* lib/Makefile.am (installcheck-local): To verify that the installed
scripts are actually executable, simply use 'test -x', instead of
resorting to perl and its '-x' file operator. Today, 'test -x'
should today be portable to any non-museum system. Since we are at
it, improve diagnostic in case of failure.
deptrack: avoid few forks in configure (minor optimization)
* m4/depend.m4 (_AM_DEPENDENCIES): Use 'echo > file' instead of
'touch file' to ensure the creation of a dummy file. Adjust the
comment that explains why we can't simply use ': > file': the
problem this latter idiom tickles is present also with Solaris 10
/bin/sh, not merely with Solaris 8 /bin/sh.
* syntax-checks.mk (sc_perl_local_no_parens): Remove, we don't even
allow the use of local but for the '$_' and '$~' special variables.
(syntax_check_rules): Adjust.
* NEWS (New in 1.11.1a): State that we prefer to quote 'like this'
or "like this" rather than `like this', as according to the new GCS
recommendations.
automake: do not quote `like this', as per GCS recommendation
* automake.in: Convert comments, diagnostic and help screen to the
use of new quoting format 'like this' or "like this" rather than
`like this'.
* tests/stdlib2.test: Adjust.
* tests/ltinstloc.test: Likewise.
* tests/location.test: Likewise.
* tests/libtool8.test: Likewise.
* tests/ldadd.test: Likewise, and prefer trailing ':' over trailing
'Exit 0'.
pm: do not quote `like this', as per GCS recommendation
This patch converts the internal perl modules used by automake and
aclocal to the use of new quoting format 'like this' or "like this"
rather than `like this'.
* lib/Automake/ChannelDefs.pm, lib/Automake/Channels.pm,
lib/Automake/Condition.pm, lib/Automake/Configure_ac.pm,
lib/Automake/DisjConditions.pm, lib/Automake/FileUtils.pm,
lib/Automake/General.pm, lib/Automake/Item.pm,
lib/Automake/Rule.pm, lib/Automake/Getopt.pm,
lib/Automake/Options.pm, lib/Automake/Struct.pm,
lib/Automake/VarDef.pm, lib/Automake/Variable.pm,
lib/Automake/Version.pm, lib/Automake/XFile.pm: Update and
adjust quoting format throughout, in comments and diagnostic.
Some minor related rewordings and reformatting since we are
at it.
m4: do not quote `like this', as per GCS recommendation
This patch converts the files comprising the automake's own build
system to the use of new quoting format 'like this' or "like this"
rather than `like this'.
* m4/auxdir.m4, m4/ccstdc.m4, m4/depend.m4, m4/depout.m4, m4/tar.m4,
m4/init.m4, m4/maintainer.m4, m4/make.m4, m4/missing.m4, m4/vala.m4,
m4/mkdirp.m4, m4/sanity.m4, m4/silent.m4, m4/strip.m4: Update and
adjust quoting format throughout, in comments and diagnostic. Some
minor related rewordings and reformatting since we are at it. Bump
serial numbers.
build: do not quote `like this', as per GCS recommendation
This patch converts the files implementing the automake's own build
system to the use of new quoting format 'like this' or "like this"
rather than `like this'.
* bootstrap, configure.ac, syntax-checks.mk, Makefile.am,
lib/Makefile.am, m4/Makefile.am, syntax-checks.mk: Update quoting
format throughout, in comments and diagnostic. Some minor related
rewordings and reformatting since we are at it.
* HACKING: Don't describe the obsolete 'git-release' target anymore.
Instead, document the new and improved targets 'git-tag-release' and
'git-upload-release' (introduced in commit v1.11-674-gaf5f939 of
25-01-2012, "release: revamp rules to tag and upload the releases").
hacking: do not quote `like this', as per GCS recommendation
* HACKING: Quote 'like this' or "like this", not `like this'. This
is done for consistency with the new recommendations in the GNU
Coding Standards, and because I've come to actually prefer the '...'
and "..." quoting format over the `...' one.
am: do not quote `like this', as per GCS recommendation
This patch converts the automake-provided '*.am' fragments, and
related files, to the use of new quoting format 'like this' or
"like this" rather than `like this'.
