platform/upstream/automake.git
12 years agoMerge branch 'maint'
Stefano Lattarini [Fri, 18 May 2012 12:31:36 +0000 (14:31 +0200)]
Merge branch 'maint'

* maint:
  tests: use 'parallel-tests' Automake option by default
  doc: fix location of Automake's own test-suite.log file
  yacc tests: avoid a spurious failure with Sun C++ 5.9
  self tests: cater for /bin/ksh symlinked to Zsh
  tests: fix spurious failures due to dpkg install-info

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agotests: use 'parallel-tests' Automake option by default
Stefano Lattarini [Tue, 15 May 2012 14:12:09 +0000 (16:12 +0200)]
tests: use 'parallel-tests' Automake option by default

This will help our transition from 'serial-tests' to 'parallel-tests'
as the default test suite driver enabled by a TESTS assignment in the
input Makefile.am.  Note that that change of default will only take
place in master, though.

* defs: Pass the 'parallel-tests' option to the AM_INIT_AUTOMAKE
invocation in the created 'configure.ac' stub, unless the variable
'am_serial_tests' is set to "yes".  Don't pay attention anymore to
the 'am_parallel_tests' variable, that's obsolete now.
* defs-static.in: Warn if the 'am_serial_tests' variable is set in the
environment; conversely, don't warn anymore about 'am_parallel_tests'
being set in the environment.
* Makefile.am (AM_TESTS_ENVIRONMENT): Nullify the 'am_serial_tests'
variable instead of the now-obsolete 'am_parallel_tests' one.
* syntax-checks.mk (sc_tests_obsolete_variables): Also warn against
uses of 'am_parallel_tests', which is now deprecated in favor of
'am_serial_tests'.  Similarly, if a use of 'parallel_tests' is seen,
suggest using 'am_serial_tests' instead, not 'am_parallel_tests'.
* gen-testsuite-part: Now that we use the 'parallel-tests' by default
in our tests, we need to completely change the logic and semantics of
generation of sibling tests for those tests that check the Automake
generated testsuite harness itself.  Do that, and give a complete
explanation of the new logic and semantics in the relevant comments.
* t/README: Update.
* Lots of test cases: Adjust.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agodoc: fix location of Automake's own test-suite.log file
Stefano Lattarini [Thu, 17 May 2012 10:36:23 +0000 (12:36 +0200)]
doc: fix location of Automake's own test-suite.log file

* doc/automake.texi (Reporting Bugs): The global testsuite log of
Automake is now (basically since we have removed make recursion from
the Automake's build system) saved in the file 'test-suite.log', not
in the file 'tests/test-suite.log'.  Adjust accordingly.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoyacc tests: avoid a spurious failure with Sun C++ 5.9
Stefano Lattarini [Wed, 16 May 2012 14:26:39 +0000 (16:26 +0200)]
yacc tests: avoid a spurious failure with Sun C++ 5.9

* t/yacc-cxx.sh (parse1.yy): When calling the 'getchar' function declared
in the #included <cstdio>, fully qualify it as std::getchar', to avoid
Sun C++ 5.9 erroring out with:
"parse1.yy", line 5: Error: The function "getc" must have a prototype.
* t/yacc-clean-cxx.sh (sub1/parsefoo.yxx): Likewise.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoself tests: cater for /bin/ksh symlinked to Zsh
Stefano Lattarini [Wed, 16 May 2012 14:10:06 +0000 (16:10 +0200)]
self tests: cater for /bin/ksh symlinked to Zsh

* t/self-check-reexec.tap: When searching for a suitable non-Bash
shells, be sure to reject any shell that is Zsh "in disguise" (as
can be found on some Debian systems, where /bin/ksh can be symlinked
to /bin/zsh4).  This is required because our testsuite does not
support older versions of Zsh, and that was causing the test to fail
in the setup just described.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agotests: fix spurious failures due to dpkg install-info
Stefano Lattarini [Wed, 16 May 2012 13:50:14 +0000 (15:50 +0200)]
tests: fix spurious failures due to dpkg install-info

* t/install-info-dir.sh: Skip some checks if 'install-info' is the one
from dpkg, not the one from GNU info; the former might try to create
files in '/var/backups/', causing spurious failures like this for any
non-root user:

    install-info(.../install-info-dir.dir/_inst/info/foo.info): \
      creating new section `Dummy utilities'
    cp: cannot create regular file `/var/backups/infodir.bak': \
      Permission denied
    install-info(.../install-info-dir.dir/_inst/info/foo.info): \
      could not backup .../install-info-dir.dir/_inst/info/dir in \
      /var/backups/infodir.bak: No such file or directory

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'maint'
Stefano Lattarini [Sun, 13 May 2012 12:10:25 +0000 (14:10 +0200)]
Merge branch 'maint'

* maint:
  tests: fix spurious failure due to i18n issue
  tests: add few missing 'cc' and 'c++' requirements

12 years agotests: fix spurious failure due to i18n issue
Stefano Lattarini [Sat, 12 May 2012 06:42:37 +0000 (08:42 +0200)]
tests: fix spurious failure due to i18n issue

Fixes automake bug#11452.

* t/parallel-tests-fork-bomb.sh: Export LANG, LANGUAGE and LC_ALL to
'C', to ensure error messages will be in English, and thus avoiding
possible false negatives in our grepping of them.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agotests: add few missing 'cc' and 'c++' requirements
Stefano Lattarini [Thu, 10 May 2012 12:32:27 +0000 (14:32 +0200)]
tests: add few missing 'cc' and 'c++' requirements

Without that requirements, the configure-time overrides of CC and CXX
wouldn't be honored in the affected tests.

* t/silent-yacc.sh ($required): Require 'cc'.
* t/silent-yacc-headers.sh: Likewise.
* t/silent-lex.sh: Likewise.
* t/lex-clean.sh: Likewise.
* t/lex-depend.sh: Likewise.
* t/lex-line.sh: Likewise.
* t/lex-depend.sh: Likewise.
* t/lex-pr204.sh: Likewise.
* t/yacc-deleted-headers.sh: Likewise.
* t/yacc-line.sh: Likewise.
* t/yacc-depend.sh: Likewise.
* t/yacc-depend2.sh: Likewise.
* t/yacc-dist-nobuild-subdir.sh: Likewise.
* t/yacc-bison-skeleton.sh: Likewise.
* t/yacc-bison-skeleton-cxx.sh ($required): Require 'c++'.
* t/yacc-clean-cxx.sh: Likewise.
* t/yacc-d-cxx.sh: Likewise.
* t/yacc-cxx.sh: Likewise.
* t/yacc-mix-c-cxx.sh ($required): Require 'cc' and 'c++'.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agofixup: s/AM_CONFIG_HEADER/AC_CONFIG_HEADERS/ in 't/confsub.sh'
Stefano Lattarini [Tue, 8 May 2012 11:22:59 +0000 (13:22 +0200)]
fixup: s/AM_CONFIG_HEADER/AC_CONFIG_HEADERS/ in 't/confsub.sh'

Issue introduced in the recent merge 'v1.12-101-g12405c9'.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'maint'
Stefano Lattarini [Tue, 8 May 2012 10:49:36 +0000 (12:49 +0200)]
Merge branch 'maint'

* maint:
  tests: fix a spurious failure with dash
  test defs: fix indentation (cosmetic change)
  tests: remove obsolete uses of $sh_errexit_works
  configure: search a sturdy POSIX shell to be used in the testsuite
  tests: shell running test scripts is now named AM_TEST_RUNNER_SHELL
  fixup: distribute t/README
  init: warn against obsolete usage of AM_INIT_AUTOMAKE

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'testsuite-saner-shell' into maint
Stefano Lattarini [Mon, 7 May 2012 20:11:28 +0000 (22:11 +0200)]
Merge branch 'testsuite-saner-shell' into maint

* testsuite-saner-shell:
  tests: fix a spurious failure with dash
  test defs: fix indentation (cosmetic change)
  tests: remove obsolete uses of $sh_errexit_works
  configure: search a sturdy POSIX shell to be used in the testsuite
  tests: shell running test scripts is now named AM_TEST_RUNNER_SHELL

12 years agotests: fix a spurious failure with dash
Stefano Lattarini [Mon, 23 Apr 2012 17:03:33 +0000 (19:03 +0200)]
tests: fix a spurious failure with dash

The dash shell, at least version 0.5.5.1, doesn't always bail out
with a syntax error when a stray "fi" in encountered:

  $ dash -c ":; fi"; echo stat = $?
  stat = 0

See also the relevant bug report:

  <http://permalink.gmane.org/gmane.comp.shells.dash/717>

This behaviour was causing a spurious error in our testsuite.  Fix it.

