platform/upstream/automake.git
12 years agoMerge branch 'testsuite-work' into master
Stefano Lattarini [Mon, 26 Dec 2011 21:49:22 +0000 (22:49 +0100)]
Merge branch 'testsuite-work' into master

At last!

Keeping the `master' and `testsuite-work' separated was becoming
quite artificial, since there is no way we are going to release
a 1.12 version of Automake that doesn't include the support for
the TAP test protocol and for user-defined test drivers -- which
are all implemented into the testsuite-work branch.

Moreover, keeping those two branches separated was causing more
maintenance and double-checking work than were warranted, was
making the automake git repository even more contorted, and, on
the top of all that, was actually *reducing coverage* by making
all the important changes in testsuite-work less visible (less
on-field testing by developers that routinely use bleeding edge
versions, less automated testing by services like Hydra/NixOS,
et cetera).

12 years agoMerge branch 'master' into testsuite-work
Stefano Lattarini [Sat, 24 Dec 2011 10:05:24 +0000 (11:05 +0100)]
Merge branch 'master' into testsuite-work

* master:
  gitignore: use only one .gitignore file, in the top-level directory
  tests: one more fixlet for gettext macros requirement
  tests: fix handling of gettext macros requirement

12 years agoMerge branch 'maint'
Stefano Lattarini [Sat, 24 Dec 2011 09:53:41 +0000 (10:53 +0100)]
Merge branch 'maint'

* maint:
  gitignore: use only one .gitignore file, in the top-level directory
  tests: one more fixlet for gettext macros requirement
  tests: fix handling of gettext macros requirement
  tests: fix failure due to debugging code forgotten into a test

12 years agogitignore: use only one .gitignore file, in the top-level directory
Stefano Lattarini [Sat, 24 Dec 2011 09:43:26 +0000 (10:43 +0100)]
gitignore: use only one .gitignore file, in the top-level directory

* doc/.gitignore, doc/amhello/.gitignore, lib/Automake/.gitignore,
lib/Automake/tests/.gitignore, tests/.gitignore: Deleted, their
contents merged into ...
* .gitignore: ... this top-level file.  Since we are at it, improve
it by anchoring files where it makes sense.

12 years agoMerge branch 'tests-gettext-macros-require' into maint
Stefano Lattarini [Sat, 24 Dec 2011 08:58:17 +0000 (09:58 +0100)]
Merge branch 'tests-gettext-macros-require' into maint

* tests-gettext-macros-require:
  tests: one more fixlet for gettext macros requirement

12 years agotests: one more fixlet for gettext macros requirement
Stefano Lattarini [Sat, 24 Dec 2011 08:56:49 +0000 (09:56 +0100)]
tests: one more fixlet for gettext macros requirement

* tests/gettext-macros.test (configure.in): Add calls to
AC_INIT and AC_PROG_CC, to avoid a spurious failure in the
later aclocal invocation.
Luckily, this wart wasn't causing any spurious SKIP or FAIL,
but only the redundant addition to the `-Wno-syntax' option
to some aclocal invocations in the gettext tests.

12 years agoMerge branch 'tests-gettext-macros-require' into maint
Stefano Lattarini [Sat, 24 Dec 2011 08:29:15 +0000 (09:29 +0100)]
Merge branch 'tests-gettext-macros-require' into maint

* tests-gettext-macros-require:
  tests: fix handling of gettext macros requirement

12 years agotests: fix handling of gettext macros requirement
Stefano Lattarini [Sat, 24 Dec 2011 08:25:00 +0000 (09:25 +0100)]
tests: fix handling of gettext macros requirement

The code introduced in the earlier change `v1.11-581-gb7d67d5'
and aimed at automatically fetching all the .m4 files provided by
gettext has proven inadequate, since it hasn't managed to truly
and always get *all* the required .m4 files.  For example, it has
failed to fetch the file `intldir.m4' (present in gettext 0.18.1,
and where the macro `AM_GNU_GETTEXT_INTL_SUBDIR' is defined),
which in turn has caused spurious SKIPs of at least the test
`gettext3.test'.

The reason for this debacle is that autopoint looks at the version
specified in AM_GNU_GETTEXT_VERSION to decide which version of the
gettext infrastructure to bring in; since we were unconditionally
specifying the older 0.10.35 version, the newer `.m4' files weren't
brought in.

* tests/gettext-macros.test: In AM_GNU_GETTEXT_VERSION, instead
of unconditionally specifying the older 0.10.35 version, specify
the version of the available `gettextize' or `autopoint' program,
dynamically extracted with the help of ...
(extract_program_version):... this new function, and saved ...
($autopoint_version): ... in this new variable.

12 years agotests: automatically handle deps also for generated tests
Stefano Lattarini [Fri, 23 Dec 2011 21:14:50 +0000 (22:14 +0100)]
tests: automatically handle deps also for generated tests

* tests/gen-testsuite-part: The test scripts are now scanned for
automatic dependency generation *after* the auto-generated tests
have been created, so they too will be scanned.  It is a little
tricky to ensure that the freshly-generated tests are correctly
scanned, and we do that with the help of ...
(@generated_tests): ... this new variable.
Other related minor changes and refactorings.

12 years agotests: auto-generate deps for tests requiring libtool/gettext
Stefano Lattarini [Fri, 23 Dec 2011 20:21:08 +0000 (21:21 +0100)]
tests: auto-generate deps for tests requiring libtool/gettext

* tests/gen-testsuite-part: Tests requiring libtool (or related
programs) will need libtool-provided m4 macros, so they should
be run after `libtool-macros.test'.  Similarly for gettext tests.
So, generate proper declarations of such dependencies.
* tests/Makefile.am: Remove now-unneeded hand-written declaration
of those dependencies.

12 years agotests: remove FIXME comments from tricks to pacify syntax checks
Stefano Lattarini [Fri, 23 Dec 2011 20:01:59 +0000 (21:01 +0100)]
tests: remove FIXME comments from tricks to pacify syntax checks

Originally, I myself had insisted on adding those "FIXME" comments
near every unusual construct or use of "creative quoting" whose
only purpose was to avoid false positive from out syntax checks.
I did that with the expectation that we could have easily and soon
added a whitelisting capability to our maintainer-specific syntax
checks.  This hasn't been the case unfortunately, and now those
"FIXME" in the comments are just distracting noise, making it more
difficult to grep the test cases for stuff that could really use a
fix or an improvement.  So just get rid of them.

* tests/instdir-ltlib.test: Remove that extra "FIXME".
* tests/instdir-prog.test: Likewise.
* tests/parallel-tests3.test: Likewise.
* tests/parallel-tests-dry-run.test: Likewise.
* tests/parallel-tests-console-output.test: Likewise.
* tests/parallel-tests-no-color-in-log.test: Likewise.
* tests/tap-realtime.test: Likewise.
* tests/test-trs-recover.test: Likewise.
* tests/tap-deps.test: Likewise.