* lib/am/check.am: Update quoting format throughout, in comments
and diagnostic. Some related rewordings, reformatting, and removal
of redundant commands since we are at it.
* lib/am/configure.am: Likewise.
* lib/am/dejagnu.am: Likewise.
* lib/am/depend2.am: Likewise.
* lib/am/distdir.am: Likewise.
* lib/am/inst-vars.am: Likewise.
* lib/am/install.am: Likewise.
* lib/am/lang-compile.am: Likewise.
* lib/am/lisp.am: Likewise.
* lib/am/ltlib.am: Likewise.
* lib/am/mans.am: Likewise.
* lib/am/progs.am: Likewise.
* lib/am/remake-hdr.am: Likewise.
* lib/am/subdirs.am: Likewise.
* lib/am/tags.am: Likewise.
* lib/am/texi-vers.am: Likewise.
* lib/am/texibuild.am: Likewise.
* lib/am/texinfos.am: Likewise.
* lib/am/yacc.am: Likewise.
tests: quote 'like this', not `like this', as per GCS recommendation
This patch converts the automake testsuite and related files, to
the use of new quoting format 'like this' or "like this" rather
than `like this'.
This is done for consistency with the new recommendations in the GNU
Coding Standards, and, well, because I've come to actually prefer
the '...' and "..." quoting formats over the `...' one.
* tests/README: Update quoting format throughout. Remove some
"excessive" use of quoting, and throw in minor rewording where
appropriate.
* tests/Makefile.am: Likewise, for comments.
* tests/CheckListOfTests.am: Likewise.
* tests/defs-static.in: Likewise.
* tests/defs: Likewise.
* tests/trivial-test-driver: Likewise.
* Many tests/*.tap, tests/*.test and tests/*.sh files: Likewise,
and for the diagnostic and informational messages as well.
* tests/gen-testsuite-part: Likewise, and for the generated tests
as well.
spy: "rm -f" and "rm -rf" without args are portable
According to recent testing by Eric Blake, it seems that all
non-museum rm(1) implementations don't complain if called without
file operands, *if* the '-f' option is used. This behaviour will
likely be mandated by future versions of the POSIX standard as
well.
We'd like to start assuming this rm(1) behaviour in future Automake
releases, in order to simplify some automake-provided cleanup rules.
So, for the moment, we add a new "spy" test in our test suite, to
try ensuring that all the rm(1) implementation we encounter "in the
wild" truly behaves as expected in this regard.
For more references, see this discussion on the bug-autoconf list:
<http://lists.gnu.org/archive/html/bug-autoconf/2012-02/msg00002.html>
which sprang from coreutils bug#10819:
<http://debbugs.gnu.org/cgi/bugreport.cgi?bug=10819>
and brought to the POSIX issue#542:
<http://austingroupbugs.net/view.php?id=542>
which has been accepted and will ensure that future version of the
POSIX standards will mandate that 'rm' shouldn't complain if called
without file operands, when the '-f' option is in use.
* tests/spy-rm.test: New test.
* tests/list-of-tests.mk: Add it.
tests: avoid spurious failures under "make distcheck"
* tests/tap-setup.sh: Pass also '-f' option to 'cp' invocations,
so that files that have already been copied in without write
permissions will still be overwritten where needed, instead of
causing a spurious testsuite failure. Issue seen when the
testsuite is run under "distcheck".
* tests/defs (get_shell_script): Remove target file before
overwriting it, for the same reason. Rework code flow a bit
since we are at it.
fixup: avoid spurious failure for 'parallel-tests-many.test'
* tests/parallel-tests-many.test: Adjust grepping of "make check"
and "make recheck" output to match the new format of the testsuite
summary. Remove redundant 'set -e' call since we are at it.
With this change, our testsuite now uses 'configure.ac' as the
name for the typical autoconf input, instead of the obsolescent
'configure.in' (which has been deprecated for several years, at
least since autoconf 2.50).
