platform/upstream/automake.git
11 years agoMerge branch 'maint'
Stefano Lattarini [Sun, 24 Jun 2012 08:12:43 +0000 (10:12 +0200)]
Merge branch 'maint'

* maint:
  tests: fix automatic re-execution of tests with Zsh
  tests: drop support for older Zsh shells
  tests: allow AM_TEST_RUNNER_SHELL to be overridden
  tests: fix setup for older Zsh

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: fix automatic re-execution of tests with Zsh
Stefano Lattarini [Sat, 23 Jun 2012 11:42:10 +0000 (13:42 +0200)]
tests: fix automatic re-execution of tests with Zsh

* defs: Use '$argv0' instead of '$0'.  With Zsh not started right
away in Bourne-compatibility mode, the latter will be the path not
of the test script itself, but of the file it's currently sourcing
-- i.e., in our case, './defs'.  This would cause the automatic
re-execution code to execute './defs' (basically a no-op) rather
than re-run the test correctly.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: drop support for older Zsh shells
Stefano Lattarini [Sat, 23 Jun 2012 09:16:08 +0000 (11:16 +0200)]
tests: drop support for older Zsh shells

* defs-static.in: Here.  The fact that such support has been broken for
almost a year (only fixed by today's commit 'v1.12.1-57-gf1e0300'),
causing no bug reports from anyone, shows that such support is not truly
warranted.  And it will get in the way of future improvements in the
handling of the exit trap (because bugs in older Zsh versions will
prevent some of our planned improvements).  So just drop it.
* t/README: Remove obsolete advice for working around bugs in older
versions of Zsh; instead, indicate version 4.3 is the oldest Zsh now
supported.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: allow AM_TEST_RUNNER_SHELL to be overridden
Stefano Lattarini [Sat, 23 Jun 2012 08:39:33 +0000 (10:39 +0200)]
tests: allow AM_TEST_RUNNER_SHELL to be overridden

* defs-static.in: Here.  This will make it easier for the maintainer
to run some self checks (like those in 't/self-check-exit.tap') with
different shells, to look for possible portability problems.  Fix a
typo (doubled "the") while we are at it.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: fix setup for older Zsh
Stefano Lattarini [Sat, 23 Jun 2012 08:34:29 +0000 (10:34 +0200)]
tests: fix setup for older Zsh

* defs-static.in: After the addition of TAP-based tests and our renaming
of "simple" test scripts from 'tests/foo.test' to 't/foo.sh', a test name
is valid if it matches the wildcard "*.sh" or "*.tap", not the wildcard
"*.test".  Adjust accordingly.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoMerge branch 'missing-not-touch-just-warn'
Stefano Lattarini [Fri, 22 Jun 2012 22:32:30 +0000 (00:32 +0200)]
Merge branch 'missing-not-touch-just-warn'

* missing-not-touch-just-warn:
  missing: do not touch timestamps; only warn for out-of-date files

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoMerge branch 'maint'
Stefano Lattarini [Fri, 22 Jun 2012 22:16:26 +0000 (00:16 +0200)]
Merge branch 'maint'

* maint:
  tests: avoid one last `...` command substitution in 'test-init.sh'
  maintcheck: guard against `...` for command substitution in test cases
  tests: more uses of $(...) over `...` for command substitution
  cosmetics: quote `like this', not 'like this', in a couple of tests
  readme: clarify/extend few entries in 't/README'
  readme: better separation of entries in 't/README'
  readme: subsections "Do" and "Do not" in 't/README' merged
  tests: new requirement 'grep-nonprint'
  cosmetics: fix description of an expected error message in a test
  tests: assume automake quotes 'like this', not `like this'

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: avoid one last `...` command substitution in 'test-init.sh'
Stefano Lattarini [Fri, 22 Jun 2012 22:12:08 +0000 (00:12 +0200)]
tests: avoid one last `...` command substitution in 'test-init.sh'

* t/ax/test-init.sh (me): In the definition of this variable.  This
also shave off a couple of forks, and the need for a sanity check
only required in fringe situations.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agomaintcheck: guard against `...` for command substitution in test cases
Stefano Lattarini [Fri, 22 Jun 2012 22:06:10 +0000 (00:06 +0200)]
maintcheck: guard against `...` for command substitution in test cases

But still allow them in configure.ac, Makefile.am and shell scripts
created or used inside test cases itself, because Autoconf (as of
version 2.69) does not yet ensure that $CONFIG_SHELL will be set to
a proper POSIX shell.