12 years agotests: few minor fixlets, improvements and tweakings
Stefano Lattarini [Fri, 23 Dec 2011 19:39:46 +0000 (20:39 +0100)]
tests: few minor fixlets, improvements and tweakings

* tests/defs-static.in ($top_testbuilddir): New, for consistency
and completeness.
* tests/Makefile.am (do_subst): Substitute @abs_top_testbuilddir@
as well.
* tests/java-compile-run-mested.test: Prefer AM_TESTS_ENVIRONMENT
over TESTS_ENVIRONMENT, now that the former has become available.
* tests/java-compile-run-flat.test: Correct a botched "FIXME"
comment.
* tests/tap-realtime.test: Remove extra whitespace in comments.
* tests/missing-tar.test: Use `get_shell_script' to bring in the
`missing' script, to increase coverage.  Remove redundant call to
`set -e'.  Fix a typo in comments.
* tests/get-sysconf.test: Remove redundant definitions of
`$top_testsrcdir', `testbuilddir' and `$top_testbuilddir'.

12 years agotests: remove redundant uses of `set -e'
Stefano Lattarini [Fri, 23 Dec 2011 17:02:21 +0000 (18:02 +0100)]
tests: remove redundant uses of `set -e'

* tests/deleted-am.test: Do not set the `errexit' shell flag, as
it is already set by `tests/defs'.
* tests/deleted-m4.test: Likewise.
* tests/dist-missing-am.test: Likewise.
* tests/dist-missing-am.test: Likewise.
* tests/dist-missing-m4.test: Likewise.
* tests/dist-missing-included-m4.test: Likewise.
* tests/get-sysconf.test: Likewise.
* tests/makefile-deps.test: Likewise.
* tests/remake-m4-pr10111.test: Likewise.
* tests/remake-am-pr10111.test: Likewise.
* tests/remake-deleted-am.test: Likewise.
* tests/remake-deleted-am2.test: Likewise.
* tests/remake-deleted-am-subdir.test: Likewise.
* tests/remake-renamed-am.test: Likewise.

12 years agoMerge branch 'maint' into testsuite-work
Stefano Lattarini [Fri, 23 Dec 2011 12:18:19 +0000 (13:18 +0100)]
Merge branch 'maint' into testsuite-work

* maint:
  tests: fix failure due to debugging code forgotten into a test

12 years agoMerge branch 'missing-fixes' into maint
Stefano Lattarini [Fri, 23 Dec 2011 12:09:50 +0000 (13:09 +0100)]
Merge branch 'missing-fixes' into maint

* missing-fixes:
  tests: fix failure due to debugging code forgotten into a test

12 years agotests: fix failure due to debugging code forgotten into a test
Stefano Lattarini [Fri, 23 Dec 2011 11:47:19 +0000 (12:47 +0100)]
tests: fix failure due to debugging code forgotten into a test

* tests/missing-tar.test: Don't ever call the `missing' script
with `sh -x'; this was used for debugging, but an instance of
it slipped into the committed test case.  Bug revealed by a
failure on a Solaris 10 system with GNU tar installed as `gtar'.

12 years agoMerge branch 'master' into testsuite-work
Stefano Lattarini [Fri, 23 Dec 2011 09:15:29 +0000 (10:15 +0100)]
Merge branch 'master' into testsuite-work

* tests/gettext-macros.test: In the generated `get.sh' file,
use `skip_all_', not `skip_'.
* tests/libtool-macros.test: Likewise.
* tests/list-of-tests.mk: Update.

12 years agoMerge branch 'maint'
Stefano Lattarini [Thu, 22 Dec 2011 23:44:34 +0000 (00:44 +0100)]
Merge branch 'maint'

* maint:
  tests: avoid spurious failure of libtool and gettext tests

12 years agoMerge branch 'tests-libtool-gettext-requirements' into maint
Stefano Lattarini [Thu, 22 Dec 2011 23:42:46 +0000 (00:42 +0100)]
Merge branch 'tests-libtool-gettext-requirements' into maint

* tests-libtool-gettext-requirements:
  tests: avoid spurious failure of libtool and gettext tests

12 years agotests: avoid spurious failure of libtool and gettext tests
Stefano Lattarini [Thu, 22 Dec 2011 23:42:24 +0000 (00:42 +0100)]
tests: avoid spurious failure of libtool and gettext tests

On Solaris 10 (and presumably earlier), /bin/sh trips up on
here-documents that contains a command substitution *and* are
fed to a shell function:

  # All as expected.
  $ cat <<END
  `pwd`
  END
  /home/stefano
  $ echo status = $?
  status = 0

  # An apparently innocuous function ...
  $ kitty () { cat; }
  # ... but hilarity ensues!
  $ kitty <<END
  `pwd`
  END
  /tmp/sh137723: cannot open
  $ echo status = $?
  status = 1

We need to work around this misbehaviour in a couple of our
tests (whose failures where causing cascading failures in a
lot of other tests).

* tests/gettext-macros.test: Avoid the use of command substitution
in a here-document passed to the `indent' function, by using the
`echo' builtin instead.
* tests/libtool-macros.test: Likewise.

See also:
<http://lists.gnu.org/archive/html/bug-autoconf/2011-12/msg00001.html>

12 years agoMerge branch 'maint' into master.
Stefano Lattarini [Thu, 22 Dec 2011 21:16:12 +0000 (22:16 +0100)]
Merge branch 'maint' into master.

* m4/regex.m4: Bump serial number by some notches, for good
measure.  Prefer the deprecation message from maint over that
from master.
* m4/python.m4: Bump serial number.
* tests/Makefile.am (canon6.log): Depend on `libtool-macros.log'.
(canon7.log): Likewise.
(extradep2.log): Likewise.
(ar-lib4.log): Likewise.
(ar-lib6a.log): Likewise.
(ar-lib6b.log): Likewise.
(vartypo2): Likewise.
(posixsubst-ltlibraries.log): Likewise.
* tests/defs: Adjust the path of the included `get.sh' scripts
generated by `libtool-macros.test' and `gettext-macros.test'.
* aclocal.m4: Delete, it's auto-generated now.
* configure: Likewise.
* Makefile.in: Likewise.
* doc/Makefile.in: Likewise.
* m4/Makefile.in: Likewise.
* tests/Makefile.in: Likewise.
* lib/Makefile.in: Likewise.
* lib/am/Makefile.in: Likewise.
* lib/Automake/Makefile.in: Likewise.
* lib/Automake/tests/Makefile.in: Likewise.
* .gitignore: Correctly ignored the autogenerated files
that are not committed anymore.

12 years agohacking: distribute it, and mention it in the ChangeLog
Stefano Lattarini [Thu, 22 Dec 2011 20:47:41 +0000 (21:47 +0100)]
hacking: distribute it, and mention it in the ChangeLog

Not distributing the HACKING file might make it more difficult,
for some random curious user, to get informed about or interested
in the Automake development process, or to send us patches.

* Makefile.am (EXTRA_DIST): Add HACKING.
* HACKING: It's OK to distribute this file, and to mention it in
the ChangeLog.

12 years agoregex: deprecate the obsolete macro AM_WITH_REGEX
Stefano Lattarini [Sat, 13 Nov 2010 11:21:28 +0000 (12:21 +0100)]
regex: deprecate the obsolete macro AM_WITH_REGEX

This is a backport of commit v1.11-433-g37b0aee.

Today, practically nobody uses the GNU rx library, which, according
to its own website <http://www.gnu.org/software/rx/rx.html>, has
been "decommissioned".  Consequently, the automake-provided macro
AM_WITH_REGEX is not used nor required anymore.  Deprecate it, so
that it will be possible to safely remove it in the next major
automake version.