Most of the test cases changed by this commit have been
automatically modified with this sed command (using GNU
sed):
sed -i 's/\<configure\(\\\?\)\.in\>/configure\1.ac/g'
* syntax-checks.mk (sc_tests_no_configure_in): New syntax check.
(syntax_check_rules): Add it.
* tests/defs: Create stub for autoconf input in 'configure.ac',
not in 'configure.in'. Adjust comments.
($AUTOCONF, $AUTOHEADER, $AUTORECONF): Remove workaround for the
infamous debian autoconf wrapper: we don't need such workaround
anymore now that we name our autoconf input as 'configure.ac'.
For more information, see commit v1.11-564-g63da492 "test defs:
hack to support autoconf-wrapper programs" of 16-12-2011.
* tests/README: Use 'configure.ac', not 'configure.in'.
* Many many tests: Likewise.
* tests/backcompat3.test: Adjust to avoid spurious failures.
* tests/backcompat5.test: Likewise.
* tests/missing6.test: Likewise.
* tests/backcompat6.test: Likewise, and extend a bit since we
are at it.
Expose the command-line length limit issue that can affect the
Automake-generated parallel-tests harness, especially on systems
where this limit is smaller (e.g., MinGW/MSYS).
Suggestion by Bob Friesenhahn.
* tests/parallel-tests-many.test: New test. We have verified that
it actually exposes the bug#7868, as it passes when we opportunely
reduce the number of test cases in $(TESTS). Checked on NetBSD 5.1
64bit, Debian unstable 32bit, Solaris 10 64bit and Cygwin 1.5 32bit.
* tests/list-of-tests.mk (handwritten_TESTS): Add it.
* tests/Makefile.am (XFAIL_TESTS): Likewise.
tests: "make installcheck" shouldn't run the testsuite twice
* Makefile.am (installcheck-local): Remove. The "installcheck"
target provided by automake is already a recursive target, so
there's no need to have an 'installcheck-local' recipe that
manually recurse into 'tests/'. In fact, such a manual recursion
was causing the testsuite to be run twice by "make installcheck".
* NEWS: Get rid of form-feed characters (^L) throughout the file.
Use separator lines instead (composed by 76 '~' characters), or
just an extra blank lines for older less verbose entries (those
predating the 1.8 release).
(New in 1.11.3): Move description of planned future backward
incompatibilities that won't make it to 1.12 ...
(New in 1.11a): ... in here. Quote 'like this', not `like this'.
Other minor miscellaneous rewordings and fixlets.
Currently, automake offers a 'parallel-tests' option to enable the
use of the parallel testsuite harness in the generated Makefiles,
but no option to explicitly state the intention of using the older
serial testsuite driver (which is currently the default).
This makes the parallel test harness seems like a second-class
citizen (while actually it should be the other way around); more
importantly, it will prevent us from making the parallel harness
be the default one in future automake versions, since the users
of the serial harness would then have no way to instruct automake
to continue using it.
We solve all of this by introducing a new 'serial-tests' option.
* lib/Automake/Options.pm (_process_option_list): Recognize the
new option, and related minor refactoring.
* tests/serial-tests.test: New test.
* tests/list-of-tests.mk: Add it.
* doc/automake.texi: Update.
* NEWS: Likewise.
tests: fix a timestamp issue, and other minor buglets
Reported by Panther Martin in automake bug#10848.
* tests/conffile-leading-dot.test: Add a proper '$sleep' invocation,
to avoid spurious failures on fast systems without sub-second
timestamp resolutions. Add other minor related and unrelated
improvements and fixlets since we are at it.
automake: new option to print location of scripts and data files
With this change, we add a new automake option '--print-libdir'
that prints (on stdout) the path of the directory containing the
Automake-provided scripts (e.g., 'missing' and 'install-sh') and
data files (e.g., INSTALL or texinfo.tex).
Suggestion by Nick Bowler. See also automake bug#10473.
* automake.in (parse_arguments): Handle the new option.
(usage): Mention it.
* doc/automake.texi (automake Invocation): Document it.
* NEWS: Update.
* tests/print-dir.test: New test.
