_RECURSIVE_TARGETS. Remove trailing and
leading empty fields in the extracted arguments.
* m4/extra-recurs.m4 (_AM_EXTRA_RECURSIVE_TARGETS): Remove, no longer
needed.
(AM_EXTRA_RECURSIVE_TARGETS): Adjust by making it a simple no-op.
aclocal: AC_CONFIG_MACRO_DIRS: work around autom4te option parsing bugs
The autom4te program coming with autoconf 2.68 and earlier had a bug
which caused the "-" command line argument (with which we tell it to
read some input from from standard input) to aways be pushed at the
*end* of the command line, regardless of where the user specified it
(that bug was fixed by autoconf commit 'v2.68-120-gf4be358', "getopt:
new Autom4te::Getopt module").
This broken semantics conflict with our usage in aclocal, where we
need to pass some input to the invoked autom4te program early, and
have so far been using the stdin to do so. Now we start using an
external file instead.
* m4/internal/ac-config-macro-dirs.m4: New file, contain a fallback
definition of the AC_CONFIG_MACRO_DIRS macro for older autoconf
releases.
* aclocal.in (trace_used_macros): When invoking autom4te, use that
file instead of "abusing" standard input.
* Makefile.am (dist_automake_ac_DATA): Rename ...
(nobase_dist_automake_ac_DATA): ... like this.
Add 'm4/internal/ac-config-macro-dirs.m4' to it.
* t/aclocal-acdir.sh: Adjust to avoid spurious failures.
Helped-by: Eric Blake <eblake@redhat.com> Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
aclocal: tracing AC_CONFIG_MACRO_DIRS can work with older autoconf as well
This will allow our users to interact also with pre-2.70 autoconf without
need for the user to add ACLOCAL_AMFLAGS in Makefile.am. For example,
before this change, in order to have aclocal look for macros in 'm4/dir1'
and 'm4/dir2' also when (say) autoconf 2.69 was used, our users would
have had to add something like:
ACLOCAL_AMFLAGS = -I m4/dir1 -I m4/dir2
in Makefile.am, in addition to the
AC_CONFIG_MACRO_DIRS([m4/dir1 m4/dir2])
in configure.ac. Now, the AC_CONFIG_MACRO_DIRS call is enough.
See the long-winded discussion on automake bug#12845 for more details:
<http://debbugs.gnu.org/cgi/bugreport.cgi?bug=12845>
* aclocal.in ($ac_config_macro_dirs_fallback): New global variable,
contains m4 code to issue a fallback definition of AC_CONFIG_MACRO_DIRS
as an alias for the private macro _AM_CONFIG_MACRO_DIRS.
(trace_used_macros): Handle and trace that macro. Do some code
reorganization and fix related botched indentation while at it.
(write_aclocal): Output '$ac_config_macro_dirs_fallback' early in
the generated aclocal.m4.
* t/aclocal-macrodirs.tap: Run unconditionally, even with older
autoconf.
* t/subpkg-macrodir.sh: Likewise.
* doc/automake.texi: Document only AC_CONFIG_MACRO_DIRS, rather
than AC_CONFIG_MACRO_DIR.
* ac-config-macro-dirs:
news: we trace AC_CONFIG_MACRO_DIRS
aclocal: trace AC_CONFIG_MACRO_DIR_TRACE
aclocal: avoid spurious warnings from autom4te with AC_CONFIG_MACRO_DIRS
coverage: expose a bug in aclocal (spurious warnings)
aclocal: smash newlines in arguments of traced macros
tests: better coverage for AC_CONFIG_MACRO_DIRS
aclocal: diagnose non-existing directories in AC_CONFIG_MACRO_DIRS better
aclocal: multiple local m4 macro dirs with AC_CONFIG_MACRO_DIRS
* t/ax/am-test-lib.sh (require_tool): Here.
* t/python-too-old.sh: And here.