* m4/regex.m4 (AM_WITH_REGEX): Give a warning of the class
`obsolete' when this macro is used.
* doc/automake.texi (Public Macros): Move description of
`AM_WITH_REGEX' from here ...
(Obsolete Macros): ... to here, and declare it as obsolete
and "to be removed in a future version".
* tests/regex-obsolete.test: New test.
* tests/Makefile.am (TESTS): Add it.
* NEWS: Update.

See also:
<http://lists.gnu.org/archive/html/automake-patches/2010-11/msg00131.html>

12 years agomaint: distribute .xz tarballs, not .bz2 ones
Stefano Lattarini [Thu, 22 Dec 2011 18:51:52 +0000 (19:51 +0100)]
maint: distribute .xz tarballs, not .bz2 ones

Many GNU packages are moving towards xz-compressed tarballs, so
let's follow suit, by dropping the creation and distribution of
a bzip2-compressed tarball and switching to xz instead.
For compatibility and safeness, we will continue to create and
distribute a gzip-compressed tarball as well.

* configure.ac (AM_INIT_AUTOMAKE): Drop `dist-bzip2', add
`dist-xz'.
* NEWS: Update.

Suggested by Jim Meyering.

12 years agoMerge branch 'pr9768-fix' into maint
Stefano Lattarini [Thu, 22 Dec 2011 18:20:46 +0000 (19:20 +0100)]
Merge branch 'pr9768-fix' into maint

* pr9768-fix:
  include: avoid "deleted .am file" problem

12 years agomaint: better use of autoconf 2.68 features
Stefano Lattarini [Thu, 22 Dec 2011 18:16:48 +0000 (19:16 +0100)]
maint: better use of autoconf 2.68 features

* configure.ac: Now that Automake requires autoconf 2.68 for its
own bootstrapping and build system, we can assume that PACKAGE_URL
gets automatically AC_SUBT'd.

12 years agocosmetics: use proper m4 quoting in configure.ac
Stefano Lattarini [Thu, 22 Dec 2011 18:05:32 +0000 (19:05 +0100)]
cosmetics: use proper m4 quoting in configure.ac

* configure.ac (AC_CONFIG_SRCDIR): Use proper m4 quoting
for its arguments.
(AC_CONFIG_AUX_DIR): Likewise.
(AC_PROG_PATH): Likewise.

12 years agoinclude: avoid "deleted .am file" problem
Stefano Lattarini [Sun, 16 Oct 2011 12:57:37 +0000 (14:57 +0200)]
include: avoid "deleted .am file" problem

* automake.in (handle_configure): When processing `configure.am',
also expand `HAVE-MAKEFILE-IN-DEPS' to a boolean telling whether
`MAKEFILE-IN-DEPS' is empty or not.
* lib/am/configure.am [?HAVE-MAKEFILE-IN-DEPS?]
(%MAKEFILE-IN-DEPS%): New target without dependencies, to
avoid the "deleted .am file" problem.  Emit this only when
`?HAVE-MAKEFILE-IN-DEPS?' is true, to avoid generating an
"empty" dependency declaration.
* tests/deleted-am.test: Make grepping of error message stricter.
* tests/dist-missing-am.test: Likewise.
* tests/remake-deleted-am.test: New test.
* tests/remake-deleted-am-2.test: Likewise.
* tests/remake-deleted-am-subdir.test: Likewise.
* tests/remake-renamed-am.test: Likewise.
* tests/makefile-deps.test: Likewise.
* tests/Makefile.am (TESTS): Add the new tests.
* NEWS: Update.

Fixes automake bug#9768.

Report by Peter Johansson.

See also commit `Release-1-10-40-gd0ebf71', which fixed a similar
problem for .m4 files included by configure.ac.

12 years agotests: better handling of gettext and libtool requirements
Stefano Lattarini [Wed, 14 Dec 2011 09:35:04 +0000 (10:35 +0100)]
tests: better handling of gettext and libtool requirements

This change fixes automake bug#9807.

Before this change, the automake testsuite only looked for the
`.m4' files containing libtool and gettext macros definitions in
the directory `${prefix}/share/aclocal' (and in the directories
specified by the `dirlist' file in there, if any), where ${prefix}
was the configure-time automake installation prefix (defaulting
to `/usr/local').

This approach had various shortcomings and disadvantages.  Let's
briefly describe the three major ones.

First, on most GNU/Linux systems, a libtool or gettext installed
from distro-provided packages (e.g., by dpkg on Debian/Ubuntu, or
by rmp on RedHat/Fedora) would have `/usr', not `/usr/local', as
its ${prefix}; so, trying to run the automake testsuite with a
simple "./configure && make && make check" would have failed to
execute the libtool and gettext tests on most GNU/Linux distros.
It's true that it was quite easy to work around this issue, by
creating a proper `/usr/local/share/aclocal/dirlist' file with
an entry pointing to `/usr/share/aclocal' (a workaround in fact
used by most automake developers); but the typical user wasn't
aware of the necessity of this trick, so the libtool and gettext
tests was usually skipped on testsuite runs "in the wild", thus
needlessly reducing coverage.

Second, the older testsuite behaviour made more difficult for
the developers to run the testsuite with non-default libtool or
gettext.  For example, assume the developer is working on a system
that has a default libtool version 1.5 installed in the /usr/local
hierarchy; to improve coverage, the developer installs also a more
modern libtool version, say 2.4, in its home directory, let's say
in ~/libtool-2.4; he then tries to run the automake testsuite with
this more modern libtool by doing an (apparently) simple:
  $ PATH=$HOME/libtool-2.4:$PATH make check
But the automake testsuite would still look for libtool macros in
/usr/local/share/aclocal, not in ~/libtool-2.4/share/aclocal, so
the wrong version of the macros would be picked up, and the tests
would either fail spuriously or (which would be worse) pass without
truly covering the libtool version the developers was thinking to
be testing with.
Worse again, the automake testsuite would *unconditionally* look
for libtool macros in /usr/local/share/aclocal, so even something
like:
  $ export ACLOCAL_PATH=$HOME/libtool-2.4/share/aclocal
  $ PATH=$HOME/libtool-2.4:$PATH make check
wouldn't work.

Third and last, during a "make distcheck", automake is configured
with a ${prefix} pointing to a proper subdirectory of the build
directory (usually `pwd`/_inst), which gets created on-the-fly;
in this case, with the old approach, the automake testsuite never
found the libtool and gettext macro files, ans so the libtool and
gettext tests was *always* skipped in a "make distcheck".

* tests/libtool-macros.test: New helper test, looking (with the
help of the `libtoolize' script) for libtool macro files required
by most libtool tests, and making them easily accessible.
* tests/gettext-macros.test: New helper test, looking (with the
help of the `libtoolize' script) for libtool macro files required
by most libtool tests, and making them easily accessible.
* tests/defs.in: Update to make it rely on the results and setups
of `libtool-macros.test' and `gettext-macros.test'.
* tests/Makefile.am: Declare dependency of all the logs of libtool
tests from `libtool-macros.log', and all the logs of gettext tests
from `gettext-macros.log'.
(TESTS): Add the new tests.

12 years agofix: typos and grammaros in comments of the new test
Stefano Lattarini [Thu, 22 Dec 2011 17:17:26 +0000 (18:17 +0100)]
fix: typos and grammaros in comments of the new test

* tests/get-sysconf.test: Fix few typos, grammaros and botched
wording.  Reported by Eric Blake.