* t/self-check-exit.tap: Use a stray parentheses rather than a stray
'fi' to trigger a syntax error.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agotest defs: fix indentation (cosmetic change)
Stefano Lattarini [Wed, 28 Dec 2011 12:23:31 +0000 (13:23 +0100)]
test defs: fix indentation (cosmetic change)

* defs (exit trap): Fix indentation left botched by previous
patch 'v1.11b-51-g626bf65'.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agotests: remove obsolete uses of $sh_errexit_works
Stefano Lattarini [Wed, 28 Dec 2011 12:23:31 +0000 (13:23 +0100)]
tests: remove obsolete uses of $sh_errexit_works

After the last changes, configure will ensure that the shell
selected to run the test scripts can correctly propagate exit
status to the exit trap when 'set -e' is in effect.

* configure.ac (sh_errexit_works): Do not AC_SUBST it anymore.
* defs-static.in (sh_errexit_works): Do not initialize
from the AC_SUBST value anymore.
* defs (trap): Trap the EXIT signal unconditionally.
* t/self-check-explicit-skips.sh: Do not skip the test if
'$sh_errexit_works' is != "yes", this check doesn't make
sense anymore.
* t/self-check-cleanup.tap: Likewise.
* t/self-check-exit.tap: Assume the exit trap is always
installed by ./defs.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoconfigure: search a sturdy POSIX shell to be used in the testsuite
Stefano Lattarini [Wed, 28 Dec 2011 12:23:31 +0000 (13:23 +0100)]
configure: search a sturdy POSIX shell to be used in the testsuite

* configure.ac: Add code (partially inspired to checks in gnulib's
'tests/init.sh') to search for a good-enough, not-buggy POSIX/XSI
shell to be used in our testsuite.  Accordingly AC_SUBSTitute the
variable 'AM_TEST_RUNNER_SHELL'.
* NEWS: Update.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agotests: shell running test scripts is now named AM_TEST_RUNNER_SHELL
Stefano Lattarini [Wed, 28 Dec 2011 12:23:31 +0000 (13:23 +0100)]
tests: shell running test scripts is now named AM_TEST_RUNNER_SHELL

This is just a preparatory refactoring for future changes.

* configure.ac (AM_TEST_RUNNER_SHELL): New variable, defined
to $SHELL', and AC_SUBST'd.
* Makefile.am (LOG_COMPILER): Redefine from $(SHELL) to
$(AM_TEST_RUNNER_SHELL).
* defs: Adjust to use $AM_TEST_RUNNER_SHELL instead
of $CONFIG_SHELL and/or $SHELL, where appropriate.  Minor
opportunistic changes.
* t/self-check-explicit-skips.sh: Likewise.
* t/self-check-sanity.sh: Likewise.
* t/self-check-tap.sh: Likewise.
* t/self-check-cleanup.tap: Likewise.
* t/self-check-dir.tap: Likewise.
* t/self-check-env-sanitize.tap: Likewise.
* t/self-check-exit.tap: Likewise.
* t/self-check-me.tap: Likewise.
* t/self-check-reexec.tap: Likewise.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agofixup: distribute t/README
Stefano Lattarini [Sun, 6 May 2012 11:33:09 +0000 (13:33 +0200)]
fixup: distribute t/README

Since commit v1.11-2128-g8e02441 of 2012-04-06, "maint: no more make
recursion in Automake's build system" we haven't been distributing the
testsuite README file in our release tarballs.  Oops.

* Makefile.am (EXTRA_DIST): Add 't/README'.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoinit: warn against obsolete usage of AM_INIT_AUTOMAKE
Stefano Lattarini [Sat, 31 Mar 2012 14:39:32 +0000 (16:39 +0200)]
init: warn against obsolete usage of AM_INIT_AUTOMAKE

Support for the two- and three-arguments invocation forms of the
AM_INIT_AUTOMAKE macro, as in:

  AM_INIT_AUTOMAKE($PACKAGE, $VERSION)

or:

  AM_INIT_AUTOMAKE($PACKAGE, $VERSION, NODEFINE)

will be removed in the next major Automake release (1.13).

Such usages have already been deprecated in the documentation
starting from commit v1.11-2015-ge99690a of 23-02-2012 "docs,
news: document planned removal of obsolete macros and features".

We now start giving runtime warnings as well (in the 'obsolete'
category).

* NEWS: Update.
* m4/init.m4 (AM_INIT_AUTOMAKE): Report the two- and three-arguments
form invocation.
* automake.in (scan_autoconf_traces): Likewise.
* doc/automake.texi: Minor adjustments.  Add an @anchor to the
location where it's described how to modernize outdated invocation
of AM_INIT_AUTOMAKE, so that it can be referenced from automake
warning/error messages.
* t/aminit-moreargs-deprecation.sh: New test.
* tests/list-of-tests.mk: Add it.
* tests/ac-output-old.tap: Adjust by calling automake with the
warnings in the 'obsolete' category disabled.
* t/backcompat.test: Likewise.
* t/backcompat3.test: Likewise.
* t/backcompat5.test: Likewise.
* t/backcompat6.test: Likewise.
* t/version.test: Likewise.
* t/version2.test: Likewise.
* t/pr2.test: Modernize style of AC_INIT and AM_INIT_AUTOMAKE
invocations, and use proper m4 quoting.
* t/pr87.test: Likewise.
* t/confsub.test: Likewise.
* t/install2.test: Likewise.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'maint'
Stefano Lattarini [Sun, 6 May 2012 09:04:45 +0000 (11:04 +0200)]
Merge branch 'maint'

* maint:
  m4: prepend m4 builtins with "m4_"
  docs: fix clumsy grammar in the scripts-based testsuite chapter

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agom4: prepend m4 builtins with "m4_"
Stefano Lattarini [Sun, 6 May 2012 08:15:59 +0000 (10:15 +0200)]
m4: prepend m4 builtins with "m4_"

The comments in 'lib/autoconf/autoconf.m4' from Autoconf 2.69 read:

  We discourage the use of the non prefixed macro names: M4sugar maps
  all the builtins into 'm4_'.  Autoconf has been converted to these
  names too.  But users may still depend upon these, so reestablish
  them.

Alas, in our init.m4 files, we were still using m4 builtins without
prefixing them with "m4_".  Not nice for a package that is one of the
major clients of Autoconf, and an integral part of the GNU Autotools!

Fix the issue.

* m4/init.m4 (AM_INIT_AUTOMAKE): Use 'm4_define' and 'm4_defn' instead
of 'define' and 'defn'.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agorequire: autoconf >= 2.65; related simplifications
Stefano Lattarini [Sat, 5 May 2012 19:48:49 +0000 (21:48 +0200)]
require: autoconf >= 2.65; related simplifications

The new support for Objective C++ (to be added in Automake 1.12.1) can be
simplified if we assume Autoconf version 2.65 or later (2.65 being the
version that introduced support for Objective C++).  Since such an Autoconf
version is two and half years old now (and will likely be almost three
years old when Automake 1.13 gets released), requiring it is acceptable.
This will also simplify testing and maintenance of Automake, because we'll
need to test with fewer Autoconf version, and possibly rely on new Autoconf
features.

* NEWS (Version requirements): Automake 1.13 will require Autoconf
2.65 or later.
* configure.ac ($required_autoconf_version): Bump to 2.65.
* m4/init.m4 (AM_INIT_AUTOMAKE): Require Autoconf >= 2.65.
Assume AC_PROG_OBJCXX is unconditionally defined.
* m4/depout.m4: Adjust a comment about a nugget of defensive programming.
* t/ext.sh: Assume that AC_PROG_OBJCXX is unconditionally defined,
and that we are using Autoconf >= 2.65.  Related simplifications.
* t/objc-megademo.sh: Likewise.
* t/objcxx-basic.sh: Likewise.
* t/objcxx-deps.sh: Likewise.
* t/objcxx-flags.sh: Likewise.
* t/objcxx-minidemo.sh: Likewise.
* t/nodep2.sh: Likewise.
* t/backcompat3.sh: Assume AC_INIT accepts an URL argument.
* t/depend5.sh: Adjust a comment.
* syntax-checks.mk (sc_test_names): Delete this check as now useless:
autoconf >= 2.65 (>= 2.63, actually) can handle arguments to AC_INIT
that contain or are m4 builtins or predefined macros.
(m4_builtin): Delete this now-useless variable as well, was used only
by the check above.
(syntax_check_rules): Update.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agodocs: fix clumsy grammar in the scripts-based testsuite chapter
Nick Alcock [Fri, 4 May 2012 14:14:43 +0000 (15:14 +0100)]
docs: fix clumsy grammar in the scripts-based testsuite chapter

* doc/automake.texi (Scripts-based Testsuites): "We'll have later" is
a rare example of English in the Automake manual clearly not written
by a native English-speaker: while comprehensible, it can be better
formulated.