* tests/list-of-tests.mk: Add it.
docs: prefer the awk+sh TAP driver over the perl one
We had completed our shell+awk implementation of the TAP driver
months ago, but never documented it in the manual, continuing to
document only the "prototype" written in perl instead. Time to
fix this, before the 1.12 release.
* doc/automake.texi (Use TAP with the Automake test harness): Now
we document the use of the awk+shell implementation of our TAP
driver, rather than of the perl one.
* tests/tap-doc2.test: Adjust to use the shell+awk implementation
of the TAP driver.
tests: be stricter in determining whether "make -jN" works
* tests/parallel-tests3.test: With at least Solaris 10 CCS make,
"make -jN" does not fail, but merely prints a warning about
"DistributedMake -j option" being ignored. This was causing this
test case to fail spuriously. So we now use a stricter check in
determining whether make works in parallel mode. Since we are at
it, we try to cater also to Sun Distributed make, which understands
"-j N" but *not* "-jN".
* tests/parallel-tests-extra-programs.test: Unless $MAKE is GNU
make, don't trust the exit status of "$MAKE -k". Fixes a spurious
failure with OpenBSD 5.0 make and NetBSD 5.1 make.
parallel-tests: fix another BSD parallel make issue
When BSD make is run in parallel mode, it apparently strips any
leading directory component from the automatic variable '$*' (of
course, against what POSIX mandates). This was causing FreeBSD 9.0
make and NetBSD 5.1 make to spuriously fail with automake-generated
test harnesses if subdir tests were present *and* make was being
run in parallel mode. This issue affected also the Automake own
testsuite.
* lib/am/check2.am (am__set_b): New internal variable.
(%OBJ%, %EXT%.log, %EXT%$(EXEEXT).log): Use it to work around
the described BSD make issue.
* tests/parallel-tests3.test: Enhanced to expose the bug.
* tests/parallel-tests-subdir.test: Enhance a little, since we
are at it.
* NEWS: Update.
tests: avoid FreeBSD make VPATH issues in more tests (see bug#7884)
See also similar change 'v1.11-755-g818bc40' of 07-02-2012,
"tests: work around bug#7884 in many yacc/lex tests". The
rationale for this change is basically the same.
* tests/lex-lib-external.test: Use "yl_distcheck" instead of
bare "$MAKE distcheck" to avoid extra failures caused by
automake bug#7884.
* tests/lex-noyywrap.test: Likewise.
* tests/lex-libobj.test: Likewise.
* tests/man6.test: This test suffers from the same FreeBSD make
incompatibility in VPATH handling that is the source of automake
bug#7884. Since this is caused by rules that are defined in the
Makefile.am by the test itself, rather that being generated by
automake, the best thing to do is to skip this test if we detect
the presence of said make incompatibility (through the function
'useless_vpath_rebuild').
* tests/man8.test: Likewise.
* tests/parallel-tests-extra-programs.test: FreeBSD make seems to
write (some) diagnostic messages on stdout instead than on stderr,
so now we also grep the stdout for the expected make diagnostic.
fixup: always use correct path for automake/aclocal wrappers
Otherwise, bootstrapping from a git checkout won't work anymore.
Issue introduced in today's commit 'v1.11-1974-ga98e58f'.
* doc/Makefile.am ($(srcdir)/amhello-1.0.tar.gz): Export PATH in a
way that makes the automake and aclocal wrappers truly accessible
(the location of this wrappers has been modified in a recent commit,
but this makefile hadn't been updated accordingly).
(update_mans): Likewise.
The 'is' helper script, used by some test cases since commit
'v1.11-1830-g96401cb' of 08-02-2012, "tests: better way to compare
lists in Makefile rules", is not available in VPATH builds, since
it resides in $(srcdir), nor in $(builddir), and only $(builddir)
is added to $PATH in the test script. This was causing spurious
failures.
Solve this issue by re-organizing a bit the layout and placing of
wrapper and help scripts used by the testsuite. This reorganization
is by no mean complete, but should be enough to solve the VPATH
issues.