12 years agotests: report useful system information in 'test-suite.log'
Stefano Lattarini [Thu, 22 Dec 2011 09:34:28 +0000 (10:34 +0100)]
tests: report useful system information in 'test-suite.log'

It has already happened various times that a user has run the
automake testsuite, experienced a failure, read the messages
telling him "See tests/test-suite.log" and "Please report to
bug-automake@gnu.org", and done exactly that -- sending us only
the contents of `tests/test-suite.log', which are usually not
enough to start debugging the reported failure.  So we have to
ask him for more details, and usually also for the `config.log'
file generated by configure.  It's time to fix this recurring
feedback inefficiency.  We do so by creating a dummy test case
that takes care of copying the contents of `config.log', plus
other useful system information, in the final `test-suite.log'.

* tests/get-sysconf.test: New test, gathering system information
and then always terminating with a SKIP, so that its output gets
copied in `test-suite.log'.
* tests/Makefile.am (TESTS): Add it.

12 years agopython: remove relics for Python 1.5 support
Reuben Thomas [Mon, 5 Dec 2011 22:40:48 +0000 (23:40 +0100)]
python: remove relics for Python 1.5 support

* m4/python.m4: The comments in here claim to support only
Python >= 2.0, yet this file still has specific support for
Python 1.5.  Just remove it, python 1.5 is 12 years old now,
and practically defunct.
* NEWS: Update.

See also commit `Release-1-10-205-gd5bec12', "Support for
Python 3.0, drop support for pre-2.0."

12 years agoconfigure: remove extraneous 'eval's from AM_RUN_LOG invocations
Stefano Lattarini [Wed, 21 Dec 2011 18:58:15 +0000 (19:58 +0100)]
configure: remove extraneous 'eval's from AM_RUN_LOG invocations

* configure.ac: Remove extra 'eval's from AM_RUN_LOG invocations;
for example, instead of "AM_RUN_LOG([eval $PERL --version])",
simply use "AM_RUN_LOG([$PERL --version])"

12 years agoconfigure: report TeX version in config.log
Stefano Lattarini [Wed, 21 Dec 2011 18:21:41 +0000 (19:21 +0100)]
configure: report TeX version in config.log

* configure.ac: If possible, report the version of the selected
TeX program; this should render the logs more informative.

12 years agomaint: snapshots from `maint' are still development snapshots
Stefano Lattarini [Thu, 22 Dec 2011 13:27:47 +0000 (14:27 +0100)]
maint: snapshots from `maint' are still development snapshots

The maintenance-oriented development line in the `maint' branch,
while being usually pretty stable and 99% backward-compatible,
is not always right off production-quality; but until now, the
Automake package version declared in configure.ac hid this fact,
since it appeared to be the version of a stable release (e.g.,
11.1).  Fix this.

* configure.ac (AC_INIT): Bump version to "1.11.0a"

12 years agoconfigure: print proper message for test releases
Stefano Lattarini [Thu, 8 Dec 2011 20:48:25 +0000 (21:48 +0100)]
configure: print proper message for test releases

* configure.ac: If the current release is detected to be a test
release or a development snapshot, print a proper warning for
the user.
* README-alpha: Delete, it's obsolete now (and in fact this file
hasn't been touched in eleven years, since release 1.4b or so).
* HACKING (Release procedure): Don't say to update README-alpha.

12 years agodevel: help in comparing Makefile.in from different commits
Stefano Lattarini [Fri, 9 Dec 2011 17:46:54 +0000 (18:46 +0100)]
devel: help in comparing Makefile.in from different commits

Now that the generated Makefile.in, configure and aclocal.m4 files
are no longer committed in Automake's git repository, a simple
"git diff" or "git log" no longer shows if and how a change in
Automake results in changes to the Makefile.in files and/or
configure script of its own build system.  Still, the ability to
peek so easily at such differences has proved itself quite useful
in the past, often revealing inconsistencies and blunders, and
sometimes even bugs; so it would be a pity to lose that altogether.

With this change, we add a new maintainer recipe that re-introduces
much of that capability, by generating and comparing on the fly the
Makefile.in, configure and aclocal.m4 derived from two arbitrary
commits of the Automake repository.

* Makefile.am (autodiffs, compare-autodiffs): New phony targets.

12 years agorepo: don't commit generated files in the git repository anymore
Stefano Lattarini [Thu, 8 Dec 2011 21:54:18 +0000 (22:54 +0100)]
repo: don't commit generated files in the git repository anymore

It has been quite some time since autoconf and libtool have stopped
committing the generated autotools files in their git repositories,
with no significant ill effects we're aware of.  It's true that the
autoconf bootstrap process has now the minor annoyance that a
pre-installed autoconf is required to complete it; but luckily
automake will not have a similar annoyance, since our bootstrap
script take care, through some hoops, to use the very automake and
aclocal versions from the current git checkout to generate the
required aclocal.m4 and Makefile.in files.  In fact, this has been
a necessity also in the past, because automake has been known to
use in its own build system new development features that hadn't
been present in any previously released automake distribution.

* aclocal.m4: Delete.
* configure: Likewise.
* Makefile.in: Likewise.
* doc/Makefile.in: Likewise.
* lib/Automake/Makefile.in: Likewise.
* lib/Automake/tests/Makefile.in: Likewise.
* lib/Makefile.in: Likewise.
* lib/am/Makefile.in: Likewise.
* m4/Makefile.in: Likewise.
* tests/Makefile.in: Likewise.
* .gitignore: Ignore configure, aclocal.m4, and all the
Makefile.in files.
* configure.ac (AC_PREREQ): New macro call, to require the
latest autoconf (2.68 for the moment).

12 years agoMerge branch 'missing-fixes' into maint
Stefano Lattarini [Thu, 22 Dec 2011 11:47:26 +0000 (12:47 +0100)]
Merge branch 'missing-fixes' into maint

* missing-fixes:
  missing: don't try to re-run tar with a munged command line
  missing: inform the user if GNU tar is called
  missing: if GNU tar exists but fails when called, give up
  missing: miscellaneous fixlets

12 years agomaint: remove executable bit from automake.in
Stefano Lattarini [Fri, 9 Dec 2011 08:51:21 +0000 (09:51 +0100)]
maint: remove executable bit from automake.in

* automake.in: This file is not meant to be executed, only to
be preprocessed to create the `automake' script; so don't leave
it executable.

12 years agomissing: don't try to re-run tar with a munged command line
Stefano Lattarini [Sun, 30 Oct 2011 10:55:28 +0000 (11:55 +0100)]
missing: don't try to re-run tar with a munged command line

* lib/missing: If the default `tar' program fails with the given
arguments, and GNU tar is not available, don't try to re-run the
default `tar' with a munged command line (e.g., ditching possibly
unportable options), as that could be subtly alter the intended
semantics (and maybe even create a somewhat corrupted tarball).
Also, it's worth noting that the main purpose of the `missing'
script is to allow a non-developer to build the package in the
face of slightly-skewed timestamps, not to provide wrappers for
all the maintainer tools -- so we don't have to try too hard
when `missing' is just called to wrap `tar'.
* tests/missing-tar.test: New test.
* tests/Makefile.am (TESTS): Add it.

12 years agomissing: inform the user if GNU tar is called
Stefano Lattarini [Sun, 30 Oct 2011 09:30:09 +0000 (10:30 +0100)]
missing: inform the user if GNU tar is called

* lib/missing: If the code trying to run GNU tar is reached, it
means that the previous attempt to run the default tar program
has failed, very likely producing some error message.  At this
point, just running GNU tar without further comments might be
confusing.