* syntax-checks.mk (sc_tests_command_subst): New check.
(syntax-check_rules): Add it.
* t/tap-global-log.sh: Minimal tweakings to avoid triggering the
new maintainer check.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: more uses of $(...) over `...` for command substitution
Stefano Lattarini [Fri, 22 Jun 2012 21:31:00 +0000 (23:31 +0200)]
tests: more uses of $(...) over `...` for command substitution

Somehow missed by the previous changes.  No big deal, fix them now.

* t/ax/depcomp.sh: Here.
* t/autodist.sh: And here.
* t/autodist-no-duplicate.sh: And here.
* t/autodist-subdir.sh: Ad here.
* t/remake11.sh t/self-check-me.tap: And here.
* t/perf/testsuite-summary.sh: And here.
* t/perf/testsuite-recheck.sh: And here.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agocosmetics: quote `like this', not 'like this', in a couple of tests
Stefano Lattarini [Fri, 22 Jun 2012 21:18:54 +0000 (23:18 +0200)]
cosmetics: quote `like this', not 'like this', in a couple of tests

* t/ansi2knr-no-more.sh: Here.
* t/aclocal-verbose-install.sh: And here.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoreadme: clarify/extend few entries in 't/README'
Stefano Lattarini [Fri, 22 Jun 2012 21:14:38 +0000 (23:14 +0200)]
readme: clarify/extend few entries in 't/README'

* t/README (Writing test cases): Here.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoreadme: better separation of entries in 't/README'
Stefano Lattarini [Fri, 22 Jun 2012 21:06:55 +0000 (23:06 +0200)]
readme: better separation of entries in 't/README'

* t/README (Writing test cases): Prepend different entries with a '*'
character acting like a bullet in a list.  This make different entries
better separated, visually-wise.  Minor related reformatting.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoreadme: subsections "Do" and "Do not" in 't/README' merged
Stefano Lattarini [Fri, 22 Jun 2012 21:01:37 +0000 (23:01 +0200)]
readme: subsections "Do" and "Do not" in 't/README' merged

* t/README (Writing test cases): Merge subsections "Do" and "Do not".
The distinction was rather artificial, and more confusing than helpful.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: new requirement 'grep-nonprint'
Stefano Lattarini [Fri, 22 Jun 2012 20:31:35 +0000 (22:31 +0200)]
tests: new requirement 'grep-nonprint'

* t/ax/test-inist.sh ($esc): New, a literal escape character.
(grep-nonprint): New requirement, check that the grep implementation
available that can handle non-printing characters correctly.
* t/color.sh: Use it instead of hand-rolled equivalent, and do not
(re)define '$esc' explicitly.
* t/color2.sh: Likewise.
* t/tap-color.sh: Likewise.
* t/vtexi4.sh: Likewise.
* t/parallel-tests-no-color-in-log.sh: Likewise.  Also prefer the use
of grep over that of $FGREP, because the new requirement only check
grep, and ensure we only grep non-printing characters from a pipe, to
avoid hitting a BSD grep limitation.
* t/parallel-tests-reset-term.sh: Likewise.
* t/ax/tap-summary-aux.sh: Use '$esc' instead of hard-coding the
literal escape character.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agocosmetics: fix description of an expected error message in a test
Stefano Lattarini [Fri, 22 Jun 2012 19:59:26 +0000 (21:59 +0200)]
cosmetics: fix description of an expected error message in a test

* t/vartypo2.sh: Here.  This reflects the change from `this style'
of quoting to 'this style'.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: assume automake quotes 'like this', not `like this'
Stefano Lattarini [Fri, 22 Jun 2012 19:44:15 +0000 (21:44 +0200)]
tests: assume automake quotes 'like this', not `like this'

* t/add-missing.tap: Here.  This slightly simplifies and/or enhances
some grepping checks on automake diagnostic.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoMerge branch 'maint'
Stefano Lattarini [Fri, 22 Jun 2012 16:55:01 +0000 (18:55 +0200)]
Merge branch 'maint'

* maint:
  tests: prefer using 'is_newest' over 'ls -t' hacks
  tests: implement is_newest as an auxiliary script, not shell function
  tests: fix some spurious failures in VPATH setup
  tests: work in VPATH setup again

11 years agotests: prefer using 'is_newest' over 'ls -t' hacks
Stefano Lattarini [Fri, 22 Jun 2012 16:15:21 +0000 (18:15 +0200)]
tests: prefer using 'is_newest' over 'ls -t' hacks

To verify that a file 'new' is newer than the file 'old', it's much
cleaner to use "is_newest new old" rather than the more clumsy
"ls -t". Adjust several of our tests accordingly.