* tests/parallel-tests-empty-testlogs.test: When $(TESTS) is empty,
NetBSD 5.1 make ends up defining $(TESTS_LOGS) to ".log" rather than
to the empty string (as would be expected). This behaviour was
causing a check in this test case to fail. But that check was in
fact overly strict, since our recipes are smart enough to work
around the botched-up substitution. Also, the failure wasn't a
regression from Automake 1.11, since that version too would have
triggered the same error with NetBSD 5.1 make. So we just relax
the test a little to avoid extra noise in the testsuite.
tests: TAP-based tests should error if no plan was specified
After this change, it will still be OK for TAP-based tests not to
report any explicit TAP plan -- but they should state *explicitly*
that they want to do so, by calling the 'plan_' function with a
proper argument (currently, either "later" or "unknonwn").
Motivated by a recent issue introduced by yesterday's commit
'v1.11-1959-g0699a84' (pointed out by Peter Rosin, and fixed
by follow-up commit 'v1.11-1961-gea44190').
* tests/tap-functions.sh (plan_): Get rid of '$have_tap_plan_',
and refactor use and initialization of '$planned_' in a way that
will allow client script to get more information on which kind
of plan (if any) has been declared by the former 'plan_' invocation.
(skip_all_): Adjust.
* tests/defs (exit trap): Only print the "lazy" TAP plan if 'plan_'
has requested that *explicitly*.
* syntax-check.mk (sc_tests_tap_plan): Remove: it's obsolete now,
since a TAP-based test that doesn't call 'plan_' explicitly won't
print any TAP plan, and the TAP driver will catch that and report
an error.
(syntax_check_rules): Adjust.
Currently, the internal $(am__dry_run) macro can fail in weird ways
when $(MAKEFLAGS) contains shell metacharacters. Let's expose this
limitation in the testsuite (and fix a couple of related weaknesses
since we are at it).
* tests/make-dryrun.test: Moved ...
* tests/make-dryrun.tap: ... here, converted to TAP, and extended to
expose the described limitations. Also ...
(am_parallel_tests): Define this so that the 'gen-testsuite-part'
script won't generate a useless wrapper script.
* test/list-of-tests.mk: Update.
Peter Rosin [Fri, 17 Feb 2012 08:52:07 +0000 (09:52 +0100)]
maint: support amending the generated ChangeLog, and fix a typo
* Makefile.am (gitlog_to_changelog_options): Add support for
amending the generated ChangeLog.
(EXTRA_DIST): Update.
* .git-log-fix: New file with things to amend to the generated
ChangeLog, starting with a fix for a typo in v1.11-1963-g3b369e6
"maint: use AC_PACKAGE_BUGREPORT to avoid duplication" from
yesterday.
cosmetics: normalize copyright notice in some tests
* tests/lisp3.test: Normalize copyright notice to read the
same of that of other tests.
* tests/lisp4.test: Likewise.
* tests/lisp5.test: Likewise.
* tests/lisp6.test: Likewise.
* tests/lisp7.test: Likewise.
* tests/lisp8.test: Likewise.
* tests/lispdry.test: Likewise.
maint: use AC_PACKAGE_BUGREPORT to avoid duplication
* configure.ac: In the message reporting whether the user is about
to build an alpha or beta version, use the autoconf-provided
AC_PACKAGE_BUGREPORT macro instead of duplicating the bur reporting
address.
Today's commit 'v1.11-1959-g0699a84' has caused the generated
'depcomp*.tap' tests to call the 'plan_' function *before* it is
defined, and has also caused the extra checks required for better
libtool coverage to be run only when libtool is *not* in use,
rather than the other way around. Fix this mess.
Report by Peter Rosin.
* tests/gen-testsuite-part: Include './defs' in the generated
tests, rather than doing so from 'depcomp.sh'. Adjust the count
of TAP tests for 'depcomp-lt-auto.test' (it's 84, not 72).