12 years agomissing: if GNU tar exists but fails when called, give up
Stefano Lattarini [Sun, 30 Oct 2011 09:12:28 +0000 (10:12 +0100)]
missing: if GNU tar exists but fails when called, give up

* lib/missing: If the code trying to run GNU tar is reached,
don't continue if the invoked GNU tar program fails, as there
is little point in doing so (and can even be confusing and
counter-productive).

12 years agomissing: miscellaneous fixlets
Stefano Lattarini [Sun, 30 Oct 2011 08:57:50 +0000 (09:57 +0100)]
missing: miscellaneous fixlets

* lib/missing: Some shells, such as Solaris or FreeBSD /bin/sh,
warn about missing programs before performing redirections.
Therefore, where we have to silently check whether a program
exists, perform redirections on a subshell.
Remove redundant uses of double-quotes in variable definitions.
Delete an extra blank line.

12 years agotests: fix spurious failure on systems lacking unistd.h
Peter Rosin [Tue, 20 Dec 2011 20:42:54 +0000 (21:42 +0100)]
tests: fix spurious failure on systems lacking unistd.h

This is for automake bug#10324.

* tests/silent-lex-generic.test (foo.l): Add a dummy #define of
YY_NO_UNISTD_H, so that the generated foo.c file won't require
unistd.h to be present (it is not present when compiling with,
e.g., MSVC 9).

12 years agotests: fix spurious failure on MSYS/MinGW and Cygwin
Stefano Lattarini [Tue, 20 Dec 2011 19:27:09 +0000 (20:27 +0100)]
tests: fix spurious failure on MSYS/MinGW and Cygwin

* tests/primary-prefix-couples-force-valid.test (Makefile.am):
Correctly append $(EXEEXT) to the path of built binaries.

12 years agoMerge branch 'master' into testsuite-work
Stefano Lattarini [Tue, 20 Dec 2011 13:16:25 +0000 (14:16 +0100)]
Merge branch 'master' into testsuite-work

12 years agoMerge branch 'maint'
Stefano Lattarini [Tue, 20 Dec 2011 13:14:08 +0000 (14:14 +0100)]
Merge branch 'maint'

* maint:
  test defs: hack to support autoconf-wrapper programs
  tests: fix a minor spurious failure with FreeBSD make
  tests: make two test scripts executable
  readme: reference webpages for automake mailing lists
  readme: update advice about testsuite execution
  readme: don't reference the old homepage at sources.redhat.com
  hacking: some more fixlets
  * NEWS: Fix typo in 'make dist-bzip2' description.
  release: don't run "make distcheck" automatically
  hacking: update on-line documentation for stable releases only
  hacking: described release procedure applies to beta releases too
  readme: the documentation is production quality now
  hacking: tell about platform-testers mailing list
  hacking: we don't use sources.redhat.com anymore
  hacking: miscellaneous minor fixes

12 years agotest defs: hack to support autoconf-wrapper programs
Stefano Lattarini [Fri, 16 Dec 2011 10:12:31 +0000 (11:12 +0100)]
test defs: hack to support autoconf-wrapper programs

* tests/defs.in ($AUTOCONF): Add a dummy `-B' option to the
autoconf invocation, so that, when the Debian autoconf wrapper
is involved, it will correctly dispatch an autoconf >= 2.50
instead of defaulting to autoconf 2.13.
($AUTOHEADER, $AUTORECONF): Likewise, but for autoheader and
autoreconf respectively.

Reported by Bruno Haible:
<http://lists.gnu.org/archive/html/automake/2011-12/msg00039.html>

12 years agotests: fix a minor spurious failure with FreeBSD make
Stefano Lattarini [Fri, 16 Dec 2011 00:46:14 +0000 (01:46 +0100)]
tests: fix a minor spurious failure with FreeBSD make

* tests/remake-am-pr10111.test: Avoid using `#' comments in
makefile recipes, as these have been confusing FreeBSD make.
The failure was masked by the fact that this test is currently
expected to fail.

Suggested by a report from Bruno Haible.

12 years agotests: make two test scripts executable
Stefano Lattarini [Fri, 16 Dec 2011 00:22:57 +0000 (01:22 +0100)]
tests: make two test scripts executable

* tests/remake-am-pr10111.test: Make executable.
* tests/remake-m4-pr10111.test: Likewise.

12 years agotests: fix spurious failure in 'color2.test'
Stefano Lattarini [Thu, 15 Dec 2011 22:52:46 +0000 (23:52 +0100)]
tests: fix spurious failure in 'color2.test'

* tests/color2.test: (expect-make): Add an "expect eof" directive,
so that the collected output from the spawned make program will be
displayed on stdout, as desired.
Since we are at it, also correctly remove a temporary file which
we was trying to delete using a wrong filename.

Problem introduced in merge `v1.11-1579-g8d3466c', probably by a
botched edit or conflict resolution.

12 years agotests: fix spurious failure in 'lex3.test'
Stefano Lattarini [Thu, 15 Dec 2011 22:12:07 +0000 (23:12 +0100)]
tests: fix spurious failure in 'lex3.test'

* tests/lex3.test (foo.l): Remove duplicated definition of
`yywrap'.  Problem introduced in merge `v1.11-1579-g8d3466c'.

12 years agoreadme: reference webpages for automake mailing lists
Stefano Lattarini [Wed, 14 Dec 2011 20:14:04 +0000 (21:14 +0100)]
readme: reference webpages for automake mailing lists

* README: Rationalize and reorganize the (brief) description of
the automake mailing lists.   In particular, instead of suggesting
the reader to write to the `-request' addresses to subscribe to
mailing lists, point him to the relevant webpages, where he can
also subscribe via a web form.

12 years agoreadme: update advice about testsuite execution
Stefano Lattarini [Wed, 14 Dec 2011 20:10:20 +0000 (21:10 +0100)]
readme: update advice about testsuite execution

* README: Now that the automake testsuite uses the parallel-tests
driver, there is no need for the user to capture the stdout of
"make check" to determine which tests have failed: a detailed log
is automatically saved into the `tests/test-suite.log' file.
Since we are at it, improve the wording by dropping an extra
"please".

12 years agoreadme: don't reference the old homepage at sources.redhat.com
Stefano Lattarini [Wed, 14 Dec 2011 19:53:10 +0000 (20:53 +0100)]
readme: don't reference the old homepage at sources.redhat.com

* README: Don't reference the old homepage at sources.redhat.com,
which is no longer active; reference the homepage on www.gnu.org
instead.  See also automake bug#10157 and bug#10248.
* tests/README: Likewise, and remove related extra-pedantic advice
about copyright papers for test cases (we'll ask for those papers
explicitly when we think they are warranted).

12 years agohacking: some more fixlets
Stefano Lattarini [Mon, 12 Dec 2011 09:20:00 +0000 (10:20 +0100)]
hacking: some more fixlets

* HACKING (Release Procedure): Place the list of pre-release
bootstrap-and-test commands on a line of its own, so it's easy to
select and then paste it into a terminal window.
Fix the explanation of "make git-release", as, since the previous
change, "make git-release" would simply run "make dist" rather
than "make distcheck".

Suggestion from Jim Meyering.