Copyright-paperwork-exempt: yes
Signed-off-by: Nick Alcock <nick.alcock@oracle.com>
Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'maint'
Stefano Lattarini [Sat, 5 May 2012 17:28:35 +0000 (19:28 +0200)]
Merge branch 'maint'

* maint:
  tests: use append mode to capture parallel make output
  parallel-tests: separate different logs with an empty line

12 years agotests: use append mode to capture parallel make output
Stefano Lattarini [Sat, 5 May 2012 17:22:40 +0000 (19:22 +0200)]
tests: use append mode to capture parallel make output

This should fix the spurious failure reported in automake bug#11413.

This is due to the fact that the redirected output of parallel make
can racily loose lines.  For example, if GNU make (3.82) is run on
a Makefile like this:

  all = 0 1 2 3 4 5 6 7 8 9
  default: $(all)
  $(all):
        @sleep 0.$$(($RANDOM % 10)); echo $@

and has its standard output redirected in overwrite-mode to a regular
file, it looses a line of that output every 15 runs or so on a Fedora
17 system with 64 ppc64 cores and Linux 3.3.1.  Redirection in append
mode does not suffer of this problem.

See also similar commit 'Release-1-10-280-g6426999' of 2009-03-10,
"Use append mode to capture parallel test output", which tackled a
similar problem for 't/parallel-tests3.sh' and 't/lisp8.sh'.

* t/tap-more.sh: Use append mode for output from "make -j", to avoid
dropped lines.
* t/parallel-tests3.sh, t/lisp8.sh: Enhance comments.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'parallel-tests-performance' into maint
Stefano Lattarini [Sat, 5 May 2012 14:39:10 +0000 (16:39 +0200)]
Merge branch 'parallel-tests-performance' into maint

* parallel-tests-performance:
  parallel-tests: separate different logs with an empty line

12 years agoparallel-tests: separate different logs with an empty line
Stefano Lattarini [Sat, 5 May 2012 14:33:36 +0000 (16:33 +0200)]
parallel-tests: separate different logs with an empty line

After commit v1.12-21-g5eeb366 "parallel-tests: optimize global log
creation", the log files copied in the global log are not anymore
separated by a blank line, causing unclear output like this:

  FAIL: test1
  ===========

  output from test 1
  FAIL: test2
  ===========

  output from test 2

where we would want something like this:

  FAIL: test1
  ===========

  output from test 1

  FAIL: test2
  ===========

  output from test 2

Fix this regression.

* lib/am/check.am (am__create_global_log): Print an extra empty line
after having displayed the content of a '.log' file.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'maint'
Stefano Lattarini [Sat, 5 May 2012 08:57:27 +0000 (10:57 +0200)]
Merge branch 'maint'

* maint:
  maintcheck: avoid a couple of spurious failures
  maintcheck: reduce code duplication, increase coverage

12 years agomaintcheck: avoid a couple of spurious failures
Stefano Lattarini [Fri, 4 May 2012 16:09:02 +0000 (18:09 +0200)]
maintcheck: avoid a couple of spurious failures

* t/objcxx-deps.sh, t/objcxx-minidemo.sh: In C++ files, use
"std::cout <<", not "cout <<", to avoid triggering a spurious
failure by the syntax check 'sc_tests_here_document_format'.

Cherry-picked from yesterday's commit v1.12-81-g7f3f467, that
had erroneously been applied to 'master' only rather than to
'maint'.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agomaintcheck: reduce code duplication, increase coverage
Stefano Lattarini [Sat, 5 May 2012 00:01:20 +0000 (02:01 +0200)]
maintcheck: reduce code duplication, increase coverage

* syntax-checks.mk: Move ...
(sc_tests_plain_aclocal,
sc_tests_plain_autoconf,
sc_tests_plain_perl,
sc_tests_plain_autoupdate,
sc_tests_plain_automake,
sc_tests_plain_make,
sc_tests_plain_autoheader,
sc_tests_plain_autoreconf,
sc_tests_plain_autom4te): ... all these targets ...
(sc_tests_plain_egrep_fgrep): ... and part of this target ...
(sc_tests_plain_check_rules): ... into this variable, and
reimplement their recipes ...
($(sc_tests_plain_check_rules)): ... with this static pattern
rule.  Enhance the coverage they offer a little since we are
at it.
(sc_m4_am_plain_egrep_fgrep): New rule, the part of the old
'sc_tests_plain_egrep_fgrep' rule that checked for use of
'fgrep' and 'egrep' in *.am and *.m4 fragments (rather than
in test scripts).
(syntax_check_rules): Update.
* t/dist-formats.tap: Fix bad use of "make" instead of "$MAKE"
revealed by the extra coverage.
* t/cond5.sh, t/auxdir-autodetect.sh: Cosmetic changes to avoid
spuriously triggering the maintainer checks due to the
enhanced coverage.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agomaintcheck: avoid a couple of spurious failures
Stefano Lattarini [Fri, 4 May 2012 16:09:02 +0000 (18:09 +0200)]
maintcheck: avoid a couple of spurious failures

* t/objcxx-deps.sh, t/objcxx-minidemo.sh: In C++ files, use
"std::cout <<", not "cout <<", to avoid triggering a spurious
failure by the syntax check 'sc_tests_here_document_format'.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'maint'
Stefano Lattarini [Fri, 4 May 2012 15:46:51 +0000 (17:46 +0200)]
Merge branch 'maint'

* maint:
  tests: avoid spurious failure on missing Obj C/C++ compiler
  objc, objc++: test support for compilation flags
  objc, objc++: add stress test
  objc, objc++: test automatic dependency tracking
  objc: reorganize basic tests
  objc++, objc: add first semantic tests
  objc++: add first basic test
  news: announce initial support for Objective C++
  objc++: test support for '.mm' suffix in _SOURCES entries
  objc++: add documentation
  objc++: initial support for Objective C++

12 years agoMerge branch 'objc++' into maint
Stefano Lattarini [Fri, 4 May 2012 15:32:08 +0000 (17:32 +0200)]
Merge branch 'objc++' into maint

* objc++:
  tests: avoid spurious failure on missing Obj C/C++ compiler

12 years agotests: avoid spurious failure on missing Obj C/C++ compiler
Stefano Lattarini [Fri, 4 May 2012 15:31:06 +0000 (17:31 +0200)]
tests: avoid spurious failure on missing Obj C/C++ compiler

* t/objc-megademo.sh: Arrange the test to SKIP, not FAIL, if there is
no Objective C or no Objective C++ compiler.  Since we are at it, remove
some checks that were actually testing Autoconf (not Automake) behaviour.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'objc++' into maint
Stefano Lattarini [Fri, 4 May 2012 11:24:23 +0000 (13:24 +0200)]
Merge branch 'objc++' into maint

* objc++:
  objc, objc++: test support for compilation flags
  objc, objc++: add stress test
  objc, objc++: test automatic dependency tracking
  objc: reorganize basic tests
  objc++, objc: add first semantic tests
  objc++: add first basic test
  news: announce initial support for Objective C++
  objc++: test support for '.mm' suffix in _SOURCES entries
  objc++: add documentation
  objc++: initial support for Objective C++

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agodist: don't bother putting README first in $(DIST_COMMON)
Stefano Lattarini [Thu, 3 May 2012 15:30:56 +0000 (17:30 +0200)]
dist: don't bother putting README first in $(DIST_COMMON)

Comments on some of our automake-time pre-processing of $(DIST_COMMON)
said that it was done in order to "put README first because it then
becomes easier to make a Usenet-compliant shar file".  But such a
format is hardly relevant anymore, and not worth the (albeit small)
added complexity.