* syntax-checks.mk (sc_tests_ls_t): New check, guard against uses
of "ls -t" and similar.
(syntax_check_rules): Add it.
* t/maken.sh: Adjust.
* t/autohdr4.sh: Likewise.
* t/extradep2.sh: Likewise.
* t/lex-depend-cxx.sh: Likewise.
* t/extradep.sh: Likewise.
* t/yacc-depend2.sh: Likewise.
* t/yacc-pr204.sh: Likewise.
* t/lex-pr204.sh: Likewise.
* t/yacc8.sh: Likewise.
* t/acloca13.sh: Likewise.
* t/lex-depend.sh: Likewise, and enhance.
* t/acloca14.sh: Likewise.
* t/aclocal7.sh: Likewise.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: implement is_newest as an auxiliary script, not shell function
Stefano Lattarini [Fri, 22 Jun 2012 14:18:28 +0000 (16:18 +0200)]
tests: implement is_newest as an auxiliary script, not shell function

This will allow to also use it in the makefile recipes used in our
test cases.

* t/ax/test-init.sh (is_newest): Remove.
* t/ax/is_newest: New script.
* Makefile.am (EXTRA_DIST): Add it.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: fix some spurious failures in VPATH setup
Stefano Lattarini [Fri, 22 Jun 2012 13:34:12 +0000 (15:34 +0200)]
tests: fix some spurious failures in VPATH setup

* t/self-check-cleanup.tap: No need to copy the 'ax/t/test-init.sh'
file over in our temporary directory.
* t/self-check-reexec.tap: Likewise.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: work in VPATH setup again
Stefano Lattarini [Fri, 22 Jun 2012 13:14:20 +0000 (15:14 +0200)]
tests: work in VPATH setup again

* defs: Drop overly paranoid sanity checks that was causing all the tests
to fail spuriously when run in a VPATH setup, with a message like:
"../t/nodef.sh: ./t/ax/test-init.sh: not found in current directory".
Those checks looked for invariants that, even if broken, would still
cause the test to fail very early and with a pretty clear error message
anyway.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoMerge branch 'maint'
Stefano Lattarini [Fri, 22 Jun 2012 12:52:21 +0000 (14:52 +0200)]
Merge branch 'maint'

* maint:
  tests: automatic re-execution works for non-POSIX shells too
  tests: use more POSIX shell features our test scripts

+ Extra non-trivial edits:

* t/ax/test-init.sh: Apply the diffs between past versions of
'defs' in maint and in master; i.e., remove handling of $required
entries 'texi2dvi-o' and 'makeinfo-html', and adjust to the fact
that the parallel testsuite harness is now the default.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: automatic re-execution works for non-POSIX shells too
Stefano Lattarini [Fri, 22 Jun 2012 09:40:39 +0000 (11:40 +0200)]
tests: automatic re-execution works for non-POSIX shells too

Some inferior shells are too greedy in parsing their input.  If a
non-POSIX Bourne shell (like Solaris 10 /bin/sh) was used to launch
one of our test scripts, it would fail unconditionally, because it
unexpectedly saw some (by it) unsupported constructs, notwithstanding
such constructs being placed *after* the code implementing automatic
test re-execution with a better shell.  In conclusion, the shell
bailed out like this:

    $ /bin/sh t/ar.sh
    $ t/ar.sh: syntax error at line 257: `is_newest_files=$' unexpected

By moving all the potentially problematic code in a separate file, to
be sourced only after the code for automatic re-execution with a better
shell, we ensure that inferior shell cannot see such code by mistake.

* defs: All code after automatic shell re-execution moved out ...
* t/ax/test-init.sh: ... to this new file.
* syntax-checks.mk (xdefs): Add it.
* Makefile.am (dist_check_DATA): Add it.  Also move in 'defs' from
a less explicit 'check_DATA' declaration.
(nodist_check_DATA):  Move in 'defs-static' from a less explicit
'check_DATA' declaration.
(check_DATA): Remove.
* t/self-check-sanity.sh: Remove, it was actually too hacky and brittle,
sanity-checking situations we don0t actually care about.
* t/list-of-tests.mk: Adjust.
* t/self-check-explicit-skips.sh: Adjust, and fix a botched heading
comments while we are at it.
* t/self-check-reexec.tap: Adjust.
* t/self-check-cleanup.tap: Likewise.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: use more POSIX shell features our test scripts
Stefano Lattarini [Thu, 14 Jun 2012 11:13:58 +0000 (13:13 +0200)]
tests: use more POSIX shell features our test scripts

Since commit 'v1.12-36-g2d68fd9' of 2012-05-07, "configure: search a
sturdy POSIX shell to be used in the testsuite", the shell running
our test script is assured to be a POSIX-conforming shell, so we can
use the more modern and flexible idioms and features that we couldn't
use when we also aimed at compatibility with non-POSIX Bourne shells,
like Solaris /bin/sh.