12 years ago* NEWS: Fix typo in 'make dist-bzip2' description.
Peter Rosin [Mon, 12 Dec 2011 08:10:56 +0000 (09:10 +0100)]
* NEWS: Fix typo in 'make dist-bzip2' description.

12 years agorelease: don't run "make distcheck" automatically
Stefano Lattarini [Sat, 10 Dec 2011 17:26:52 +0000 (18:26 +0100)]
release: don't run "make distcheck" automatically

* Makefile.am (git-dist): The developers should test extensively
before finally creating the release tarball; so don't run "make
distcheck" on their behalf here; instead ...
* HACKING (Release procedure): ... state here that "make check"
and "make distcheck" should be run before calling "make git-dist".

12 years agohacking: update on-line documentation for stable releases only
Stefano Lattarini [Mon, 5 Dec 2011 19:18:00 +0000 (20:18 +0100)]
hacking: update on-line documentation for stable releases only

* HACKING (Release procedure): On-line documentation at www.gnu.org
should be updated for stable releases only.

12 years agohacking: described release procedure applies to beta releases too
Stefano Lattarini [Mon, 5 Dec 2011 19:15:46 +0000 (20:15 +0100)]
hacking: described release procedure applies to beta releases too

* HACKING (Release procedure): Explicitly state that the described
procedure is valid for alpha/beta versions as well.

12 years agoreadme: the documentation is production quality now
Stefano Lattarini [Mon, 5 Dec 2011 19:10:18 +0000 (20:10 +0100)]
readme: the documentation is production quality now

* README-alpha: The documentation is now quite mature, no need
to decry it.

12 years agohacking: tell about platform-testers mailing list
Stefano Lattarini [Mon, 5 Dec 2011 18:57:07 +0000 (19:57 +0100)]
hacking: tell about platform-testers mailing list

* HACKING (Release procedure): Announces for alpha and beta releases
should be sent to the `platform-testers@gnu.org' mailing list.  Some
related rewording.

12 years agohacking: we don't use sources.redhat.com anymore
Stefano Lattarini [Mon, 5 Dec 2011 18:48:34 +0000 (19:48 +0100)]
hacking: we don't use sources.redhat.com anymore

* HACKING (Release procedure): Don't explain how to upload the
 web pages at sources.redhat.com; we don't use that anymore.

12 years agohacking: miscellaneous minor fixes
Stefano Lattarini [Mon, 5 Dec 2011 18:41:19 +0000 (19:41 +0100)]
hacking: miscellaneous minor fixes

* HACKING (Administrivia): If a commit fixes a bug registered at GNU
debbugs, its bug number be reported in the ChangeLog entry.  Re-order
the entries to give more visibility to the advice on how to verify
that a commit really fixes a bug.
(Working with git): Improve advice about which pre-existing branch
a topic branch should be based on.

12 years agoMerge branch 'master' into testsuite-work
Stefano Lattarini [Fri, 9 Dec 2011 23:58:13 +0000 (00:58 +0100)]
Merge branch 'master' into testsuite-work

12 years agoMerge branch 'maint'
Stefano Lattarini [Fri, 9 Dec 2011 23:46:13 +0000 (00:46 +0100)]
Merge branch 'maint'

Commits merged from maint:
  dist-xz, dist-bzip2: don't hard-code -9, honor envvar settings
  news: fix typos, grammaros and suboptimal wording
  maint: sync auxiliary files from upstream
  tests: fix spurious failures due to missing 'yywrap()' function
  depcomp: spelling fix
  tests: fix 'distcheck-override-infodir.test' on Cygwin
  cosmetics: typofix in comments
  coverage: undistributed '.am' and '.m4' files are diagnosed
  coverage: required but missing '.am' and '.m4' files are diagnosed
  coverage: expose automake bug#10111 in the testsuite
  fix: regenerate doc/Makefile
  texinfo: work around Solaris 10 xpg4 shell bug in install rules
  configure: report perl version in config.log
  tests: fix spurious failure with older install-info
  maintcheck: fix spurious failure in 'color2.test'
  tests: fix spurious error in 'uninstall-fail.test' on Solaris
  tests: fix typo in 'uninstall-fail.test'

Extra edits:
  * ChangeLog: Remove mention of commit 6da46f31 (2010-10-05), which
  has been superseded by commit c8e01d58 (2011-12-09).  Fix entry
  relative to that commit, to adapt to differences between maint and
  master.

12 years agodist-xz, dist-bzip2: don't hard-code -9, honor envvar settings
Jim Meyering [Fri, 9 Dec 2011 22:17:18 +0000 (23:17 +0100)]
dist-xz, dist-bzip2: don't hard-code -9, honor envvar settings

Before the present change, automake-generated `dist-xz' rule used
a hard-coded `xz -9'.  That was a problem because on this front,
xz differs from gzip and bzip2.  While the latter two don't incur
any run-time decompression penalty for using a higher compression
level, specifying -9 with xz imposes a potentially fatal virtual
memory requirement on any client that wants to decompress your
tar.xz file.
People have complained that a tarball compressed with -9 cannot
be uncompressed in a low-memory environment (wrt-based embedded).
Hence, instead of defaulting to -9, which is useful only for very
large tarballs, it defaults to -e (equivalent to -6e).  This
limits the default memory requirements imposed on decompressors,
yet still gives very good compression ratios.

* lib/am/distdir.am (dist-xz): Do not hard-code xz's -9: that made
it impossible to override.  Actually don't default to -9, either,
since that induced inordinately large virtual memory usage when
merely decompressing.  Instead, use its XZ_OPT envvar, defaulting
to -e if not defined.  Suggested by Lasse Collin.
(dist, dist-all) [?XZ?]: Likewise
(dist-bzip2): Similarly, do not hard-code -9, but do continue to
use -9 by default.  Honor the BZIP2 envvar.
(dist, dist-all) [?BZIP2?]: Likewise
* NEWS: Update.
* doc/automake.texi (The Types of Distributions): Describe the
newly enabled environment variables.

This is inspired to commit v1.11-389-g6da46f3, with additional
changes to reflect that the xz compression level should default
to -e, not -9.

12 years agonews: fix typos, grammaros and suboptimal wording
Stefano Lattarini [Fri, 9 Dec 2011 11:47:53 +0000 (12:47 +0100)]
news: fix typos, grammaros and suboptimal wording

* NEWS: Fix typos, grammaros and suboptimal wording.

Reported by Jim Meyering.

12 years agomaint: sync auxiliary files from upstream
Stefano Lattarini [Wed, 7 Dec 2011 20:02:39 +0000 (21:02 +0100)]
maint: sync auxiliary files from upstream

* lib/texinfo.tex: Synced from upstream, by "make fetch".
* lib/config.guess: Likewise.
* lib/config.sub: Likewise.

12 years agotests: fix spurious failures due to missing 'yywrap()' function
Stefano Lattarini [Mon, 24 Oct 2011 21:18:34 +0000 (23:18 +0200)]
tests: fix spurious failures due to missing 'yywrap()' function

The AC_PROG_LEX Autoconf macro does not diagnose a failure to find
the "lex library" expected to provide a `yywrap' function (function
which is required to link most lex-generated programs).  On the
contrary, when all the link attempts (i.e., with `-ll' and `-lfl')
fail, configure declares that no lex library is needed, and simply
proceeds with the configuration process -- only for the build to
possibly fail later, at make time.