* automake.in (handle_dist): Don't sort @dist_common.
(for_dist_common): Delete this function, is not used anymore.
* lib/am/distdir.am (DISTFILES): Remove obsolete comment.
* t/distcom4.sh: Adjust.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'always-support-silent-rules'
Stefano Lattarini [Wed, 2 May 2012 16:15:41 +0000 (18:15 +0200)]
Merge branch 'always-support-silent-rules'

* always-support-silent-rules:
  silent rules: support for them is always active now

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'maint'
Stefano Lattarini [Wed, 2 May 2012 16:14:55 +0000 (18:14 +0200)]
Merge branch 'maint'

* maint:
  news: document improvement in parallel-tests performance
  tests: fix spurious failure due to autom4te caching
  news: fix typos
  aclocal: error out again on unrecognized arguments
  parallel-tests: save few forks when possible
  parallel-tests: optimize global log creation
  parallel-tests: optimize 'recheck' target for speed
  coverage: add performance tests on some parallel-tests aspects

12 years agonews: document improvement in parallel-tests performance
Stefano Lattarini [Wed, 2 May 2012 16:04:25 +0000 (18:04 +0200)]
news: document improvement in parallel-tests performance

* NEWS: Document the improved performances of the parallel-tests targets
'check' and 'recheck'. See recent commits:

  - v1.12-20-gf5733aa of 2012-04-26,
    "parallel-tests: optimize 'recheck' target for speed"

  - v1.12-21-g5eeb366 of 2012-02-26,
    "parallel-tests: optimize global log creation"

  - v1.12-22-g4c2dd6e of 2012-04-30,
    "parallel-tests: save few forks when possible"

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years ago tests: fix spurious failure due to autom4te caching
Stefano Lattarini [Wed, 2 May 2012 15:47:56 +0000 (17:47 +0200)]
 tests: fix spurious failure due to autom4te caching

Reported by Peter Breitenlohner :
<http://lists.gnu.org/archive/html/automake-patches/2012-05/msg00045.html>

* t/suffix.sh: Remove stale autom4te.cache directories, to prevent racy,
spurious failures (using 'aclocal --force' was not enough, since the cache
was still picked up by the following automake call).

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoobjc, objc++: test support for compilation flags
Stefano Lattarini [Tue, 1 May 2012 13:04:37 +0000 (15:04 +0200)]
objc, objc++: test support for compilation flags

* t/objc-flags.sh, t/objcxx-flags.sh: New tests.
* t/list-of-tests.mk: Add them.

Co-authored-by: Peter Breitenlohner <peb@mppmu.mpg.de>
Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoobjc, objc++: add stress test
Peter Breitenlohner [Mon, 30 Apr 2012 14:11:03 +0000 (16:11 +0200)]
objc, objc++: add stress test

* t/objc-megademo.sh: New test, trying out a package using all
of C, C++, Objective C and Objective C++ at the same time.
* t/list-of-tests.mk: Add it.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoobjc, objc++: test automatic dependency tracking
Stefano Lattarini [Mon, 30 Apr 2012 13:17:13 +0000 (15:17 +0200)]
objc, objc++: test automatic dependency tracking

* t/objc-deps.sh, t/objcxx-deps.sh: New tests.
* t/list-of-tests.mk: Add them.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoobjc: reorganize basic tests
Stefano Lattarini [Mon, 30 Apr 2012 12:30:42 +0000 (14:30 +0200)]
objc: reorganize basic tests

* t/objc.sh, t/objc2.sh: Removed, merged into ...
* t/objc-basic.sh: ... this new test.
* t/objcxx-basic.sh: Add reference to new sister test.
* t/list-of-tests.mk: Update.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoobjc++, objc: add first semantic tests
Stefano Lattarini [Mon, 30 Apr 2012 12:17:44 +0000 (14:17 +0200)]
objc++, objc: add first semantic tests

* t/objcxx-minidemo.sh: New test.
* t/objc-minidemo.sh: Likewise.
* t/list-of-tests.mk: Add them.

Co-authored-by: Peter Breitenlohner <peb@mppmu.mpg.de>
Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoobjc++: add first basic test
Peter Breitenlohner [Thu, 26 Apr 2012 23:00:27 +0000 (01:00 +0200)]
objc++: add first basic test

* t/objcxx-basic.sh: New test.
* t/list-of-tests.mk: Add it.

Signed-off-by: Peter Breitenlohner <peb@mppmu.mpg.de>
Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'parallel-tests-performance' into maint
Stefano Lattarini [Wed, 2 May 2012 10:51:09 +0000 (12:51 +0200)]
Merge branch 'parallel-tests-performance' into maint

* parallel-tests-performance:
  parallel-tests: save few forks when possible
  parallel-tests: optimize global log creation
  parallel-tests: optimize 'recheck' target for speed
  coverage: add performance tests on some parallel-tests aspects

12 years agonews: fix typos
Stefano Lattarini [Wed, 2 May 2012 07:48:11 +0000 (09:48 +0200)]
news: fix typos

* NEWS (Bugs fixed in 1.12.1): Fix botched indentation of an entry,
add a forgotten "and", remove a redundant "the".

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoaclocal: error out again on unrecognized arguments
Stefano Lattarini [Tue, 1 May 2012 22:13:10 +0000 (00:13 +0200)]
aclocal: error out again on unrecognized arguments

Starting from commit v1.11-662-g52246cc of 2102-02-18, "cmdline parsing:
move into a dedicated perl module", aclocal has been silently ignoring
non-option arguments instead of correctly reporting them.
Fix this regression.

* t/aclocal.sh: Update to catch the regression.
* aclocal.in (parse_arguments): Explicitly reject non-option arguments.
* NEWS: Update.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'cygnus-remove'
Stefano Lattarini [Wed, 2 May 2012 08:52:18 +0000 (10:52 +0200)]
Merge branch 'cygnus-remove'

* cygnus-remove:
  cygnus: remove support for Cygnus-style trees

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'maint'
Stefano Lattarini [Wed, 2 May 2012 08:26:12 +0000 (10:26 +0200)]
Merge branch 'maint'

* maint:
  cosmetics: fix typo in aclocal m4 directory README
  cygnus: warn when it's used (flagged as 'obsolete' now)

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'cygnus-deprecate' into maint
Stefano Lattarini [Wed, 2 May 2012 07:59:46 +0000 (09:59 +0200)]
Merge branch 'cygnus-deprecate' into maint

* cygnus-deprecate:
  cygnus: warn when it's used (flagged as 'obsolete' now)

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agocosmetics: fix typo in aclocal m4 directory README
Andrew Eikum [Tue, 1 May 2012 14:18:56 +0000 (09:18 -0500)]
cosmetics: fix typo in aclocal m4 directory README

* m4/acdir/README: Add forgotten "if".
* THANKS: Update.

Copyright-paperwork-exempt: yes
Signed-off-by: Andrew Eikum <aeikum@codeweavers.com>
Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agosilent rules: support for them is always active now
Stefano Lattarini [Tue, 1 May 2012 12:23:04 +0000 (14:23 +0200)]
silent rules: support for them is always active now

Before this change, support for silent rules was optional and activated
only if the 'silent-rules' option was specified.

The rationale behind that behaviour was that the silent-rules machinery
originally only worked with make implementations supporting nested
variables expansions, which isn't (or wasn't) a POSIX-specified features;
so that packages wanting to be "extra-portable" couldn't use the
silent-rules machinery at all.

But after Paul Eggert's commit 'v1.11-598-g8493499' of 25-12-2011,
"silent-rules: fallback for makes without nested vars" (motivated
by automake bugs #9928 and #10237), Makefiles generated when the
'silent-rules' option is enabled can work also with those make
implementations that don't support nested variables expansion (albeit
the configure-time selected verbosity cannot be overridden at make
runtime in such case).

In light of that, and since silent rules are disabled by default even
when the 'silent-rules' option is given (unless the developer explicitly
makes them enabled by default be calling "AM_SILENT_RULES([yes])" in
configure.ac), we can now have the support for silent rules always
enabled.

The 'silent-rules' option will thus become a no-op, but will reaming a
valid option for the time being, for better backward-compatibility.

A collateral effect of this change is that the use of the 'silent-rules'
option will not anymore automatically disable the warnings in the
'portability-recursive' category.

