install-sh: a slightly better diagnostic, and tests enhancements
* lib/install-sh: When called with no non-option arguments and the '-t'
option with an argument that is not an existing directory, have the
diagnostic output complain about the lack of required arguments rather
than about the bad argument passed to '-t'.
* t/install-sh-unittests.sh: Enhance to also check diagnostic printed
in cases of expected failure.
testsuite harness: report test exit status in log file
The exit status of a test should be reported in the test logs, so
that one can see at a glance whether the test has succeeded or failed,
without having to look also into the corresponding .trs file.
This fixes automake bug#11814.
* lib/test-driver: Also report the test script exit status in the
test log (as the last line).
* t/check-exit-status-reported.sh: Test this new behaviour.
* t/list-of-tests.mk: Add the new test.
* t/ax/test-lib.sh( am_exit_trap): No longer log the test exit status;
this has been made redundant by the change to 'test-driver'. While at
it, fix an imperfect quoting.
TAP driver: no need to invoke AC_PROG_AWK directly
It is already required by AM_INIT_AUTOMAKE anyway.
* doc/automake.texi: Adjust examples.
* t/tap-doc2.sh: Adjust documentation-tracking test.
* m4/init.m4 (AM_INIT_AUTOMAKE): Explicitly tell that the AC_PROG_AWK
requirement is also needed whenever the TAP driver is used.
TAP driver: remove perl implementation (move it into contrib/)
That implementation was only meant as a standard against which the
portable awk+shell implementation was to be measured. Now, since
Automake 1.12, the latter implementation is fully functional and
already used in the wild, and in fact feature-par with the perl
implementation. So the perl implementation is now just slowing
down and complicating our testsuite. Let's move it to 'contrib/'
(we don't want to remove it, in case someone is actually using it
in the wild).
* lib/tap-driver.pl: Move it ...
* contrib/tap-driver.pl: ... here. While at it, convert quoting
`like this' to quoting 'like this', and remove an obsolescent FIXME
comment.
* lib/Makefile.inc (dist_script_DATA): Drop '%D%/tap-driver.pl'.
* Makefile.am (EXTRA_DIST): Add 'contrib/tap-driver.pl'.
* doc/automake.texi: Remove one stray reference to 'tap-driver.pl',
and reference 'tap-driver.sh' instead, as intended.
* t/ax/am-test-lib.sh ($am_tap_implementation): Delete definition and
uses.
(fetch_tap_driver): Simplify to unconditionally assume the shell+awk
implementation of the TAP driver is used.
(get_shell_script): Make more flexible so that it can cater to the
needs of 'fetch_tap_driver()'.
* t/tap-bad-prog.tap: Likewise.
* t/tap-bailout-leading-space.sh: Likewise.
* t/tap-signal.tap: Likewise.
* t/tap-test-number-0.sh: Likewise.
* t/test-driver-cond.sh: Use 'tap-driver.sh' instead of 'tap-driver.pl'.
* gen-testsuite-part (%test_generators): Do not generate sister tests
that use the perl TAP driver rather than the shell+awk one.
* NEWS: Update.
* micro:
post-release: micro version bump to 1.14.1a devel version
release: stable micro release 1.14.1
HACKING: minor clarification
tests: make install-info-dir.sh print more debugging info
tests: remove too-brittle test tap-realtime.sh
maintainer: am-ft: add option to cater to clock skews
sync: update INSTALL, config.guess and config.sub from upstream
TAP driver: cosmetic fixes
* t/tap-realtime.sh: Delete. It has always been brittle, but now
it's also causing spurious failures when mawk is used as the awk
implementation in tap-driver.sh (see bug#14601).
* t/list-of-tests.mk: Adjust.
maintainer: am-ft: add option to cater to clock skews
* maintainer/am-ft: Add option '-S', giving a number of seconds to sleep
after copying the tarball to the remote system and before unpacking,
building and testing it. This is to cater to situations where the clock
of the remote system is skewed (in the past) w.r.t. the clock the local
system the tarball has been built on.
* modernize-install-sh:
cosmetics: untabify the install-sh script
install-sh: assume that "set -f" and "set +f" work...
install-sh: assume ${var:-value} works as expected
install-sh: assume 'dirname' is available and working correctly
* micro:
cosmetics: fix typo in a user-facing message in tests
automake: account for perl hash order randomization
tests: avoid use of intervals to capitalize letters
distcheck: don't allow overriding of --prefix and --srcdir by the user
tests: expose bug#14991 (relates to 'distcheck')
automake: account for perl hash order randomization
Try to explicitly order the keys of some perl hashes when looping
on them to do sanity/correctness checks and possibly display warning
messages; this should ensure a more reproducible output. Not really
a big deal, but I prefer to keep the order of such output reproducible
if possible.