* tests/depcomp.sh: Adjust accordingly, and fix inverted logic
about the decision of when to run extra checks for libtool.
build: clearer diagnostic if "make" is invoked before "./configure"
Since a GNUmakefile wrapper has been introduced in our build
system, a "make" invocation issued before a "./configure" one
has caused GNU make to die with the cryptic message:
GNUmakefile:18: Makefile: No such file or directory
GNUmakefile:19: /syntax-checks.mk: No such file or directory
make: *** No rule to make target `/syntax-checks.mk'. Stop.
* GNUmakefile: Detect when we are being run from an unconfigured
tree, and give helpful diagnostic.
This avoids a spurious maintcheck failure, and also simplifies
the 'depcomp.sh' helper script a little.
* tests/gen-testsuite-part: Create and place also a proper 'plan_'
call (to emit the TAP plan) in each generated 'depmod*.tap' test.
* tests/depcomp.sh: Remove 'plan_' invocations, and simplify a
little since we are at it.
* update-copyright: Also exclude the COPYING and INSTALL files
found in the top-level directory.
* COPYING, INSTALL: Revert update of copyright years done in
today's commit 'v1.11-1948-g641a5a4'.
Instead of tweaking or dumbing down some of our sources or tests
in order not to make them spuriously trigger few maintainer checks,
we tweak such checks instead.
* syntax-checks.mk (sc_tests_overriding_macros_on_cmdline): Also
allow the variable 'exp' to be overridden on the make command line.
Whitelist the test 'make-dryrun.test' where appropriate.
(sc_mkinstalldirs): When whitelisting the 'mkinstalldirs' occurrence
in 'lib/Makefile.am', do not look for the exact line number: it is
obviously too fragile.
* tests/parallel-tests-extra-programs.test: Run make with the '-k'
option where appropriate, to ensure it doesn't stop at the first
error, which would prevent the building of the extra programs we
expect it to build.
* tests/tap-signal.tap: Sending a SIGQUIT on Cygwin 1.5 can cause a
segmentation fault instead (sometimes at least). Don't let this
older bug pollute the results of our testsuite.
* tests/suffix3.tap (foo.zoo): This C++ file fails to compile with
older g++ (3.4.4) on Cygwin 1.5.25 if we #include <iostream>. Don't
do that, since it's not strictly required, and the "using namespace"
directive is already enough to ensure the file contents is valid C++
but invalid C.
* doc/automake.texi (Program variables): Don't list "-d -t" twice
in the examples of valid ways to specify the '-d' flag to Yacc;
instead, use "-d -t" and "-t -d". Add missing period.
tests: another spurious failure with older autoconf fixed
* tests/aclocal8.test (configure.in): Also call 'AC_INIT', it's
required by some older autoconf versions (e.g., 2.63).
* tests/acloca15.test: Likewise. Add trailing ':' command since
we are at it.
* tests/conffile-leading-dot.test: Relax grepping of 'config.status'
stderr, to cater for older autoconf. Issue revealed by a failure
with autoconf 2.63 on Cygwin 1.5.25.
update-copyright: don't touch files synced from external packages
* Makefile.am (update-copyright): Do not update copyright years of
files synced from external packages, as given by the '$(FETCHFILES)'
variable. Silence the recipe since we are at it.
tests: fix spurious failure when testing wildcards in EXTRA_DIST
* tests/extra10.test: Remove the "make distcheck" call, since, at
least with GNU make 3.82, it trips on the fact that $(srcdir) is
not explicitly used in the wildcard in EXTRA_DIST.
While the 'update-copyright' script shouldn't be needed by final
users, that's not a good reason to keep it out by the distribution,
as doing so might be unhelpful toward occasional but curious users
interested in peeking into the automake build system.