* t/README: Suggest to use POSIX shell features liberally in test cases,
with possible exception of Makefile recipes and configure shell code.
* Several tests: Adjust to use more POSIX shell features; e.g., $(...)
rather than `...`, $((...)) rather than `expr ...`, "if ! CMD; then ..."
instead of "if CMD; then :; else ...", and so on.
In several places, when using the 'test' built-in, prefer '-eq' over
'=' for numeric comparisons, and prefer "grep -c PATTERN FILE" over
"grep PATTERN FILE | wc -l".
Throw in other low-hanging easy improvements and fixlets while we are
at it.
* t/ax/depcomp.sh, t/ax/tap-summary-aux.sh, t/ax/tap-functions.sh,
defs, defs-static.in: Likewise.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoMerge branch 'maint'
Stefano Lattarini [Thu, 21 Jun 2012 11:15:05 +0000 (13:15 +0200)]
Merge branch 'maint'

* maint:
  docs: avoid failures with development version of Texinfo (4.13.90)
  typofix: s/test derivers/test drivers/ in check.am comments
  typofix: s/env/even/ in comments in GNUmakefile

11 years agodocs: avoid failures with development version of Texinfo (4.13.90)
Stefano Lattarini [Thu, 21 Jun 2012 11:13:04 +0000 (13:13 +0200)]
docs: avoid failures with development version of Texinfo (4.13.90)

* doc/automake-history.texi: Use '@item' instead of '@itemx' where
appropriate.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agomissing: do not touch timestamps; only warn for out-of-date files
Stefano Lattarini [Fri, 15 Jun 2012 09:13:00 +0000 (11:13 +0200)]
missing: do not touch timestamps; only warn for out-of-date files

Before this change, the missing script had a twofold role:

  - it warned the user if some required maintainer tools was missing,
    or too old;

  - in such a case, it tried to "fix" the timestamp of the files that
    should have been rebuilt by that tool (without actually updating
    the file contents, of course), to allow the build to continue.

The second capability used to be quite useful in the days when most
projects committed files generated by maintainer-only tools in their
VCS repository (today the trend is not to keep such generated files
VCS-committed anymore).  In such a setup, the "timestamp-fixing"
capability of 'missing' was quite useful, in that it allowed users
lacking some required maintainer tool to build from a VCS checkout
in the face of skewed timestamps (as could have been caused by
"cvs update" or "git checkout").

But then, when the automatic remake rules kicked in due to the
generated files being *actually out-of-date* (e.g., because the user
had modified 'configure.ac' but lacked a modern-enough autoconf to
rebuild it), that behaviour of 'missing' caused the same problem that
plagued AM_MAINTAINER_MODE; i.e., the user would get non-dependable
builds and inconsistent statuses of the build tree -- changes to
source files don't reflect on generated files, and this can be very
confusing and cause hard-to-spot errors).

So we now believe that the best approach to deal with timestamp-related
issues is not to have 'missing' to "automagically" try to resolve
them (with all the risk and brittleness entailed), but rather to
suggest those projects still keeping generated files committed in their
VCS to provide a proper (say) 'fix-timestamp.sh' script that touches
the timestamp of the checked-out files, to ensure no spurious rebuild
will be triggered.  As a bonus, such a script can be more aware of the
particularities, nooks and corner cases of a project, and thus more
reliable than the old 'missing' script.

An example of this approach is offered by GNU awk (release 4.0.1, Git
tag 'gawk-4.0.1', commit b85b04e8).  The GNU awk maintainers commit
the Autotools-generated files (configure, Makefile.in, etc.) and other
generated in the project's Git repository, but offer a useful script
'bootstrap.sh' that fixes the timestamps of those files, to ensure no
useless remake is triggered in a freshly cloned repository:

    #! /bin/sh
    # bootstrap.sh --- touch relevant files to avoid out-of-date issues
    #                  in Git sandboxes
    touch aclocal.m4
    find awklib -type f -print | xargs touch
    sleep 1
    touch configure
    sleep 2
    touch configh.in
    sleep 1
    touch test/Maketests
    find . -name Makefile.in -print | xargs touch
    touch doc/*.info
    touch po/*.gmo
    touch po/stamp-po
    touch awkgram.c
    touch command.c
    touch version.c

A similar, simplified script is also reported as an example in the
manual.