Issue revealed by spurious testsuite failures with perl 5.18, as
reported in automake bug#14891. See also:
<http://search.cpan.org/dist/perl-5.18.0/pod/perldelta.pod#Hash_randomization>
<http://onionstand.blogspot.ie/2012/12/are-you-relying-on-hash-keys-being.html>
* lib/Automake/Variable.pm (variables): Explicitly order the values of
the returned Automake::Variable instances.
(variables_dump): Simplify, using the knowledge that 'variables()' now
sorts its output.
* t/preproc-errmsg.sh: Adjust.
install-sh: assume that "set -f" and "set +f" work...
... and disable/enable shell globbing, respectively. This is
mandated by POSIX, and supported even by Solaris 9 /bin/sh (one
of the most braindead shells we still support).
distcheck: don't allow overriding of --prefix and --srcdir by the user
Not through AM_DISTCHECK_FLAGS, nor through DISTCHECK_FLAGS. Apparently,
some packages got in the habit of relaying all the options passed to the
original ./configure invocation through to the configure invocations
in "make distcheck". This was causing problems, because it also passed
through the original --srcdir and --prefix options.
Fixes: expose bug#14991 (relates to 'distcheck')
* lib/am/distdir.am (distcheck): Pass the hard-coded --srcdir and
--prefix options *after* both the developer-defined options in
$(AM_DISTCHECK_FLAGS) and the user-defined options in $(DISTCHECK_FLAGS).
* t/list-of-tests.mk (XFAIL_TESTS): Remove the now-passing test
'distcheck-no-destdist-or-srcdir-override.sh'.
* doc/automake.texi (Checking the Distribution): Update.
* NEWS: Likewise.
* micro:
tests: fix spurious failure when zip is present but unzip is not
tests: fix spurious failure due to localization issues
NEWS: update with the changes since v1.14
docs: correct typos in the fix-timestamp.sh script
python: byte-compile nobase_*_PYTHON files only once
cosmetics: typofix in the 'missing' script
test: avoid false positives in 'cc-no-c-o' script
test harness: improve catching of usage errors in script 'test-driver'
tests: fix a spurious failure on NetBSD-current
am-ft: make the environment available earlier
NEWS: post-release tweaks (for 1.14.x series)
tests: avoid a spurious failure on MacOS X 10.6.8
tests: don't risk hanging on the 'cl' requirement
tests: fix spurious failure when zip is present but unzip is not
Fixes automake bug#15181.
* t/dist-formats.tap (have_compressor): When checking that zip(1), also
check for unzip(1), otherwise "make distcheck" will be unable to extract
the zip tarball it creates, which will cause spurious failures. While
at it, reorganize the existing code a bit.
* THANKS, NEWS: Update.
tests: fix spurious failure due to localization issues
Fixes automake bug#15237.
* t/autohdr-subdir-pr12495.sh: Ensure make is run in the C locale, so that
we can expect error messages in English when grepping its output.
* THANKS, NEWS: Update.
docs: correct typos in the fix-timestamp.sh script
* doc/automake.texi: Here. The original version of this example script
makes no sense at all, using 'configure' instead of the intended 'touch'
in few key places.
Copyright-paperwork-exempt: yes Signed-off-by: Peter Breitenlohner <peb@mppmu.mpg.de> Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
* t/ax/cc-no-c-o.in: Be more careful in determining whether both the
'-c' and '-o' options have been passed on the command line to the
compiler. In particular, do not spuriously complain in the face of
options like '-compatibility_version' or '-current_version' (seen on
Mac OS X 10.7).
* THANKS: Update.
test harness: improve catching of usage errors in script 'test-driver'
Fixes automake bug#14840.
* lib/test-driver: Catch and report usage errors where the caller has
forgotten to specify one of the mandatory options (--test-name,
--log-file, --trs-file) or has not passed any non-option argument.
Also, be sure to work correctly even when no '--' special argument
is passed to separate option from non-options arguments.
* THANKS: Update.
* tests/silent-custom.sh: Be prepared to handle creative
quoting in the output of the shell run for the make recipes
when the shell traces are active ("set -x").