* yl-work-for-master: (44 commits)
tests: truly use Lex and Yacc selected by configure
news: fix grammaros related to Yacc
yacc/lex tests: avoid spurious skips with non-GNU make
fixup: remove double news entry
tests: workaround for shells with broken 'set -e'
tests: avoid spurious failure with non-bison yacc
tests: fix spurious failures due to missing 'yywrap()' function
fixup: list of yacc/lex tests was botched
tests: work around bug#7884 in many yacc/lex tests
yacc/lex tests: remove an unneeded workaround for bug#8485
yacc/lex tests: avoid too much failures with FreeBSD make
yacc-dist-nobuild.test: add reference to bug#7884
yacc tests: fix bug in 'yacc-cxx.test'
lex tests: fix spurious link errors on Solaris
lex: "make clean" removes .c files from non-distributed .l
lex tests: make test on Lex dependency tracking more "semantic"
lex tests: remove erroneous check about ylwrap distribution
yacc tests: "make clean" removes C++ files from non-distributed .y
tests: add a workaround for automake bug#8485
tests: add a couple of inter-tests reference
...
cscope: is only expected to work with GNU make in VPATH setup
The 'cscope' functionality does not properly handle VPATH rewrites;
so we explicitly document that, for now, it is only ensured to work
with GNU make when doing a VPATH build, and we adjust testsuite
requirements accordingly.
Issue revealed by a failure of 'cscope.test' with Solaris 10 CCS
make.
* doc/automake.texi (Tags): Document the limitations of the cscope
support. Couple of fixlets to formatting since we are at it.
* tests/cscope.test: Converted to TAP, and thus moved ...
* tests/cscope.tap: ... here. Only run tests with VPATH builds when
using GNU make. Adjust so that no C, C++, Fortran or Java compiler,
nor any Emacs program, will be required to run this test.
* tests/list-of-tests.mk: Adjust.
Jim Meyering [Sun, 12 Feb 2012 15:57:42 +0000 (16:57 +0100)]
maint: add a rule to use gnulib's update-copyright
* lib/update-copyright: New file, from gnulib.
* Makefile.am (FETCHFILES): Add update-copyright to the list.
(fetch): Fetch it.
(update-copyright): New rule.
check: fix test-suite.log formatting with NetBSD sed
* lib/am/check.am (am__rst_section): Tweak to ensure en extra final
newline is printed also with NetBSD sed. Issue revealed by a
failure of 'test-metadata-global-result.test'.
* tests/rst-formatting.test: New test, checking the internal
variables '$(am__rst_title)' and '$(am__rst_section)'.
* tests/list-of-tests.mk: Add it.
Some shells, like NetBSD 5.1 /bin/ksh, have a bug that causes
wildcards resulting from an unquoted parameter expansion not to
be expanded as expected in some cases:
$ touch a b c d
$ /bin/sh -c 'var="[ab]" && echo $var' # As expected.
a b
$ /bin/ksh -c 'var="[ab]" && echo $var' # Oops.
[ab]
$ /bin/ksh -c 'var="?" && echo $var' # But this as expected.
a b c d
This was causing a failure in our testsuite. Work around that.
See also:
<http://lists.gnu.org/archive/html/automake-patches/2012-02/msg00149.html>
* tests/extra11.test: Work around the described issue by using the
$(wildcard) builtin in the EXTRA_DIST definition. This is not a
problem, since this test was assuming GNU make anyway. Since we are
at it, enhance coverage a little by also running "make distcheck".
* tests/extra10.test: Similarly, enhance coverage by also running
"make distcheck", and fix a minor bug in the process. But continue
using a "bare" wildcard expression in EXTRA_DIST, without involving
the $(wildcard) builtin, to ensure coverage of similar usages "in
the wild".
fixup: fix test tap-signal.tap to avoid false negatives
Problem introduced by yesterday commit v1.11-1851-g79b2c6a,
"tests: avoid yet other spurious failures on NetBSD"
* tests/tap-signal.tap: Truly ensure the exit status of the make
call under test is correctly reflected in the TAP result displayed
by the corresponding 'command_ok_' invocation.
* maint:
fixup: do not tickle makefile-deps.test with the make_dryrun code
gitattributes: custom merge driver for ChangeLog not needed anymore
gitattributes: use "union" merge driver for NEWS
gitattributes: custom merge driver for ChangeLog not needed anymore
* .gitattributes: We don't maintain a version-controlled ChangeLog
file anymore, so we have no need to specify a custom merge driver
for it anymore. Update accordingly.
tests: truly use Lex and Yacc selected by configure
When 'yacc-work' was merged into 'yl-work-for-master' in commit v1.11-1910-ge6c40d4 (12-02-2012), the values for $YACC and $LEX
stopped being substituted directly by config.status, and started
being substituted by a Makefile rule using the contents of the
$(YACC) and $(LEX) makefile variables.