* t/python-vars.sh: And here.
* t/python-virtualenv.sh: And here. Also add some sanity
checks while at it.
python: improve support for modern python (CPython 3.2 and PyPy)
This fixes automake bug#8847.
* m4/python.m4 (AM_PATH_PYTHON): Add python3.3 to
_AM_PYTHON_INTERPRETER_LIST.
* lib/py-compile: Fix compiled filenames for PEP-3147, currently
implemented in CPython 3.2 and newer. Do not create '.pyo' files
for PyPy.
This is now the documented way, and indeed the only correct way, to
support AC_CONFIG_MACRO_DIR and AC_CONFIG_MACRO_DIRS in third-party
tools. See autoconf commit v2.69-45-g5269030 of 2012-11-09,
"AC_CONFIG_MACRO_DIRS: improve tracing and add sanity checks".
* aclocal.in (trace_used_macros): Trace 'AC_CONFIG_MACRO_DIR_TRACE'
rather than 'AC_CONFIG_MACRO_DIRS'. We still have to trace
'AC_CONFIG_MACRO_DIR' explicitly though, for compatibility with
Autoconf versions before 2.70.
* t/aclocal-macrodirs.tap: Enhance w.r.t. precedence of arguments
in one and several AC_CONFIG_MACRO_DIRS calls.
aclocal: avoid spurious warnings from autom4te with AC_CONFIG_MACRO_DIRS
When some macro expanded in configure.ac calls AC_REQUIRE on another
macro that is defined in one of the local m4 macro dirs specified
with AC_CONFIG_MACRO_DIRS, aclocal prints spurious warnings like:
configure.ac:4: warning: MY_BAR is m4_require'd but not m4_defun'd
configure.ac:3: MY_FOO is expanded from...
Such warnings come from autom4te, and are due to the fact that the
*first* autom4te invocation issued by aclocal is not yet able to
"see" the m4 macro definitions in the local m4 dirs (because they
can be looked for only after the AC_CONFIG_MACRO_DIRS call has been
traced, and tracing it requires running autom4te).
To allow us to work around this issue, autom4te has introduced a new
"witness" macro 'm4_require_silent_probe', that, when defined, allows
us to silence that particular kind of warnings (and only it).
Reported by Nick Bowler; see point (4) of:
<http://lists.gnu.org/archive/html/autoconf-patches/2012-11/msg00000.html>
* aclocal.in (trace_used_macros): Pre-define the special macro
'm4_require_silent_probe' when invoking autom4te.
* t/aclocal-macrodirs.tap ("AC_CONFIG_MACRO_DIR interaction with
AC_REQUIRE"): This test passes now: remove the "TODO" directive.
* t/aclocal-macrodir.tap ("AC_CONFIG_MACRO_DIRS interaction with
AC_REQUIRE"): Likewise.
* t/acloca17.sh: Remove.
* t/list-of-tests.mk: Adjust.
coverage: expose a bug in aclocal (spurious warnings)
When some macro expanded in configure.ac calls AC_REQUIRE on another
macro that is defined in one of the local m4 macro dirs specified
with one of the macros AC_CONFIG_MACRO_DIRS or AC_CONFIG_MACRO_DIR,
aclocal prints spurious warnings like:
configure.ac:4: warning: MY_BAR is m4_require'd but not m4_defun'd
configure.ac:3: MY_FOO is expanded from...
Expose this weakness in our testsuite.
Reported by Nick Bowler; see point (4) of:
<http://lists.gnu.org/archive/html/autoconf-patches/2012-11/msg00000.html>
* t/aclocal-macrodir.tap ("AC_CONFIG_MACRO_DIR interaction with
AC_REQUIRE"): New test, still xfailing.