* t/depcomp2.sh: Strip, from the redirected ./configure stderr, the
possible error message "rm: conftest.dSYM: is a directory", generated
by cleanup code that doesn't cater to the existence of *.dSYM
directories sometimes created by the compiler on MacOS X. This
"massaging" of ./configure stderr is legitimate, since the spurious
error message is due not to automake-related code, but to a know
buglet/limitation of either Autoconf or Mac OS X bundles gcc:
<http://lists.gnu.org/archive/html/bug-autoconf/2007-11/msg00017.html>
Actually, from that link it appears that the original Autoconf issue
had been fixed, but it must have been re-introduced in the meantime :-(
On the GNU/Linux boxes of some users that run our testsuite there
is a '/usr/local/bin/cl' executable, from the IRAF package:
<http://iraf.noao.edu/>
The test 'compile4.sh' (and other tests) try to invoke the 'cl'
command to check whether it's a Microsoft compiler; the IRAF cl
is an interactive program, so it hangs on such invocation. In
conclusion, the testsuite hangs for those users which have the
IRAF cl early in PATH.
Fix the issue by redirecting the input of cl from /dev/null when
invoking it, which is enough to prevent the cl program from IRAF
from hanging, and should have no effect on the behaviour of the
Microsoft compiler.
This change fixes automake bug#14707.
* t/ax/am-test-lib.sh (require_tool): Adjust the handling of
the 'cl' requirement.
* maint:
release: stable minor release 1.14
NEWS: one more minor fixlet
maint: port check-minimal-autoconf to VPATH builds
tests: avoid spurious failures in Linux -> MinGW cross-compilation mode
tests: simplify checks for some expected variables values in Makefiles
NEWS: improve and update wording
tests: fix/improve few heading comments
coverage: new test on Texinfo @include support
tests: tighten a grepping check
tests: more significant names for some tests
rename-tests: rework some code for clarity and safety
tests: cosmetic changes in t/extra-sources.sh
rename-tests: inform the user about the pre-filled commit msg
typofix: in comments in 'maintainer/rename-tests'
rename-tests: also "git add" list-of-tests.mk
tests: rename t/exsource.sh -> t/extra-sources.sh
tests: some improvements to Gettext tests
maint: version bump after beta release 1.13b
release: beta release 1.13b (will become 1.14)
automake: assume we can always pass '-o' to the C compiler
NEWS: fix typo
NEWS: on assuming "rm -f" without arguments work
docs: AM_PROG_CC_C_O: correct imprecise statements about it
NEWS: document deprecation of 'shar' and 'compress' dist formats
m4: rename minuso.m4 -> prog-cc-c-o.m4
tests: some tests make no sense if "$CC -c -o" doesn't work
AM_PROG_CC_C_O: don't rely on AC_PROG_CC_C_O, re-implement similar logic
compile: rewrite AC_PROG_CC with AM_PROG_CC_C_O contents
tests: avoid few lingering $MAKE redirections
tests: fix a spurious failure on NetBSD 5.1
texi: build version.texi and stamp-vti in srcdir
tests: fix a botched heading comment
tests: fix another spurious with FreeBSD make
tests: fix a spurious failure with FreeBSD make
tests: remove remaining exec bits ('maint' branch)
PLANS: subdir-objects: various updates
THANKS: update Akim's e-mail address
tests: less uses of "make -e"; avoid spurious failures in 'check-cc-no-c-o'
build: be more respectful of user-specified verbosity
check-cc-no-c-o: unify initializations in a single place
check-cc-no-c-o: avoid a spurious failure
build: fixup for building in a VPATH setup
Use AC_DEFUN_ONCE to define AM_PROG_CC_C_O
compile: avoid AC_PROG_CC messy rewrite
options: tiny simplification in dealing with incompatible versions
options: try to report as much errors as possible
refactor: fix few "inverted boolean" usages
options: better name for an internal function
options: more consistency in use of return statuses to report errors
options: tiny simplification in dealing with erroneous opts
options: consistently use return statuses to report errors
options: re-enable some sanity checks
THANKS: update Eric Blake's e-mail address
NEWS: typofix
news: document new 'subdir-objects' warning
PLANS: one minor fixlet (mostly cosmetic)
PLANS: we have already dropped support for split info files in master
NEWS: fix a reference to Automake 1.14 where Automake 2.0 was intended
PLANS: fix reference to non-existent 'next' branch
PLANS: fix botched version reference
maintcheck: fix two references to old location of aclocal and automake
dist: deprecated shar and tar+compress formats
am: prefer a shorter idiom where possible
maint: re-run "make update-copyright" ...