This behaviour is intended; the Autoconf manual reads:
 ``You are encouraged to use Flex in your sources, since it is
   both more pleasant to use than plain Lex and the C source it
   produces is portable.  In order to ensure portability, however,
   you must either provide a function `yywrap' or, if you don't use
   it (e.g., your scanner has no `#include'-like feature), simply
   include a `%noyywrap' statement in the scanner's source.''

This AC_PROG_LEX behaviour is causing some spurious failures of
the Automake testsuite in environments which lack a proper library
providing `yywrap' (this happens for example on Fedora-based
systems).   The proper workaround is to simply provide a fall-back
implementation of `yywrap' in our lexers.

* tests/cond35.test: Provide a dummy `yywrap' function.
* tests/lex3.test: Likewise.
* tests/silent-lex-generic.test: Likewise.
* tests/silent-lex-gcc.test: Likewise.
* tests/silent-many-generic.test: Likewise.
* tests/silent-many-gcc.test: Likewise.

Reported by Jim Meyering:
<http://lists.gnu.org/archive/html/automake-patches/2011-10/msg00092.html>

Cherry-picked (with some edits) from commits `v1.11-1085-gb5c3968'
(master) and `v1.11-871-geb147a1' (testsuite-work).

12 years agodepcomp: spelling fix
Paul Eggert [Sun, 4 Dec 2011 11:11:30 +0000 (12:11 +0100)]
depcomp: spelling fix

* lib/depcomp (-h): Fix misspelling in usage diagnostic.

12 years agotests: fix 'distcheck-override-infodir.test' on Cygwin
Peter Rosin [Sat, 26 Nov 2011 13:54:00 +0000 (14:54 +0100)]
tests: fix 'distcheck-override-infodir.test' on Cygwin

* tests/distcheck-override-infodir.test (Makefile.am): Do not add
any `/' between $(DESTDIR) and the following paths.  Otherwise,
when $(DESTDIR) is empty, the recipes will try to access files
with a leading double slash, which have an implementation-defined
interpretation (e.g., for Cygwin, they mean UNC paths).

12 years agocosmetics: typofix in comments
Stefano Lattarini [Thu, 24 Nov 2011 11:14:49 +0000 (12:14 +0100)]
cosmetics: typofix in comments

* tests/remake-am-pr10111.test (Makefile.am): Fix typo in comments.
* THANKS: Update.

Reported by Krzysztof Żelechowski.

12 years agocoverage: undistributed '.am' and '.m4' files are diagnosed
Stefano Lattarini [Thu, 20 Oct 2011 09:55:01 +0000 (11:55 +0200)]
coverage: undistributed '.am' and '.m4' files are diagnosed

The stub rules emitted to work around the "deleted header problem"
for `.m4' files (included by autoconf in e.g., configure.ac) and
for `.am' files (included by automake in e.g., Makefile.am) should
not prevent "make" from correctly complaining when such a required
file is missing from a distribution tarball.
* tests/dist-missing-am.test: New test.
* tests/dist-missing-m4.test: Likewise.
* tests/dist-missing-included-m4.test: Likewise.
* tests/Makefile.am (TESTS): Add them.

Suggestion by Ralf Wildenhues.

12 years agocoverage: required but missing '.am' and '.m4' files are diagnosed
Stefano Lattarini [Sun, 16 Oct 2011 15:07:19 +0000 (17:07 +0200)]
coverage: required but missing '.am' and '.m4' files are diagnosed

The stub rules emitted to work around the "deleted header problem"
for `.m4' files (included by autoconf in e.g., configure.ac) and
for `.am' files (included by automake in e.g., Makefile.am) should
not prevent the remake rules from correctly erroring out when a
still-required file is missing.

* tests/deleted-am.test: New test.
* tests/deleted-m4.test: Likewise.
* tests/Makefile.am (TESTS): Add them.

12 years agocoverage: expose automake bug#10111 in the testsuite
Stefano Lattarini [Tue, 22 Nov 2011 19:50:53 +0000 (20:50 +0100)]
coverage: expose automake bug#10111 in the testsuite

* tests/remake-am-pr10111.test: New test, xfailing.
* tests/remake-m4-pr10111.test: Likewise.
* tests/Makefile.am (TESTS, XFAIL_TESTS): Add them.

12 years agofix: regenerate doc/Makefile
Stefano Lattarini [Tue, 22 Nov 2011 10:52:40 +0000 (11:52 +0100)]
fix: regenerate doc/Makefile

12 years agotexinfo: work around Solaris 10 xpg4 shell bug in install rules
Stefano Lattarini [Sat, 19 Nov 2011 18:53:57 +0000 (19:53 +0100)]
texinfo: work around Solaris 10 xpg4 shell bug in install rules

* lib/am/texinfos.am (install-html-am): Use an extra variable
indirection to work around a bug in Solaris 10 /usr/xpg4/bin/sh.
Bug revealed by a failure of `txinfo21.test'.  See also:
<http://lists.gnu.org/archive/html/bug-autoconf/2011-11/msg00005.html>
<http://debbugs.gnu.org/cgi/bugreport.cgi?bug=10026#23>

12 years ago* lib/install-sh: Spelling fix in comment.
Paul Eggert [Sun, 20 Nov 2011 07:51:15 +0000 (23:51 -0800)]
* lib/install-sh: Spelling fix in comment.

12 years agoconfigure: report perl version in config.log
Stefano Lattarini [Thu, 17 Nov 2011 08:04:10 +0000 (09:04 +0100)]
configure: report perl version in config.log

* configure.ac: Report the version of the selected perl interpreter
in config.log; this should render the logs more informative.

12 years agotests: fix spurious failure with older install-info
Stefano Lattarini [Sun, 13 Nov 2011 09:25:43 +0000 (10:25 +0100)]
tests: fix spurious failure with older install-info

* tests/install-info-dir.test (foo.texi): Also add proper
`@settitle', `@dircategory' and `@direntry' directive, otherwise
older versions of `install-info' (e.g., 4.8) will fail to create
the `dir' index file in ${infodir}.

Problem revealed by a failure on NetBSD 5.1.

12 years agomaintcheck: fix spurious failure in 'color2.test'
Stefano Lattarini [Sat, 12 Nov 2011 21:24:36 +0000 (22:24 +0100)]
maintcheck: fix spurious failure in 'color2.test'

* tests/color2.test: Avoid creative quoting to avoid a spurious
failure of the `sc_tests_Exit_not_exit' maintainer check.

12 years agotests: fix spurious error in 'uninstall-fail.test' on Solaris
Stefano Lattarini [Sat, 12 Nov 2011 17:41:53 +0000 (18:41 +0100)]
tests: fix spurious error in 'uninstall-fail.test' on Solaris

* tests/uninstall-fail.test: Solaris 10 /usr/xpg4/bin/sh can add
a line number before the `:' in the error messages issued by shell
builtins.  Account for that in our grepping of make output.

12 years agotests: fix typo in 'uninstall-fail.test'
Stefano Lattarini [Sat, 12 Nov 2011 17:25:38 +0000 (18:25 +0100)]
tests: fix typo in 'uninstall-fail.test'

* tests/uninstall-fail.test: Always use `$rm_f_is_silent_on_error'
instead of the bogus `$rm_f_is_silent_on_failure'.