* t/aclocal-macrodirs.tap ("AC_CONFIG_MACRO_DIRS interaction with
AC_REQUIRE"): Likewise.
aclocal: smash newlines in arguments of traced macros
This change fixes the existing issues with AC_CONFIG_MACRO_DIRS
containing newlines:
<http://lists.gnu.org/archive/html/autoconf-patches/2012-11/msg00000.html>
Likely, it will also allow a less involved implementation of the
AM_EXTRA_RECURSIVE_TARGETS macro (but that is left for potential
follow-up patches).
* aclocal.in (trace_used_macros): When calling autom4te, pass its
'--trace' option an argument containing '${::}%' rather than '$1'.
According to the autoconf manual (as of version 2.69), that will expand
to the concatenation, with the '::' string, of all the arguments passed
to a macro, with all newline characters in such arguments smashed.
Related adjustments when handling the macro AC_CONFIG_MACRO_DIRS, to
ensure leading whitespace in its argument are handled correctly.
* t/aclocal-macrodirs.tap ("AC_CONFIG_MACRO_DIRS: extra whitespace"):
No longer declare it as an xfailing test.
Suggested by a report from Nick Bowler:
<http://lists.gnu.org/archive/html/autoconf-patches/2012-11/msg00000.html>
* t/aclocal-macrodirs.tap: Enhance to check use of extra whitespaces
and newline characters in the arguments to AC_CONFIG_MACRO_DIRS. This
test currently fails, so mark it as "TODO". While at it, throw in other
minor edits and enhancements.
aclocal: diagnose non-existing directories in AC_CONFIG_MACRO_DIRS better
This new implementation ensures that any directory (possibly excluding
the first one, if the '--install' option is used) that is declared with
AC_CONFIG_MACRO_DIRS and that is non-existent will cause an error from
aclocal.
* aclocal.in (scan_m4_dirs): Add a new argument, telling whether it's OK
for the scanned directory to be non-existing. Adjust the implementation
accordingly.
($first_user_m4dir): Remove, no more needed.
(scan_m4_files): Update 'scan_m4_dirs' invocations so that aclocal will
not complain if the first user macro directory is non-existing and the
'--install' option is given: such directory will be created later by
aclocal itself.
* t/aclocal-macrodir.tap: Do not mark the last test as TODO anymore;
it now passes. Make stricter by ensuring a non-existing directory in
AC_CONFIG_MACRO_DIRS causes an hard error, not a warning.
aclocal: multiple local m4 macro dirs with AC_CONFIG_MACRO_DIRS
A new macro 'AC_CONFIG_MACRO_DIRS' has been recently introduced in
autoconf (and is expected to appear in the autoconf 2.70 release),
allowing us to declare several local m4 macro directories for a
package.
It can be done either passing several arguments to a single invocation:
This will allow projects to use several m4 macro local dirs, without the
need to use ACLOCAL_AMFLAGS (which we want to make obsolete and finally
remove). This is especially important for projects that are used as
nested subpackages of larger projects.
For more information and rationales, refer to these past discussions:
as well as to Automake commit v1.12.1-165-gcd1a9cc of 2012-07-03,
"aclocal: deprecate ACLOCAL_AMFLAGS, trace AC_CONFIG_MACRO_DIR instead",
autoconf commit v2.69-42-gd73770f of 2012-10-17, "AC_CONFIG_MACRO_DIRS:
new macro, mostly for aclocal".
* aclocal.in ($ac_config_macro_dir): Turn this global scalar it into ...
(@ac_config_macro_dirs): ... this global array.
(trace_used_macros): Update '@ac_config_macro_dirs' instead of
re-defining '$ac_config_macro_dir'. Cater to calls the now-preferred
macro 'AC_CONFIG_MACRO_DIRS' in addition to the "obsolescent" one
AC_CONFIG_MACRO_DIR.
(main loop): Append '@ac_config_macro_dirs', not '$ac_config_macro_dir',
to '@user_includes'.
* t/subpkg-macrodir.sh: New test.
* t/aclocal-macrodirs.tap: Likewise.