The present change is basically a backport of the Automake-NG commit
'v1.11d-75-g61ca923' of 19-04-2012, "[ng] silent rules: support for
them is always active now".  Stemmed by this thread on the Automake-NG
list:
<http://lists.gnu.org/archive/html/automake-ng/2012-04/msg00027.html>

* automake.in: Enable silent-rules machinery unconditionally.  Reword
some comments.  Do not switch off warnings in the category
'portability-recursive' anymore if the 'silent-rules' option is
given.
* m4/init.m4: Call 'AC_REQUIRE' on 'AM_SILENT_RULES' unconditionally.
* lib/Automake/Options.pm: Recognize 'silent-rules' as a dummy option.
Do not bother anymore checking that it is set in 'configure.ac'.
* t/silent.sh: Remove now-redundant calls to AM_SILENT_RULES.
* t/silent2.sh: Likewise.
* t/silent3.sh: Likewise.
* t/silent4.sh: Likewise.
* t/silent8.sh: Likewise.
* t/silent9.sh: Likewise.
* t/silent-lex.sh: Likewise.
* t/silent-many-gcc.sh: Likewise.
* t/silent-many-generic.sh: Likewise.
* t/silent-yacc-headers.sh: Likewise.
* t/silent-yacc.sh: Likewise.
* t/silentcxx-gcc.sh: Likewise.
* t/silentcxx.sh: Likewise.
* t/silentf77.sh: Likewise.
* t/silentf90.sh: Likewise.
* t/help-silent.sh: Likewise.
* t/silent-configsite.sh: Likewise, and other minor adjustments.
* t/silent7.sh: Likewise.  Also, extend to check that silent rules
are disabled by default, and remove obsoleted checks about the
'silent-rules' option being rejected in 'AUTOMAKE_OPTIONS'.
* t/silent6.sh: Remove now-redundant calls to 'AM_SILENT_RULES'.
Call automake with '-Wno-portability-recursive'.  Remove other
obsoleted checks.
* t/dollarvar.sh: Don't expect the use of 'AM_SILENT_RULES' in
'configure.ac' to disable warning in the 'portability-recursive'
category anymore.
* t/compile_f_c_cxx.sh: Adjust to avoid a spurious failure.
* t/silent-amopts.sh: Remove as obsolete.
* t/silent-nowarn.sh: Likewise.
* t/list-of-tests.mk: Adjust.
* NEWS: Update.
* doc/automake.texi: Update and simplify accordingly.
* configure.ac (AM_INIT_AUTOMAKE): Remove 'silent-rules' option.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'maint'
Stefano Lattarini [Tue, 1 May 2012 12:28:39 +0000 (14:28 +0200)]
Merge branch 'maint'

* maint:
  fixup: botched edit in bootstrap.sh

12 years agofixup: botched edit in bootstrap.sh
Stefano Lattarini [Tue, 1 May 2012 12:25:51 +0000 (14:25 +0200)]
fixup: botched edit in bootstrap.sh

* bootstrap.sh (dosubst): Fix botched option ("G" instead of "g")
in a sed "s///" expression.  Issue introduced in today's commit
'v1.12-17-g5f810d0'.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'maint'
Stefano Lattarini [Tue, 1 May 2012 12:04:56 +0000 (14:04 +0200)]
Merge branch 'maint'

* maint:
  news: fix confusing wording
  news: fixed weaknesses in build system and testsuite
  bootstrap: consistently use correct copyright year
  fixup: remove leftover "TERM=ansi" definition in 't/color.sh'

12 years agonews: fix confusing wording
Stefano Lattarini [Tue, 1 May 2012 11:42:52 +0000 (13:42 +0200)]
news: fix confusing wording

Reported by Dave Hart.

* NEWS (Future backward-incompatibilities): Fix possibly confusing
wording about the upcoming requirement of Autoconf >= 2.65.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agonews: fixed weaknesses in build system and testsuite
Stefano Lattarini [Tue, 1 May 2012 11:32:42 +0000 (13:32 +0200)]
news: fixed weaknesses in build system and testsuite

* NEWS (Bugs fixed in 1.12.1): Report that several weaknesses in the
Automake's own build system and test suite have been fixed.  See for
example recent commits:
  - commit v1.12-5-geb7e8f3, for bug#11345
  - commit v1.12-5-g63e07a9, for bug#11346
  - commit v1.12-5-gf31fe4f
  - commit v1.12-10-gab14841
  - commit v1.12-19-g807f3cf, for bug#11369
  - commit v1.12-20-gd330368, for bug#11369
  - bug#11387, pre-emptively solved by commit v1.12-4-g444618b

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agobootstrap: consistently use correct copyright year
Stefano Lattarini [Tue, 1 May 2012 10:43:22 +0000 (12:43 +0200)]
bootstrap: consistently use correct copyright year

Suggested by Peter Johansson in the discussion about automake bug#11356:
<http://debbugs.gnu.org/cgi/bugreport.cgi?bug=11356#41>

* boostrap.sh (dosubst): Don't initialize the substituted value for the
release year from the current year (using `date +%Y`); instead, do it
through ...
($RELEASE_YEAR): ... this new statically-defined variable.  To ensure it
won't get out-of-sync, it will be automatically updated ...
* Makefile.am (update-copyright): ... by this target's recipe.
Since are at it, fix a botched output redirection for an error message,
i.e., use ">&2" instead of ">&1".

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agofixup: remove leftover "TERM=ansi" definition in 't/color.sh'
Stefano Lattarini [Mon, 30 Apr 2012 21:25:18 +0000 (23:25 +0200)]
fixup: remove leftover "TERM=ansi" definition in 't/color.sh'

It should have actually be removed in commit 'v1.12-23-gb105d40'.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoparallel-tests: save few forks when possible
Stefano Lattarini [Mon, 30 Apr 2012 20:18:24 +0000 (22:18 +0200)]
parallel-tests: save few forks when possible

* lib/am/check.am (am__check_pre): Save some forks, by analyzing more
carefully the value of '$@' to decide how to extract its dirname part,
and whether that should be created as a directory.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'maint'
Stefano Lattarini [Mon, 30 Apr 2012 20:07:15 +0000 (22:07 +0200)]
Merge branch 'maint'

* maint:
  news: Automake 1.13 will require Autoconf >= 2.65
  news: put planned future backward-incompatibilities on top
  readme: update INSTALL to latest version from Autoconf
  sync: some auxiliary files synced from upstream
  hacking: copyright notices must be updated before releasing
  color-tests: coloring can be forced on non-ANSI terminals as well

12 years agonews: Automake 1.13 will require Autoconf >= 2.65
Stefano Lattarini [Mon, 30 Apr 2012 19:43:21 +0000 (21:43 +0200)]
news: Automake 1.13 will require Autoconf >= 2.65

Autoconf 2.65 is almost three years old now, so requiring it is
acceptable.  Also, the soon-to-appear support for Objective C++
in Automake will be simplified if we can assume that Autoconf
version (which is the one that introduced support for Objective
C++).

* NEWS (Planned future backward-incompatibilities): Automake 1.13
will require Autoconf 2.65 or later.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agonews: put planned future backward-incompatibilities on top
Stefano Lattarini [Mon, 30 Apr 2012 19:29:34 +0000 (21:29 +0200)]
news: put planned future backward-incompatibilities on top

* NEWS (New in 1.12): List of planned future backward-incompatibilities
moved from here ...
(New in 1.12.1): ... to here.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoreadme: update INSTALL to latest version from Autoconf
Stefano Lattarini [Mon, 30 Apr 2012 19:25:37 +0000 (21:25 +0200)]
readme: update INSTALL to latest version from Autoconf

* INSTALL: Updated to the version coming with Autoconf 2.69.  This
is warranted now that we use and require that same Autoconf version
in our build system.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agosync: some auxiliary files synced from upstream
Stefano Lattarini [Mon, 30 Apr 2012 19:18:36 +0000 (21:18 +0200)]
sync: some auxiliary files synced from upstream

* lib/config.sub: Synced from upstream, by "make fetch".

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agohacking: copyright notices must be updated before releasing
Stefano Lattarini [Mon, 30 Apr 2012 19:17:06 +0000 (21:17 +0200)]
hacking: copyright notices must be updated before releasing

* HACKING (Release procedure): Suggest to use "make update-copyright".

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agocolor-tests: coloring can be forced on non-ANSI terminals as well
Stefano Lattarini [Mon, 30 Apr 2012 18:44:50 +0000 (20:44 +0200)]
color-tests: coloring can be forced on non-ANSI terminals as well

Before this change, colorization of testsuite output was suppressed
whenever the terminal was recognized to be a "dumb" one, incapable
of handling ANSI coloring (i.e., when the environment variable TERM
had a value of "dumb").  This happened even when the AM_COLOR_TESTS
variable was set to a value of "always".

Such a behaviour was suboptimal and slightly confusing; in fact, if
a user wants to force coloring of testsuite output that is being
redirected to a regular file, he should be able to do so even if his
terminal is not capable of handling ANSI colors -- in fact, such
terminal wouldn't be involved with the testsuite output in any way,
so why should it be allowed to influence it?

Thus, we now enable coloring of test output whenever AM_COLOR_TESTS
is set to "always", irrespective of the value of the TERM environment
variable.