tests: avoid spurious failure with older flex (2.5.4)
build: move automake and aclocal in 'bin' subdir
build: break up monolithic Makefile.am in subdir-specific fragments
maint branch: we are going to become Automake 1.14
tests: typofixes in comments in t/preproc-c-compile.sh
tests: remove bashism from a test
tests: rename some with more descriptive names
typofix: in comments in t/extra2.sh
tests: fix botched cross-reference in a heading comment
automake: refactoring: factor out common cpp-like flags
NEWS (mint): reflect new Automake versioning scheme
maintcheck: avoid spurious failure
perl: perl subroutine prototypes are problematic, don't use them
maint: more adjustments to the new versioning scheme
cosmetics: fix some "docstring-like" comments in automake
style: call perl functions 'like_this()', not '&like_this()'
preproc: enhance and extend tests
preproc: add support for relative names in included fragments
maint: use more perl subroutines prototypes in the automake script
build: auto-generate perl subroutines prototypes for automake and aclocal
refactor: rip module Automake::Language out of automake script
tests: more information about Lex and Yacc programs
lint: fix spurious failure for 'sc_rm_minus_f' syntax check
maint: bump version 1.13.1a -> 1.13.2a
plans: we are not going to remove AM_PROG_MKDIR_P in Automake 1.14
init.m4: add probe to check "rm -f" without args work
subdir-objects: complain if it isn't enabled
plans: update w.r.t. latest changes
ywrap: remove an obsolete FIXME comment
ywrap: style fixes (no semantic change intended)
convenience: "make lint" as an alias for "make maintainer-check"
docs: typofix in manual
coverage: using multiple lexers in a single program
tests: remove most uses of the AM_PROG_CC_C_O obsolete macro
coverage: obsolete macro AM_PROG_CC_C_O should cause no warning nor errors
INSTALL: update copyright years
ithreads: use runtime (not configure time) detection of perl threads
copyright: add few missing copyright notices
maint: files in PLANS are to be exempted from copyright notice
maint: consistently honor the UPDATE_COPYRIGHT_YEAR environment variable
copyright: update some copyright years
compile: use 'compile' script when "-c -o" is used with losing compilers
HACKING: suggest more checks before releasing
tests: can fake a compiler not grasping "-c -o" -- globally in all tests
sync: update files from upstream with "make fetch"
typofix: in comments in GNUmakefile
Rename 'maint/' -> 'maintainer/', for Git's sake
HACKING: minor typofix
HACKING: bug-tracker, the PLANS directory, and how to plan "big" changes
HACKING: rewindable branches should live in the 'experimental/*' namespace
HACKING: fixlets about git branch rewinding policy
HACKING: commit messages are not to follow GCS ChangeLog rules too strongly
HACKING: "detailed explanation" in commit messages is almost mandatory
HACKING: we use "merge --log" even when merging master
HACKING: typofix
depend2.am: fix comments on verbosity of compilation rules
depend2.am: improve comments a little
plans: automake 1.14 is to assume "rm -f" with no args is OK
plans: we want to active subdir-objects unconditionally in automake 1.14
tests: adjust stale references to old test names
tests: rename the last aclocal test with dumb name
tests: fix an old botched change to an aclocal test
tests: fix some botched inter-test references in heading comments
coverage: compile rules used "-c -o" also with losing compilers
texi: remove extra verbosity in creation of dirstamp directory
coverage: user can avoid distributing '.info' pages
plans: add some on-going plans (already registered on the bug tracker)
docs: mention dist-hook help for EXTRA_DIST
texi: remove workaround for older Texinfo (4.1)
NEWS: improve wordings in entry deprecating suffix-less info files
build: don't enable 'color-tests' automake option explicitly
build: enable all warnings as fatal in our own build system
texi: Texinfo sources and CLEANFILES definition should co-exist peacefully
tests: make two new test executable
runtest: better command line API
tests: move runtest.in away from the top-lever directory
maint: move more maintainer files in the 'maint/' subdir
plans: add the "PLANS" directory
tests: more significant names for some tests
maint: add some of my maintainer-specific scripts
texi: deprecate hack about info files in CLEANFILES variables
texi: info files can be generated in the builddir
tests: simplify checks for some expected variables values in Makefiles
Do so by using our custom 'is' auxiliary script rather than grepping
the output from make. This is more natural, more robust, and often
shorter to write.