* t/list-of-tests.mk: Add them.
* t/aclocal-macrodir.tap: Adjust and extend a little to keep it more in
sync with 'aclocal-macrodirs.tap'.
tests: remove mostly-redundant tests on silent rules
We used to have several couples of tests named like 'silent-foo-gcc.sh'
and 'silent-foo-generic.sh'. Differently from what the names suggest,
the first test in such a couple (that is, "silent-foo-gcc.sh") was not
meant to check specific GCC-related features, but rather to check how
the silent rules behave in combination with automatic dependency tracking
when the 'gcc' depmode (that targets GCC versions before the 3.x and 4.x
release series) is forced.
Such depmode forcing was done exclusively to cover the code paths in
'lib/am/depend2.am' that actually invoke the 'depcomp' script, rather
than using the inlined, GCC-specific compiler invocation (the so-called
"fastdep" mode), which is the default with modern GCC or with other
modern compilers that can emulate the GCC command-line interface (e.g.,
clang 3.0).
But whenever we run the silent-*-generic.sh" tests with other supported
compilers which have an associated depmode different from 'gcc3', these
same "non-fastdep" code paths are covered, since in those tests we run
./configure with the '--enable-dependency-tracking' option, which causes
slower depmodes not to be rejected. Examples of such compilers are the
Sun C and C++ compilers (at least since version 5.9, a.k.a. Sun Studio
12.1), and the Tiny C Compiler (from version 0.9.26); and I run the
Automake testsuite quite regularly with those compilers.
So, the "silent-*-gcc.sh" test cases don't offer any real coverage
enhancements, while still using testsuite runtime and causing some
(admittedly minor, but still annoying) synchronization headaches with
the sister tests "silent-foo-general.sh" tests. So let's just remove
these "silent-*-gcc.sh" tests.
* t/silent-c-gcc.sh: Remove.
* t/silent-cxx-gcc.sh: Likewise.
* t/silent-lt-gcc.sh: Likewise.
* t/silent-many-gcc.sh: Likewise.
* t/silent-c-generic.sh: Rename ...
* t/silent-c.sh: ... like this, and adjust heading comments.
* t/silent-cxx-generic.sh: Rename ...
* t/silent-cxx.sh: ... like this, and adjust heading comments.
* t/silent-lt-generic.sh: Rename ...
* t/silent-lt.sh: ... like this, and adjust heading comments.
* t/silent-many-generic.sh: Rename ...
* t/silent-many-languages.sh: ... like this, and adjust heading
comments.
* t/list-of-tests.mk: Adjust.
* master:
tests: enhance coverage on silent-rules and C compilation rules a little
tests: new variable $am_testaux_builddir
tests: rename $am_testauxdir -> $am_testaux_srcdir
* master:
tests: rename few tests (on silent-rules support)
tests: merge two tests on silent-rules with libtool
vala tests: source test-init.sh, not ./defs
tests: fix a spurious typo-related failure
tests: remove spurious leftover use of 'Exit'
tests: can check our recipes avoid trailing backslashes
vala: improve comments to AM_PROG_VALAC
news: update w.r.t. recent vala changes
vala: if no proper compiler found, set $(VALAC) to 'valac'
vala: AM_PROG_VALAC should not produce an error for tool-old valac
docs: document recent changes to AM_PROG_VALAC
tests: enhance tests on AM_PROG_VALAC
vala: style fixes in vala.m4
vala: add action arguments, for when no proper vala compiler is found
tests: merge two tests on silent-rules with libtool
This helps to reduce code duplication, and also to (marginally)
speed up the involved tests.
Before this change, we had:
$ diff -u t/silent3.sh t/silent9.sh
--- t/silent3.sh 2012-11-07 19:38:05.000000000 +0100
+++ t/silent9.sh 2012-11-07 19:27:54.000000000 +0100
@@ -14,9 +14,9 @@
# You should have received a copy of the GNU General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
-# Check silent-rules mode, with libtool, standard depmode case.