* NEWS: Update.
* lib/am/check.am [%?COLOR%] (am__tty_colors): Activate colorization
of testsuite output whenever AM_COLOR_TESTS has the value of "always".
* t/ax/tap-summary-aux.sh: Export the TERM environment variable to
"dumb" when forcing colorization of the testsuite output; this should
*not* prevent such colorization from taking place, and we want to
check that this expectation really holds.
* t/ax/testsuite-summary-checks.sh: Likewise.
* t/color.sh: Likewise, and adjust some grepping checks.
* t/tap-color.sh: Likewise.  Also, remove redundant "make check"
invocation since we are at it.
* t/color2.sh: Likewise, and check that exporting TERM=dumb actually
prevents testsuite output colorization when AM_COLOR_TESTS is unset.
* t/parallel-tests-reset-term.sh: Relax, to prevent it from failing
spuriously due to the new semantic.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agocygnus: remove support for Cygnus-style trees
Stefano Lattarini [Fri, 6 Apr 2012 09:18:38 +0000 (11:18 +0200)]
cygnus: remove support for Cygnus-style trees

Support for "Cygnus-style" tree as enabled by the 'cygnus' option
has been deprecated in the documentation and with runtime warning
since Automake 1.12.1.  It's now time to remove it.

Closes automake bug#11034.

* lib/Automake/Options.pm (_process_option_list): Issue a proper
error when the 'cygnus' option is seen.
* automake.in: Don't handle the 'cygnus' option anymore.
* docs/automake.texi: Remove references to cygnus option and
Cygnus-style trees.
* lib/am/dejagnu.am, lib/am/texinfos.am: Drop special setups for
cygnus mode.
* NEWS: Update.
* t/cygnus-no-more.sh: New test.
* t/flavor.sh: Adjust.
* t/cygnus-deprecation.sh: Remove.
* t/cygnus-check-without-all.sh: Likewise.
* t/cygnus-requires-maintainer-mode.sh: Likewise.
* t/cygnus-dependency-tracking.sh: Likewise.
* t/cygnus-imply-foreign.sh: Likewise.
* t/cygnus-no-installinfo.sh: Likewise.
* t/cygnus-no-dist.sh: Likewise.
* t/clean2.sh: Likewise.
* t/txinfo5.sh: Likewise.
* t/txinfo5b.sh: Likewise.
* t/list-of-tests.mk: Update.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agocygnus: warn when it's used (flagged as 'obsolete' now)
Stefano Lattarini [Sat, 31 Mar 2012 12:53:24 +0000 (14:53 +0200)]
cygnus: warn when it's used (flagged as 'obsolete' now)

Support for "Cygnus-style" tree as enabled by the 'cygnus'
option is now reported as obsolete when the warnings in the
'obsolete' category are enabled.

See automake bug#11034.

* automake.in (check_cygnus): Report the use of 'cygnus' mode as
obsolete.
* NEWS: Update.
* docs/automake.texi: Minor adjustments.
* t/cygnus-deprecation.sh: New test.
* t/list-of-tests.mk: Add it.
* tests/cygnus-check-without-all.sh: Adjust by calling automake
with the warnings in the 'obsolete' category disabled.
* t/cygnus-requires-maintainer-mode.sh: Likewise.
* t/cygnus-dependency-tracking.sh: Likewise.
* t/cygnus-imply-foreign.sh: Likewise.
* t/cygnus-no-installinfo.sh: Likewise.
* t/cygnus-no-dist.sh: Likewise.
* t/clean2.sh: Likewise.
* t/flavor.sh: Likewise.
* t/txinfo5.sh: Likewise.
* t/txinfo5b.sh: Likewise.
* t/library.sh: Remove use of 'cygnus' and 'dejagnu' options,
and of AM_MAINTAINER_MODE macro.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoparallel-tests: optimize global log creation
Stefano Lattarini [Sat, 25 Feb 2012 23:30:18 +0000 (00:30 +0100)]
parallel-tests: optimize global log creation

With this change, the time required to execute the test case
'testsuite-summary-speed.sh' has dropped as follows:

  + Slow Debian i686 system, 1 core x 1.5 GHz, 768 MB of RAM,
    GNU make 3.81:
      - 1 run: 4 minutes => 30/40 seconds

  + Fast Solaris 10 i686 system, 4 cores x 3 GHz, 20 GB of RAM,
    Solaris CCS make:
     - 2 runs: 6 minutes => 30 seconds

  + Very fast Fedora ppc64 system, 64 cores x 3.5 GHz, 64 GB of RAM,
    GNU make 3.82:
     - 5 runs: 7 minutes => 1 minute and 10 seconds

* lib/am/check.am ($(TEST_SUITE_LOG)): Optimize for speed, by avoiding
lots of forks with the help of ...
(am__create_global_log): ... this new internal variable, basically
defining a smart awk program, and ...
(am__global_test_result_rx, am__copy_in_global_log_rx): ... these new
internal variables, used by the one above.
(am__rst_section): Remove as obsolete
* t/rst-formatting.trs: Remove.
* t/list-of-tests.mk: Update.
* t/testsuite-summary-count-many.sh: Improve its stress testing by
using 1 million tests rather than just 5 thousands: we can afford
this with our new optimization, and still have the test completed
in a reasonable time.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoparallel-tests: optimize 'recheck' target for speed
Stefano Lattarini [Thu, 26 Apr 2012 20:25:35 +0000 (22:25 +0200)]
parallel-tests: optimize 'recheck' target for speed

With this change, the time required to execute the test case
'testsuite-recheck-speed.sh' has dropped as follows:

  + Slow Debian i686 system, 1 core x 1.5 GHz, 768 MB of RAM,
    GNU make 3.81:
      - 1 run: 6 minutes => 40 seconds

  + Fast Solaris 10 i686 system, 4 cores x 3 GHz, 20 GB of RAM,
    Solaris CCS make:
      - 4 runs: 3 minutes => 1 minute

  + Very fast Fedora ppc64 system, 64 cores x 3.5 GHz, 64 GB of RAM,
    GNU make 3.82:
      - 5 runs: 4 minutes => 1 minute 30 seconds

* lib/am/check.am (recheck): Optimize for speed, by avoiding lots of
forks with the help of ...
(am__list_recheck_tests): ... this new internal variable, basically
defining a proper awk program, and ...
(am__recheck_rx): ... this other new internal variable, used by the
one above.
* t/test-trs-recover2.sh: Relax by not checking for a very corner
case ('.log' and '.trs' files both unreadable) that we don't handle
anymore.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agonews: announce initial support for Objective C++
Stefano Lattarini [Thu, 26 Apr 2012 13:14:04 +0000 (15:14 +0200)]
news: announce initial support for Objective C++

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoobjc++: test support for '.mm' suffix in _SOURCES entries
Peter Breitenlohner [Thu, 26 Apr 2012 13:12:39 +0000 (15:12 +0200)]
objc++: test support for '.mm' suffix in _SOURCES entries

* t/ext.sh: Enhance by adding a file with the '.mm' suffix in
the 'foo_SOURCES' definition.
* t/nodep2.sh: Likewise.

Co-authored-by: Stefano Lattarini <stefano.lattarini@gmail.com>
Signed-off-by: Peter Breitenlohner <peb@mppmu.mpg.de>
Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agocoverage: add performance tests on some parallel-tests aspects
Stefano Lattarini [Sun, 26 Feb 2012 08:42:11 +0000 (09:42 +0100)]
coverage: add performance tests on some parallel-tests aspects

* t/testsuite-recheck-speed.sh: New test.
* t/testsuite-summary-speed.sh: Likewise.
* t/list-of-tests.mk: Update.

The new tests are not meant to PASS/FAIL, but rather to help us to
obtain quantitative measurements of the performance improvements
offered by soon-to-appear optimization patches.

Timing before the optimizations implemented by following patches:

  + Slow Debian i686 system, 1 core x 1.5 GHz, 768 MB of RAM,
    GNU make 3.81:
     - testsuite-recheck-speed.sh: 1 run, ~ 6 minutes
     - testsuite-summary-speed.sh: 1 run, ~ 4 minutes

  + Fast Solaris 10 i686 system, 4 cores x 3 GHz, 20 GB of RAM,
    Solaris CCS make:
     - testsuite-recheck-speed.sh: 4 runs, ~ 3 minutes
     - testsuite-summary-speed.sh: 2 runs, ~ 6 minutes

  + Very fast Fedora ppc64 system, 64 cores x 3.5 GHz, 64 GB of RAM,
    GNU make 3.82:
     - testsuite-recheck-speed.sh: 5 runs, ~ 4 minutes
     - testsuite-summary-speed.sh: 5 runs, ~ 7 minutes

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoobjc++: add documentation
Peter Breitenlohner [Thu, 26 Apr 2012 13:03:02 +0000 (15:03 +0200)]
objc++: add documentation

* doc/automake.texi (Objective C++ Support): New node.
(How the Linker is Chosen, Support for Other Languages): Adjust.