* NEWS: Update.
* doc/automake.texi (Auxiliary Programs, maintainer-mod): Update.
(CVS): Do not suggest that 'missing' can "automagically" fix botched
timestamp due to a "cvs update"; this isn't true anymore.  Give an
example of a custom 'fix-timestamp.sh' script that can be used to
obtain the same effect (in a slightly more laborious but also more
explicit and less brittle way).  Fix some minor typos and improper
wordings while we are at it.
* lib/missing: Basically rewritten to implement the new semantics.
As a side effect (one of the several), the '--run' option is no more
required nor recognized.
* m4/missing.m4 (AM_MISSING_HAS_RUN): Enhance the test on '$MISSING'
to ensure it actually provides the new semantics (by trying the new
"witness" option '--is-lightweight').
* lib/am/distdir.am (distdir): No need anymore to check for "bad"
distributed man pages that were actually dummy stubs generated by
the 'missing' script.
* t/missing4.sh: Rename ...
* t/remake-aclocal-version-mismatch.sh: ... like this.
* t/missing2.sh: Rename ...
* t/missing-version-mismatch.sh: ... like this, and adjust to the
new semantics.
* t/missing3.sh: Adjust to the new semantics.
* t/man4.sh: Remove as obsolete.
* t/missing.sh: Likewise.
* t/missing5.sh: Likewise.
* t/txinfo30.sh: Likewise.
* t/man6.sh: Adjust grepping checks.
* t/remake6.sh: Likewise.
* t/list-of-tests.mk: Adjust.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotypofix: s/test derivers/test drivers/ in check.am comments
Stefano Lattarini [Thu, 21 Jun 2012 09:43:54 +0000 (11:43 +0200)]
typofix: s/test derivers/test drivers/ in check.am comments

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotypofix: s/env/even/ in comments in GNUmakefile
Stefano Lattarini [Wed, 20 Jun 2012 21:32:44 +0000 (23:32 +0200)]
typofix: s/env/even/ in comments in GNUmakefile

Spotted by Eric Blake.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoMerge branch 'maint'
Stefano Lattarini [Wed, 20 Jun 2012 16:35:57 +0000 (18:35 +0200)]
Merge branch 'maint'

* maint:
  bootstrap: overhaul and improve
  docs: quote 'like this', not `like this'

11 years agobootstrap: overhaul and improve
Stefano Lattarini [Wed, 20 Jun 2012 16:28:02 +0000 (18:28 +0200)]
bootstrap: overhaul and improve

This fixes several weaknesses and buglets in the 'bootstrap' convenience
target offered in GNUmakefile (and its supporting code).  Refer to the
extensive code comments in there for more details.

* GNUmakefile: Almost completely rewritten.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agodocs: quote 'like this', not `like this'
Stefano Lattarini [Wed, 20 Jun 2012 15:27:57 +0000 (17:27 +0200)]
docs: quote 'like this', not `like this'

* doc/automake.texi: Fix the reported warnings and informative messages
from automake to quote 'like this' rather than as `like this'.  Do the
same for comments and some text in our examples.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agofixup: tests: special requirement 'makeinfo-html' removed from ./defs
Stefano Lattarini [Wed, 20 Jun 2012 12:23:28 +0000 (14:23 +0200)]
fixup: tests: special requirement 'makeinfo-html' removed from ./defs

This should have ideally been part of earlier commit v1.12.1-93-g3c64d54.

* defs (makeinfo-html): Remove.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agomaintcheck: guard against obsolete $required entries
Stefano Lattarini [Wed, 20 Jun 2012 11:49:18 +0000 (13:49 +0200)]
maintcheck: guard against obsolete $required entries

* syntax-check.mk: Add rules guarding against the use of the
obsolete test requirements 'texi2dvi-o' and 'makeinfo-html'.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: drop requirement 'makeinfo-html'; 'makeinfo' is enough
Stefano Lattarini [Wed, 20 Jun 2012 10:27:11 +0000 (12:27 +0200)]
tests: drop requirement 'makeinfo-html'; 'makeinfo' is enough

* defs: Since we now require Texinfo 4.9 or later in the Automake
generated Makefiles, and since the '--html' option of makeinfo, as
well as its capability to generate HTML output, has been supported
since Texinfo 4.0 (according to the Texinfo NEWS file), the
requirement 'makeinfo-html' is obsolete -- the simpler requirement
'makeinfo' is enough.
* t/instdir-texi.sh ($required): Adjust by using simply 'makeinfo'
instead of 'makeinfo-html'.
* t/silent8.sh: Likewise.
* t/txinfo21.sh: Likewise.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: drop requirement 'texi2dvi-o'; 'texi2dvi' is enough
Stefano Lattarini [Wed, 20 Jun 2012 09:52:20 +0000 (11:52 +0200)]
tests: drop requirement 'texi2dvi-o'; 'texi2dvi' is enough