+# Check silent-rules mode, with libtool, nodep case.
-# Please keep this file in sync with 'silent4.sh' and 'silent9.sh'.
+# Please keep this file in sync with 'silent3.sh' and 'silent4.sh'.
And a related fix that solves a spurious testsuite failures in
VPATH builds.
* t/ax/test-defs.in ($am_testaux_builddir): New, counterpart of
$am_testaux_srcdir, but pointing inside the build directory.
* t/self-check-shell-no-trail-bslash.sh: Use $am_testaux_builddir,
not $am_testaux_srcdir, when fetching the 'shell-no-trail-bslash'
script.
This is just a preparatory change in view of a future commit.
* t/ax/test-defs.in: Here.
* t/ax/tap-summary-aux.sh: And here.
* t/ax/testsuite-summary-checks.sh: And here.
* t/distcheck-missing-m4.sh: And here.
* t/distcheck-outdated-m4.sh: And here.
* t/self-check-shell-no-trail-bslash.sh: And here.
* t/test-driver-acsubst.sh: And here.
* t/test-driver-cond.sh: And here.
* t/test-driver-custom-multitest.sh: And here.
* t/test-driver-custom-multitest-recheck.sh: And here.
* t/test-driver-custom-multitest-recheck2.sh: And here.
* t/testsuite-summary-count-many.sh: And here.
* maint:
vala tests: source test-init.sh, not ./defs
tests: fix a spurious typo-related failure
tests: remove spurious leftover use of 'Exit'
tests: can check our recipes avoid trailing backslashes
vala: improve comments to AM_PROG_VALAC
news: update w.r.t. recent vala changes
vala: if no proper compiler found, set $(VALAC) to 'valac'
vala: AM_PROG_VALAC should not produce an error for tool-old valac
docs: document recent changes to AM_PROG_VALAC
tests: enhance tests on AM_PROG_VALAC
vala: style fixes in vala.m4
vala: add action arguments, for when no proper vala compiler is found
* vala-work:
vala: improve comments to AM_PROG_VALAC
news: update w.r.t. recent vala changes
vala: if no proper compiler found, set $(VALAC) to 'valac'
vala: AM_PROG_VALAC should not produce an error for tool-old valac
docs: document recent changes to AM_PROG_VALAC
tests: enhance tests on AM_PROG_VALAC
vala: style fixes in vala.m4
vala: add action arguments, for when no proper vala compiler is found
tests: can check our recipes avoid trailing backslashes
This is related to commit v1.11-1704-g254227b of 2012-05-01,
"parallel-tests: avoid trailing backslashes in make recipes",
and automake bug#10436.
Recipes with a trailing backslash character (possibly followed by
blank characters only) can cause spurious syntax errors with at
least older bash versions (e.g., bash 2.05b), and can be potentially
be unportable to other weaker shells.
So provide a target that runs the testsuite looking for this kind
of breakage (without requiring a real bugged shell).
* t/ax/shell-no-trail-bslash.in: New, a "shell" that chokes on '-c'
commands having a trailing '\' (possibly followed by whitespace only).
* Makefile (t/ax/shell-no-trail-bslash): Generate this script from it.
(noinst_SCRIPTS, CLEANFILES): Add it.
(EXTRA_DIST): Add 't/ax/shell-no-trail-bslash.in'.
(check-no-trailing-backslash-in-recipes): New target, runs the testsuite
with 'shell-no-trail-bslash' as the CONFIG_SHELL, to catch possible
recipes having a trailing backslash character (possibly followed by
* .gitignore: Update.
* t/self-check-shell-no-trail-bslash.sh: New testsuite self-check.
* t/parallel-tests-trailing-bslash.sh: Remove as obsolete.
* t/list-of-tests.mk: Adjust.
projects / thirdparty / automake.git / log