Signed-off-by: Peter Breitenlohner <peb@mppmu.mpg.de>
Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoobjc++: initial support for Objective C++
Peter Breitenlohner [Thu, 26 Apr 2012 13:01:23 +0000 (15:01 +0200)]
objc++: initial support for Objective C++

Original thread (dating back to almost three years ago):
<http://lists.gnu.org/archive/html/automake-patches/2009-07/msg00016.html>

* automake.in: Register new language 'objcxx'.
(lang_objcxx_rewrite): New subroutine.
(resolve_linker): Add OBJCXXLINK.
(%_am_macro_for_cond): Add am__fastdepOBJCXX and AC_PROG_OBJCXX.
(%_ac_macro_for_var): Add OBJCXX and OBJCXXFLAGS.
* m4/depend.m4 (_AM_DEPENDENCIES): Add OBJCXX.
* m4/init.m4 (AM_INIT_AUTOMAKE): Add AC_PROG_OBJCXX hook.

Signed-off-by: Peter Breitenlohner <peb@mppmu.mpg.de>
Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'mkdirp-remove' into master
Stefano Lattarini [Sun, 29 Apr 2012 20:45:00 +0000 (22:45 +0200)]
Merge branch 'mkdirp-remove' into master

* mkdirp-remove:
  AM_PROG_MKDIR_P: remove as obsolete

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'maint'
Stefano Lattarini [Sun, 29 Apr 2012 20:43:27 +0000 (22:43 +0200)]
Merge branch 'maint'

* maint:
  AM_PROG_MKDIR_P: deprecate, to be removed in Automake 1.13

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'mkdirp-deprecate' into maint
Stefano Lattarini [Sun, 29 Apr 2012 20:35:16 +0000 (22:35 +0200)]
Merge branch 'mkdirp-deprecate' into maint

* mkdirp-deprecate:
  AM_PROG_MKDIR_P: deprecate, to be removed in Automake 1.13

12 years agoMerge branch 'maint'
Stefano Lattarini [Sun, 29 Apr 2012 20:32:20 +0000 (22:32 +0200)]
Merge branch 'maint'

* maint:
  tests: root can write on files without write perms; cater for this
  tests: require lex, not yacc, in a lex test case
  maint: fix a comment typo

12 years agotests: root can write on files without write perms; cater for this
Stefano Lattarini [Sat, 28 Apr 2012 21:18:45 +0000 (23:18 +0200)]
tests: root can write on files without write perms; cater for this

Fixes last part of automake bug#11369.

* t/dist-readonly.sh ($required): Add 'non-root'.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agotests: require lex, not yacc, in a lex test case
Stefano Lattarini [Sat, 28 Apr 2012 19:36:42 +0000 (21:36 +0200)]
tests: require lex, not yacc, in a lex test case

See automake bug#11369.

* t/lex-clean.sh ($required): Require 'lex', not 'yacc'.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoAM_PROG_MKDIR_P: remove as obsolete
Stefano Lattarini [Fri, 6 Apr 2012 09:24:11 +0000 (11:24 +0200)]
AM_PROG_MKDIR_P: remove as obsolete

Support for the obsolescent 'AM_PROG_MKDIR_P' m4 macro (and its
output variable '$(mkdir_p)') has in the documentation and with
runtime warning since Automake 1.12.1.  It's now time to remove it.

* automake.in (scan_autoconf_traces): Remove special handling of
'AM_PROG_MKDIR_P'.
* m4/mkdirp.m4: Delete.
* Makefile.am (dist_automake_ac_DATA): Remove it.
* t/mkdirp-deprecation.sh: Delete.
* t/list-of-tests.mk: Remove it.
* doc/automake.texi: Remove references to the 'AM_MKDIR_P' m4 macro
and the '$(mkdir_p)' output variable.
* t/gettext-macros.sh: Ensure the gettext-requiring tests will
still see the now-removed 'AM_PROG_MKDIR_P' macro as an alias for
the 'AC_PROG_MKDIR_P' macro, since even recent versions of gettext
uses 'AM_PROG_MKDIR_P' in their '.m4' files.
* NEWS: Update.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoAM_PROG_MKDIR_P: deprecate, to be removed in Automake 1.13
Stefano Lattarini [Mon, 2 Apr 2012 15:05:42 +0000 (17:05 +0200)]
AM_PROG_MKDIR_P: deprecate, to be removed in Automake 1.13

Support for the obsolescent 'AM_PROG_MKDIR_P' macro (and its
output variable '$(mkdir_p)') has already been deprecated in
the documentation.  It's now time to make its use actively
emit warnings in the 'obsolete' category as well.

* automake.in (scan_autoconf_traces): Warn about 'AM_PROG_MKDIR_P'
obsolescent if that macro is seen.
* m4/mkdirp.m4 (AM_PROG_MKDIR_P): Warn about its own obsolescence.
* m4/init.m4: Require 'AC_PROG_MKDIR_P', not 'AM_PROG_MKDIR_P'.
* t/mkdirp-deprecation.sh: New test.
* t/list-of-tests.mk: Add it.
* syntax-check.mk (sc_mkdir_p): Delete this check as obsolete.
(syntax_check_rules): Remove it.
* t/gettext-macros.sh: Ensure the gettext-requiring tests will
call aclocal and automake with the 'obsolete' warnings disabled,
since even recent versions of gettext used the now-deprecated
'AM_PROG_MKDIR_P' m4 macro.
* NEWS: Update.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agomaint: fix a comment typo
Jim Meyering [Sat, 28 Apr 2012 08:40:26 +0000 (10:40 +0200)]
maint: fix a comment typo

* configure.ac: Fix typo in comment: s/out/our/

12 years agoMerge branches 'remove-acdir-option' and 'remove-obsolete-m4'
Stefano Lattarini [Sat, 28 Apr 2012 08:53:20 +0000 (10:53 +0200)]
Merge branches 'remove-acdir-option' and 'remove-obsolete-m4'

* remove-acdir-option:
  aclocal: remove the --acdir option

* remove-obsolete-m4:
  m4: delete several obsolete macros

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'maint'
Stefano Lattarini [Sat, 28 Apr 2012 08:47:22 +0000 (10:47 +0200)]
Merge branch 'maint'

* maint:
  fixup: avoid unconditional re-bootstrapping on "make dist"
  build: ensure release year in copyright notice is up-to-date
  cosmetics: fix spacing in THANKS
  readme: explicitly state we use ranges in copyright years
  bootstrap: add convenience make target
  automake, aclocal: update copyright year in output

12 years agofixup: avoid unconditional re-bootstrapping on "make dist"
Stefano Lattarini [Sat, 28 Apr 2012 08:21:31 +0000 (10:21 +0200)]
fixup: avoid unconditional re-bootstrapping on "make dist"

Since our 'bootstrap' script is distributed, the 'distdir' target depends
on it.  But in our GNUmakefile, we also have a 'bootstrap' target declared
.PHONY, which when called re-bootstrap the Automake distribution.  Thus,
whenever we run "make dist", GNU make sees it must remake the 'bootstrap'
target (as it is a dependency of the 'distdir' target), and thus ends up
re-bootstrapping all the package (because of the .PHONY 'bootstrap' target
in GNUmakefile).

We fix this issue by renaming our bootstrap script to 'bootstrap.sh'.

* bootstrap: Renamed ...
* bootstrap.sh: ... to this.
* GNUmakefile (bootstrap): Adjust.
* Makefile.am (EXTRA_DIST): Likewise.
(autodiffs): Likewise.
* HACKING: Likewise.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agobuild: ensure release year in copyright notice is up-to-date
Stefano Lattarini [Fri, 27 Apr 2012 14:27:08 +0000 (16:27 +0200)]
build: ensure release year in copyright notice is up-to-date

From a suggestion by Eric Blake.  See automake bug#11356.
This is a follow-up to previous patch 'v1.12-12-gb99b5be'.