* defs: Since we now require Texinfo 4.9 or later in the Automake
generated Makefiles, and since the '-o' option of texi2dvi has been
supported since Texinfo 4.1, the requirement 'texi2dvi-o' is
obsolete -- the simpler requirement 'texi2dvi' is enough.
* t/silent8.sh ($required): Adjust by using simply 'texi2dvi'
instead of 'texi2dvi-o'
* t/txinfo-no-clutter.sh: Likewise.
* t/txinfo13.sh: Likewise.
* t/txinfo23.sh: Likewise.
* t/txinfo24.sh: Likewise.
* t/txinfo25.sh: Likewise.
* t/txinfo28.sh: Likewise.
* t/txinfo33.sh: Likewise.
* t/vtexi4.sh: Likewise.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotexi: require Texinfo >= 4.9, related enhancements
Stefano Lattarini [Sun, 17 Jun 2012 10:23:18 +0000 (12:23 +0200)]
texi: require Texinfo >= 4.9, related enhancements

We start passing the '--build-dir' option to the texi2dvi and texi2pdf
invocations done in our Texinfo-related rules.  The argument for this
option will be specific for each info_TEXINFOS entry, to work around a
texi2dvi bug that could otherwise cause racy failures in parallel make
builds:
<http://lists.gnu.org/archive/html/automake-patches/2012-06/msg00073.html>
That bug seems already fixed in the development version of texi2dvi,
though:
<http://lists.gnu.org/archive/html/automake-patches/2012-06/msg00074.html>
so the details of our '--build-dir' usage can be revisited once we can
assume Texinfo 5.0 or later.

Since the '--build-dir' options has been only supported since Texinfo
4.9, we lose compatibility with all older Texinfo versions.

But we also get some real improvements.

First, the '--build-dir' option implies the '--tidy' option; this, like
the '--clean' option that we were formerly using, prevents cluttering of
the build directory with all the TeX and Texinfo auxiliary files and
build by-products, but, differently from '--clean', keeps them around
for later re-runs to use; this shortens the rebuild times considerably.

Second, the use of '--build-dir' also allows us to specify different
build directories for the PDF and DVI output, preventing the rules
building them from stomping on each other's feet when run in parallel.
This is demonstrated by the test cases 'txinfo-no-clutter.sh' and
'txinfo33.sh', which before this change used to fail when run with
MAKE="make -j4", but now succeed even in that case.

* NEWS: Update.
* lib/am/texibuilds.am: Do not try to avoid texi2dvi/texi2pdf option
'-o'; it has been supported since Texinfo 4.1, and we now require
Texinfo >= 4.9.  Remove an obsolete comment.  Modify calls to texi2dvi
and texi2pdf to use the '--build-dir' option, so that TeX auxiliary
files and build by-products for a 'foo.texi' input will be placed in
a 'foo.t2d' directory (for texi2dvi) or in a 'foo.t2p' directory (for
texi2pdf).
* automake.in (scan_texinfo_file): Greatly simplify the implementation,
since we don't need anymore to keep track of the TeX auxiliary files and
build by-products to clean.  Drop the '@CLEAN_FILES' part of the return
value.
(handle_texinfo_helper): Adjust the 'scan_texinfo_file()' call to its
new signature.  Append the "work directories" now created by texi2dvi
and texi2pdf invocation to the list of files/directories removed upon
"make mostlyclean".
* lib/am/texinfos.am: Update the cleaning rules (and comments)
accordingly.
* t/txinfo-no-clutter.sh: Minimal adjustment to avoid spurious failures.
* .gitignore: Update.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agodocs: it's not true that DISTCHECK_CONFIGURE_FLAGS is maintainer-reserved
Stefano Lattarini [Sat, 16 Jun 2012 07:16:53 +0000 (09:16 +0200)]
docs: it's not true that DISTCHECK_CONFIGURE_FLAGS is maintainer-reserved

At least, not anymore since commit 'v1.11-372-g9760039' of 2001-06-10,
"distcheck: add support for AM_DISTCHECK_CONFIGURE_FLAGS".  See also
automake bug#8784.

* doc/automake.texi (Flag Variables Ordering): Do not report
'DISTCHECK_CONFIGURE_FLAGS' as a "variables that are only useful
to the maintainer that has no user counterpart": now it is a
user-reserved variable, its maintainer-reserved counterpart
being 'AM_DISTCHECK_CONFIGURE_FLAGS' .