12 years agoMerge branch 'maint'
Stefano Lattarini [Thu, 10 Nov 2011 11:55:00 +0000 (12:55 +0100)]
Merge branch 'maint'

* maint:
  tests: avoid a spurious failure of 'ltinit.test' MinGW
  tests: testsuite is now safe to run with dmake in parallel mode
  tests: fix spurious failures w.r.t. parallel make and colorization
  tests: fix spurious failure in 'distcheck-override-infodir.test'
  tests: avoid another failure of 'uninstall-fail.test' on Solaris

12 years agotests: avoid a spurious failure of 'ltinit.test' MinGW
Stefano Lattarini [Thu, 10 Nov 2011 11:43:54 +0000 (12:43 +0100)]
tests: avoid a spurious failure of 'ltinit.test' MinGW

* tests/ltinit.test: Be laxer in grepping configure output, to
avoid spurious failures on systems which lack POSIX dynamic
linking (e.g., MinGW), or when cross-compiling for such systems.

See also commit `v1.11-855-ge9e5d4a'.

Report and suggestion from Peter Rosin.

12 years agotests: testsuite is now safe to run with dmake in parallel mode
Stefano Lattarini [Tue, 8 Nov 2011 18:50:29 +0000 (19:50 +0100)]
tests: testsuite is now safe to run with dmake in parallel mode

* tests/defs.in: Unset variables DMAKE_CHILD, DMAKE_DEF_PRINTED
and DMAKE_MAX_JOBS, which are exported by Solaris dmake when run
in parallel mode, and which might confuse make processes spawned
by our testsuite.

12 years agotests: fix spurious failures w.r.t. parallel make and colorization
Stefano Lattarini [Tue, 8 Nov 2011 17:41:30 +0000 (18:41 +0100)]
tests: fix spurious failures w.r.t. parallel make and colorization

* tests/color2.test: Skip the test if the $MAKE program fails to
consider the standard output as a tty when spawned by `expect'.
This is required for make implementations, like FreeBSD make and
Solaris dmake, that redirect the output of recipes to temporary
files or pipes when run in parallel mode.  Since we are at it,
simplify the detection of a working `expect' program, and throw
in other minor simplifications.

12 years agoMerge branch 'fix-pr9579' into maint
Stefano Lattarini [Mon, 7 Nov 2011 15:07:25 +0000 (16:07 +0100)]
Merge branch 'fix-pr9579' into maint

* fix-pr9579:
  tests: fix spurious failure in 'distcheck-override-infodir.test'

12 years agotests: fix spurious failure in 'distcheck-override-infodir.test'
Stefano Lattarini [Mon, 7 Nov 2011 15:05:48 +0000 (16:05 +0100)]
tests: fix spurious failure in 'distcheck-override-infodir.test'

* tests/distcheck-override-infodir.test ($required): Add
'install-info'.

12 years agotests: avoid another failure of 'uninstall-fail.test' on Solaris
Stefano Lattarini [Mon, 7 Nov 2011 14:59:43 +0000 (15:59 +0100)]
tests: avoid another failure of 'uninstall-fail.test' on Solaris

* tests/uninstall-fail.test: On Solaris 10, if `/bin/rm' is run
with the `-f' option, it doesn't print any error message when
failing to remove a file (due to e.g., "Permission denied").
Yikes.  Cater to this incompatibility, by relaxing the test when
a faulty `rm' is detected.

12 years agoMerge branch 'master' into testsuite-work
Stefano Lattarini [Sat, 5 Nov 2011 20:39:54 +0000 (21:39 +0100)]
Merge branch 'master' into testsuite-work

* master:
  ar-lib: fix configure output for "unrecognized archiver interface"

12 years agoMerge branch 'msvc'
Stefano Lattarini [Sat, 5 Nov 2011 20:38:52 +0000 (21:38 +0100)]
Merge branch 'msvc'

* msvc:
  ar-lib: fix configure output for "unrecognized archiver interface"

12 years agoar-lib: fix configure output for "unrecognized archiver interface"
Stefano Lattarini [Sat, 5 Nov 2011 20:35:40 +0000 (21:35 +0100)]
ar-lib: fix configure output for "unrecognized archiver interface"

* m4/ar-lib.m4: Ensure that, even when an error is hit while trying
to determine the archiver interface kind, the "checking archiver
interface" message from configure is properly terminated before
an error message is printed, to avoid slightly garbled output.
* tests/ar4.test: Enhance.
* tests/ar5.test: Likewise.

12 years agotests: more faithful 'installcheck' support in few tests
Stefano Lattarini [Sat, 5 Nov 2011 20:30:39 +0000 (21:30 +0100)]
tests: more faithful 'installcheck' support in few tests

* tests/ar-lib5a.test: Use the `$am_scriptdir' variable instead
of "$testsrcdir/../lib" or "$top_testsrcdir/lib", to test more
faithfully under "make installcheck".
* tests/ar-lib5a.test: Likewise.
* tests/libobj-basic.test: Likewise.
* tests/libobj19.test: Likewise.
* tests/suffix5.test: Likewise.

12 years agotests: remove redundant settings of `errexit' shell flag
Stefano Lattarini [Sat, 5 Nov 2011 19:41:49 +0000 (20:41 +0100)]
tests: remove redundant settings of `errexit' shell flag

* tests/aclocal-path-install-serial.test: Do not set the
`errexit' shell flag, as it is already set by `tests/defs'.
* tests/ar-lib2.test: Likewise.
* tests/ar-lib3.test: Likewise.
* tests/ar-lib4.test: Likewise.
* tests/ar-lib5a.test: Likewise.
* tests/ar-lib5b.test: Likewise.
* tests/ar-lib6a.test: Likewise.
* tests/ar-lib6b.test: Likewise.
* tests/ar-lib7.test: Likewise.
* tests/ar3.test: Likewise.
* tests/ar4.test: Likewise.
* tests/ar5.test: Likewise.
* tests/extra-portability.test: Likewise.
* tests/extra-portability2.test: Likewise.
* tests/extra-portability3.test: Likewise.
* tests/install-info-dir.test: Likewise.
* tests/maintmode-configure-msg.test: Likewise.

12 years agogenerated tests: avoid extra "FIXME" comments
Stefano Lattarini [Sat, 5 Nov 2011 19:33:19 +0000 (20:33 +0100)]
generated tests: avoid extra "FIXME" comments

* tests/gen-testsuite-part: Avoid putting too much "FIXME"
comments in the generated test scripts, since they end up
being just counter-productive noise when one has to grep
the test scripts looking for real "FIXME" items.

12 years agoMerge branch 'prove' into testsuite-work
Stefano Lattarini [Sat, 5 Nov 2011 15:22:22 +0000 (16:22 +0100)]
Merge branch 'prove' into testsuite-work

* prove:
  cosmetics: fix typo in 'prove-runner' script

12 years agoMerge branch 'master' into testsuite-work
Stefano Lattarini [Sat, 5 Nov 2011 10:52:55 +0000 (11:52 +0100)]
Merge branch 'master' into testsuite-work

Commits merged from master:
 - test defs: new functions to analyze configure help screen
 - warnings: fix buglets for portability warnings
 - tests: extend tests on 'extra-portability' warning category

Extra edits to the testsuite:
* tests/extra-portability.test: Use `$am_original_AUTOMAKE'
instead of the obsolete `$original_AUTOMAKE'.
* tests/extra-portability3.test: Likewise.
* tests/list-of-tests.mk: Update.