* configure.ac (RELEASE_YEAR): New AC_SUBST'd variable, should hold
the value of the current year.
* Makefile.am (update-copyright): Be sure to also update the
definition of 'RELEASE_YEAR' in configure.ac.
* lib/Automake/Config.in ($RELEASE_YEAR): New exported variable,
initialized from the value substituted for '@RELEASE_YEAR@'.
(@EXPORT): Add it.
* automake.in, aclocal.in: Use '$RELEASE_YEAR' (which will be
substituted at make time) instead of hard-coding the release
year.  This should ensure the copyright range in the version
message and in the generated files (Makefile.in and aclocal.m4)
are automatically kept up-to-date.
* bootstrap (dosubst): Update, also substitute '@RELEASE_YEAR@'.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agocosmetics: fix spacing in THANKS
Stefano Lattarini [Fri, 27 Apr 2012 21:56:17 +0000 (23:56 +0200)]
cosmetics: fix spacing in THANKS

* THANKS: Use spaces, not tabs, to separate a reporter's name from
his e-mail address.  Also, increase such spacing, to allow a more
consistent formatting with longer names.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoreadme: explicitly state we use ranges in copyright years
Stefano Lattarini [Sat, 28 Apr 2012 07:29:34 +0000 (09:29 +0200)]
readme: explicitly state we use ranges in copyright years

Reported by Jack Kelly in automake bug#11356.

* README: Explicitly state we use ranges in copyright years; this is
mandated by the GNU Coding Standards for each package that wants to
actually use ranges in copyright years.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoaclocal: remove the --acdir option
Stefano Lattarini [Fri, 6 Apr 2012 08:46:27 +0000 (10:46 +0200)]
aclocal: remove the --acdir option

It not documented in the manual anymore, its use has already been
causing a deprecation warning, and its oncoming removal has already
been announced in the NEWS file.  So let's finally remove it.

* aclocal.in (parse_arguments): Don't recognize the '--acdir' option
anymore.
(handle_acdir_option): Remove subroutine.
* configure.ac ($ACLOCAL): Adjust.
($AUTOMAKE): Cosmetic changes, for consistency.
* tests/aclocal-acdir.test: Remove usages of the '--acdir' option.
* tests/aclocal-print-acdir.test: Likewise.
* NEWS: Update.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agobootstrap: add convenience make target
Stefano Lattarini [Fri, 27 Apr 2012 16:48:27 +0000 (18:48 +0200)]
bootstrap: add convenience make target

Dependencies in the Automake build system are not completely specified
(see for example the commit log of recent commit 'v1.12-10-gab14841',
"build: avoid too greedy rebuilds in the testsuite").  In fact, some of
them cannot even be; for example, Makefile is generated at configure
time from Makefile.in, which should be regenerated by our bleeding-edge
automake script, which is generated by out Makefile -- specifying the
complete chain of dependencies here would bring to a circular dependency
issue.

For this reason, before testing or deploying a change, we are often
forced to perform a full re-bootstrap of the Automake package, to ensure
all our files are actually up-to-date.  Until now, this has to be done
manually, thus causing wasted keystrokes and more possibilities of error.

With this change, we introduce a new 'bootstrap' make target to
automatize all the (easy) steps of this re-bootstrapping (plus some
minor bells & whistles since we are at it).

* GNUmakefile: Rewrite to allow an easy bootstrapping and clean rebuild
of the whole package, in particular with the help of ...
(bootstrap): ...  this new target.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agom4: delete several obsolete macros
Stefano Lattarini [Fri, 6 Apr 2012 08:10:12 +0000 (10:10 +0200)]
m4: delete several obsolete macros

Support for several obsolescent macros has already been deprecated,
and their removal already announced in the documentation and in the
NEWS file.  It's now time to finally remove them.

* m4/dmalloc.m4 (fp_WITH_DMALLOC): Don't define anymore.
* m4/maintainer.m4 (jm_MAINTAINER_MODE): Likewise.
* m4/lispdir.m4 (ud_PATH_LISPDIR): Likewise.
* m4/protos.m4 (fp_C_PROTOTYPES): Likewise.
* m4/ccstdc.m4: Delete file.
* m4/header.m4: Likewise.
* m4/obsol-gt.m4: Likewise.
* m4/obsol-lt.m4: Likewise.
* m4/obsolete.m4: Likewise.
* Makefile.am (dist_automake_ac_DATA): Don't list the deleted '.m4'
files anymore.
* t/confh.sh: Use 'AC_CONFIG_HEADERS' instead of the now-removed
AM_CONFIG_HEADER.
* t/confh.sh: Likewise.
* t/confh4.sh: Likewise.
* t/confh6.sh: Likewise.
* t/confh7.sh: Likewise.
* t/confh8.sh: Likewise.
* t/confsub.sh: Likewise.
* t/stamph2.sh: Likewise.
* t/subdir6.sh: Likewise.
* t/autohdr2.sh: Remove as obsolete.
* t/obsolete.sh: Likewise.
* t/list-of-tests.mk. Don't list the deleted test cases anymore.
* NEWS: Update.
* doc/automake.texi: Likewise.  In particular ...
(Obsolete Macros): Remove this node altogether.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoautomake, aclocal: update copyright year in output
Stefano Lattarini [Fri, 27 Apr 2012 09:14:53 +0000 (11:14 +0200)]
automake, aclocal: update copyright year in output

Fixes automake bug#11356.

* aclocal.in, automake.in: Add copyright year '2012' in the version
message.  Ditto for the copyright notice in the generated Makefile.in
or aclocal.m4 files; and for those, since we are at it, prefer using
a range for copyright years, rather than listing each copyright year
individually.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'maint'
Stefano Lattarini [Thu, 26 Apr 2012 15:52:31 +0000 (17:52 +0200)]
Merge branch 'maint'

* maint:
  maint: don't use "union" merge driver for NEWS anymore
  build: avoid too greedy rebuilds in the testsuite
  tests: avoid yet another spurious environment influence
  tests: fix a failure in Lex/C++ tests on Mac OS X
  tests: fix a spurious failure due to truncated timestamps
  build: require autoconf 2.69
  build: don't require minimal automake version in AM_INIT_AUTOMAKE

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agomaint: don't use "union" merge driver for NEWS anymore
Stefano Lattarini [Thu, 26 Apr 2012 15:47:14 +0000 (17:47 +0200)]
maint: don't use "union" merge driver for NEWS anymore

The use of the "union" merge driver to avoid conflicts in NEWS has
proven to be problematic, as it often silently produced mis-merges
(instead of spurious conflicts, but a spurious conflict is better
than a mis-merge).  Also, our new branching policy (see automake
bug#11153 for a description) should greatly mitigate the problem of
merge conflicts in NEWS.  For further discussion, see:
<http://lists.gnu.org/archive/html/automake-patches/2012-04/msg00148.html>

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agobuild: avoid too greedy rebuilds in the testsuite
Stefano Lattarini [Thu, 26 Apr 2012 13:05:05 +0000 (15:05 +0200)]
build: avoid too greedy rebuilds in the testsuite

The autogenerated makefile fragment 't/testsuite-part.am' used to depend
on the list of *all* the hand-written files.  This was technically correct,
since that makefile fragment was (and still is) generated by the script
'gen-testsuite-part', which scans all the hand-written files to detect
implicit dependencies and to decide for which tests a further wrapper test
should be generated.

However, the presence of such a dependency implied that, whenever *any*
test case was modified (no matter how slightly), the 't/testsuite-part.am'
file was rebuilt, and since that is included by our 'Makefile.am', the
'Makefile.in' file was rebuilt as well by automake.  In order to do so,
automake scanned our 'configure.ac' file, which requires the latest
Autoconf version (2.69 at the time of writing), and the casual user can
easily lack that on his machine (and that should be allowed, as automake
currently supports any autoconf version >= 2.62).

The described situation could hinder in-field testing or debugging by
users (even experienced ones) that are not automake developers; see for
example automake bug#11347.

So we drop the explicit dependency of 't/testsuite-part.am' on the
hand-written test cases.  (As an aside, note that this has already
been done in Automake-NG, albeit for other reasons; see the commit
'v1.11b-129-g1690aca' of 2012-04-23, "[ng] build: define $(TESTS)
through a wildcard").

This is not a serious regression in the faithfulness of the Automake
build systems, since it unfortunately already had several undeclared
dependencies; e.g., 'Makefile.in' should depend on the automake script,
and 'aclocal.m4' should depend on the aclocal script (but this isn't
possible, as it would mean that distributed files depend on ones generated
at make time).  Similarly, 'Makefile.in' should depend on some of the
'lib/am/*.am' files, but does not.

Currently, the workaround to get a faithful and correct rebuild is to run

  ./bootstrap && ./config.status --recheck && make clean all

and the present commit does not change that.

* Makefile.am ($(srcdir)/t/testsuite-part.am): Don't depend on
't/list-of-tests.mk' nor on '$(handwritten_TESTS)' anymore.
($(generated_TESTS)): Likewise, and don't depend on 'Makefile.am'
either.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>