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoMerge branch 'am-preprocess-drop-tricky-transform'
Stefano Lattarini [Fri, 15 Jun 2012 13:07:05 +0000 (15:07 +0200)]
Merge branch 'am-preprocess-drop-tricky-transform'

* am-preprocess-drop-tricky-transform:
  automake: stop supporting "%KEY?iftrue:iffalse%" transforms
  am: stop using "%KEY:iffalse?iftrue%" transforms

11 years agoMerge branch 'maint'
Stefano Lattarini [Fri, 15 Jun 2012 08:27:27 +0000 (10:27 +0200)]
Merge branch 'maint'

* maint:
  tests: minor reorganization of few tests
  fixup: adjust t/list-of-tests.mk for renamed/added tests
  subdirs: enhance coverage, tweak and rename few tests
  sync: update files from upstream with "make fetch"
  py-compile: consistently quote 'like this', not `like this'.
  docs: recursive make considered harmful
  docs: clean rules are not run in reverse order of build rules anymore
  silent: new $(AM_V_P) variable, tell if we're running in silent mode
  refactor: silent rules handling (a little)
  refactor: &define_verbose_var: accept a third optional argument

+ Extra non-trivial edits:

These are due to the fact that support for silent rules is enabled
unconditionally these days (since commit 'v1.12-34-g14141f2' of
2012-05-01, "silent rules: support for them is always active now").

* automake.in: In the new silent related code, do not ever check
whether the 'silent-rules' option is active; just assume support
for silent rules is enabled.
* t/silent-obsolescent-warns.sh: Remove as obsolete.
* t/list-of-tests.mk: Adjust.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoMerge branch 'docs-recursion' into maint
Stefano Lattarini [Fri, 15 Jun 2012 08:09:19 +0000 (10:09 +0200)]
Merge branch 'docs-recursion' into maint

* docs-recursion:
  docs: recursive make considered harmful

11 years agoMerge branch 'silent-custom' into maint
Stefano Lattarini [Fri, 15 Jun 2012 08:09:04 +0000 (10:09 +0200)]
Merge branch 'silent-custom' into maint

* silent-custom:
  silent: new $(AM_V_P) variable, tell if we're running in silent mode
  refactor: silent rules handling (a little)
  refactor: &define_verbose_var: accept a third optional argument

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoautomake: stop supporting "%KEY?iftrue:iffalse%" transforms
Stefano Lattarini [Thu, 14 Jun 2012 21:45:08 +0000 (23:45 +0200)]
automake: stop supporting "%KEY?iftrue:iffalse%" transforms

And the similar "reduced forms" "%KEY:iffalse%" and "%KEY?iftrue%"
as well.

They are convoluted, never used (after the previous change), and will
get in the way in the Automake-NG branch, where we'll soon want to use
GNU make static pattern rules in our internal '*.am' fragments.

* automake.in (preprocess_file, transform): Simplify not to support
those transform patterns.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoam: stop using "%KEY:iffalse?iftrue%" transforms
Stefano Lattarini [Thu, 14 Jun 2012 21:41:58 +0000 (23:41 +0200)]
am: stop using "%KEY:iffalse?iftrue%" transforms

And the similar "reduced forms" "%KEY:iffalse%" and "%KEY?iftrue%"
as well.

They are harder to grasp, can be easily reformulated in function of
other "plainer" transforms ("?KEY?" and "?!KEY?"), and we'll remove
support for them anyway in a later change.

* lib/am/data.am, lib/am/libs.am, lib/am/lisp.am, lib/am/ltlib.am,
lib/am/progs.am, lib/am/python.am, lib/am/scripts.am: Adjusted not
to rely on those kinds of transforms.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: minor reorganization of few tests
Stefano Lattarini [Thu, 14 Jun 2012 21:27:42 +0000 (23:27 +0200)]
tests: minor reorganization of few tests

* t/subdir3.sh: Rename ...
* t/sourcefile-in-subdir.sh: ... like this.
* t/subdir4.sh: Rename ...
* t/depcomp-implicit-auxdir.sh: ... like this.
* t/subdir6.sh: Rename ...
* t/confh-subdir-clean.sh: ... like this.
* t/subdir7.sh: Renamed ...
* t/dir-named-obj-is-bad.sh: ... like this, and enhance.
* t/srcsub.sh, t/srcsub2.sh: Unify ...
* t/src-acsubst.sh: ... as this test.
* t/list-of-tests.mk: Adjust.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agofixup: adjust t/list-of-tests.mk for renamed/added tests
Stefano Lattarini [Thu, 14 Jun 2012 21:04:33 +0000 (23:04 +0200)]
fixup: adjust t/list-of-tests.mk for renamed/added tests

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agosubdirs: enhance coverage, tweak and rename few tests
Stefano Lattarini [Thu, 14 Jun 2012 20:47:28 +0000 (22:47 +0200)]
subdirs: enhance coverage, tweak and rename few tests