Unfortunately, we can't do that in all cases: sometimes we really need
to match the content of a variable against a regular expressions, and
we can't know nor are interested in its exact value.
This is basically a follow-up on commit v1.11-1830-g96401cb of
2012-02-08 (tests: better way to compare lists in Makefile rules).
* t/subst-no-trailing-empty-line.sh: Adjust.
* t/pluseq10.sh: Likewise.
* t/check5.sh: Likewise, and enhance a little while at it.
* t/check7.sh: Likewise.
* t/exeext.sh: Likewise.
Mostly to bring them more in sync with the ones in Automake-NG.
See also commit v1.12.2-824-g5468d52 of 2012-08-10([ng] tests:
reorganize gettext tests a bit) in Automake-NG.
* t/gettext.sh: Rename ...
* t/gettext-basics.sh: ... like this, enhance a little, and
move checks on requirement of 'config.rpath' out into ...
* t/gettext-config-rpath.sh: ... into this new test, and move
checks about PR/381...
* t/gettext-pr381.sh: ... into this new test.
* t/gettext2.sh: Rename ...
* t/gettext-external-pr338.sh: ... like this, and enhance a
little.
* t/gettext3.sh: Rename ...
* t/gettext-intl-subdir.sh: ... like this, and add trailing
':' command.
* lib/config.guess: This. Actually, only the timestamp has been
updated (apparently, it was mistakenly not updated in the previous
real change to the script).
lang, suffix rules: don't require C stuff needlessly
This change fixes automake bug#14560: when two or more user-defined suffix
rules were present in a single Makefile.am, automake would needlessly
include definition of some make variables related to C compilation in the
generated Makefile.in.
* automake.in (handle_languages): Fix logic to decide whether or not to
include definitions of C compilation related variables in the generated
Makefile.in: instead of doing so when two or more user-defined suffix
rules are seen (which is a completely bogus criterion), do so when two
or more compiled languages are used.
* lib/Automake/Rule.pm (suffix_rules_count): Remove as no longer used.
(@EXPORT): Adjust.
* t/list-of-tests.mk (XFAIL_TESTS): No longer list the test script
'suffix-extra-c-stuff-pr14560.sh', which now passes.
* NEWS: Update.
Automake needlessly generates definition of make variables related
to C compilation when two or more user-defined suffix rules are
present in a single Makefile.am.
* t/suffix-extra-c-stuff-pr14560.sh: New test, exposing the bug.
This test is still xfailing.
* t/no-extra-c-stuff.sh: New test, check for a potential related
regression. This regression is not actually present here (so this
test passes), but it still took place in our first attempt at
fixing bug#14560 -- so this test has proven to be actually useful.
* t/no-extra-makefile-code.sh: Improve comments, and tighten the
grepping checks a little.
* t/list-of-tests.mk (handwritten_TESTS): Add the new tests.
(XFAIL_TESTS): Add the new xfailing test.
* micro:
configure: remove an obsolete TODO comment
post-release: micro version bump (1.13.3a)
release: stable micro release 1.13.3
typofix: fix grammaro in comments in t/tags-pr12372.sh
* micro:
NEWS: minor fixlets, re-wording, and better text wrapping
tests: slightly stricter checks in t/cxx-demo.sh
tests: fix spurious failure when 'etags' program is Exuberant Ctags
tests: fix spurious failure due to missing sleeps
tests: fix spurious failure when 'etags' program is Exuberant Ctags
Fixes automake bug#14517.
* t/tags-pr12372.sh: If the 'etags' program in use supports the
'--langmap' option, use it. That is required to avoid spurious
failures with Exuberant Ctags (at least version 5.8), which by
default do not generate any tags for file extensions it doesn't
recognize.
* THANKS: Give credit to that bug's original reporter.
* t/remake-configure-dependencies.sh: Add few missing '$sleep'
invocations. I thought that the sleeps implicit in the configure
invocation were enough, but they were not, actually. Here is what
can happen:
1. The config.status script is generated by a configure run.
2. ./config.status and make are run.
3. The 'print-version' script is modified.
4. Since that script is listed in $(CONFIGURE_DEPENDENCIES),
autoconf is re-run.
5. On a fast-enough machine, the three steps 2-4 above, even
combined, might have taken less than a second to run;
6. If the filesystem doesn't have a sub-second timestamp
resolution, that means the newly-generated configure has
the same timestamp of the old config.status;
7. So, config.status is not re-run, and the Makefiles are
not updated.
8. Spurious failure!
projects / thirdparty / automake.git / log