Unfortunately, our configure.ac was calling AM_SUBST_NOTMAKE on
'YACC' and 'LEX' (see commit v1.11-284-g34ca92b of 29-01-2011,
"tests: don't define YACC and LEX in the Makefiles", for a
rationale), so that the $(YACC) and $(LEX) in the Makefile simply
hold the default make-provided values (that is, "yacc" and "lex"
respectively), independently of what yacc and lex programs were
defined at configure time.
Apart from reducing coverage and creating confusion, this behaviour
also caused spurious failures on systems lacking a yacc and/or lex
programs.
Problem reported by Peter Rosin. References:
<http://lists.gnu.org/archive/html/automake-patches/2012-02/msg00131.html>
<http://lists.gnu.org/archive/html/automake-patches/2012-02/msg00132.html>
* configure.ac (AM_SUBST_NOTMAKE): Don't call anymore on 'YACC' nor
on 'LEX'.
Peter Rosin [Tue, 14 Feb 2012 01:51:55 +0000 (02:51 +0100)]
depcomp tests: avoid redundant runs for non-libtool test
When running the dependency tests without libtool, it's
not meaningful to run the tests three-fold, the libtool
configure options are simply ignored anyway.
* tests/depcomp.sh: Only run the tests three-fold when
libtool is in use.
tests: avoid yet other spurious failures on NetBSD
* tests/tap-signal.tap: Add a "strategically placed" extra 'echo'
command and a temporary silencing of xtraces; they are required
to avoid possible garbled output with NetBSD make, which would
miss some final newlines in the expected places and thus mess up
our TAP output.
* tests/test-driver-is-distributed.test: Rewrite to avoid having to
edit configure.in; such editing was subtly botched on NetBSD (the
'AC_CONFIG_AUX_DIR' call got misplaced, ending up *before* the
'AC_INIT' call), and that caused an hard-to-track bug. Since we
are at it, extend coverage a little.
tests: avoid spurious failure with NetBSD /bin/ksh
* tests/parallel-tests-fd-redirect-exeext.test (configure.in): Tweak
so that we don't end up with the generated configure script having
two nested subshells that start with '((':
Peter Rosin [Mon, 13 Feb 2012 20:21:48 +0000 (21:21 +0100)]
tests: increase coverage for depcomp tests
On platforms requiring that no undefined symbols exist in order
to build shared libraries (e.g. Windows DLLs), you have to
explicitly declare that the libtool library you are building
does not actually have any undefined symbols, for libtool to
even try to build it as a shared library. Without that
explicit declaration, libtool falls back to a static library
only, regardless of any --enable-shared flags etc.
tests/depcomp.sh (Makefile.am): Add -no-undefined if a libtool
library (.la) is built.
(src/Makefile.am): Likewise.
yacc/lex tests: avoid spurious skips with non-GNU make
* tests/defs (useless_vpath_rebuild): Fix botched logic that
was causing this function to always diagnose non-GNU make
implementations as being affected by the "useless rebuilds
in VPATH setup" bug (that affects FreeBSD make and causes
automake bug#7884).
* tests/dist-formats.tap: Some versions of the BSD shell wrongly
bail out when the 'errexit' shell flag is active and the left-hand
command in a "&&" list fails and that list is the *last* command
of a "case" statement. This was causing an incorrect initialization
of the '$missing_compressors' variable, and thus potential spurious
failures when non-existing programs were assumed to be available.
* tests/yacc-cxx.test: Some versions of the BSD shell wrongly bail
out when the 'errexit' shell flag is active if the left-hand
command in a "&&" list fails and such list is the *last* command
of a "for" or "while" loop or of an "if" construct. Work around
this issue.
* tests/check12.test: Likewise.
projects / thirdparty / automake.git / log