* t/subdir5.sh: Rename ...
* t/subdir-add-pr46.sh: ... like this.
* t/subdir8.sh: Rename ...
* t/subdir-add2-pr46.sh: ... like this.
* t/cond2.sh: Rename ...
* t/subdir-cond-err.sh: ... like this.
* t/subdir9.sh: Rename ...
* t/subdir-with-slash.sh: ... like this; improve heading comments.
* t/subcond.sh: Rename ...
* t/subdir-cond-gettext.sh: ... like this; improve m4 quoting.
* t/subcond2.sh: Rename ...
* t/subdir-am-cond.sh: ... like this; adjust heading comments.
* t/subcond3.sh: Rename ...
* t/subdir-ac-subst.sh: ... like this; adjust heading comments.
* t/subdir2.sh: Rename ...
* t/subdir-subsub.sh: ... like this; enhance so that it also runs
'./configure' and 'make'.
* t/subdir-order.sh: New test, check that the $(SUDBIRS) entries
are processed in the order they are specified.
* doc/automake.texi: Adjust references to tests.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agosync: update files from upstream with "make fetch"
Stefano Lattarini [Thu, 14 Jun 2012 14:32:47 +0000 (16:32 +0200)]
sync: update files from upstream with "make fetch"

* lib/config.guess, lib/gitlog-to-changelog, lib/texinfo.tex: Update.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agopy-compile: consistently quote 'like this', not `like this'.
Stefano Lattarini [Thu, 14 Jun 2012 14:30:16 +0000 (16:30 +0200)]
py-compile: consistently quote 'like this', not `like this'.

* lib/py-compile (usage_error): Here.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agodocs: recursive make considered harmful
Stefano Lattarini [Tue, 12 Jun 2012 15:24:21 +0000 (17:24 +0200)]
docs: recursive make considered harmful

In the documentation, we shouldn't confuse "packages using subdirectories"
with "packages using recursive make invocations".  A package can have a
careful organization in subdirectories, make no (or almost no) use of make
recursion (examples of such packages are GNU bison, GNU cppi, and Automake
itself).

In fact, make recursion should be used as seldom as possible, because it
makes the build system more brittle and dependency declarations less
faithful.  See Peter Miller's article "Recursive Make Considered Harmful"
for more a more in-depth discussion:

    http://miller.emu.id.au/pmiller/books/rmch/

* doc/automake.texi (Directories): Clarify how a build system based on
make recursion does.  Observe that such a system, albeit being very
widespread, has its own issues and drawbacks, and that one can have a
non-recursive setup also for projects using complex directory layout.
(Recursing subdirectories): Speak of "packages that use make recursion"
rather than of "packages with subdirectories".

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoMerge branch 'subdirs-simplify' into maint
Stefano Lattarini [Tue, 12 Jun 2012 14:12:45 +0000 (16:12 +0200)]
Merge branch 'subdirs-simplify' into maint

* subdirs-simplify:
  docs: clean rules are not run in reverse order of build rules anymore

11 years agodocs: clean rules are not run in reverse order of build rules anymore
Stefano Lattarini [Tue, 12 Jun 2012 14:10:19 +0000 (16:10 +0200)]
docs: clean rules are not run in reverse order of build rules anymore

At least since commit 'v1.12.1-12-gec6a135' of 2012-06-10, "subdirs: unify
rules for "cleaning" and "normal" recursive targets"

* doc/automake.texi (Recursing subdirectories): Adjust.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoMerge branch 'maint'
Stefano Lattarini [Mon, 11 Jun 2012 19:39:46 +0000 (21:39 +0200)]
Merge branch 'maint'

* maint:
  maintcheck: fix failures, both real and spurious
  subdir tests: avoid an use of "make -j4", for portability
  tests: avoid failure due to libtool quirks in C++ demo test
  tests: fix spurious failures due to missing '$sleep'
  subdirs: unify rules for "cleaning" and "normal" recursive targets
  tests: add a "demo" test on C support
  tests: look for '.lo' rather than '.o' object when using Libtool with C++
  cosmetics: few typofixes in older ChangeLogs, suggested by "codespell.py"
  maint: grammar fixes: s/all these/all of these/
  subdir-objects: improve "make mostlyclean" efficiency and flexibility

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoMerge branch 'subdir-objects-pr10697' into maint
Stefano Lattarini [Mon, 11 Jun 2012 19:27:12 +0000 (21:27 +0200)]
Merge branch 'subdir-objects-pr10697' into maint

* subdir-objects-pr10697:
  maintcheck: fix failures, both real and spurious

11 years agomaintcheck: fix failures, both real and spurious
Stefano Lattarini [Mon, 11 Jun 2012 19:21:32 +0000 (21:21 +0200)]
maintcheck: fix failures, both real and spurious

* syntax-checks.mk (sc_rm_minus_f): Whitelist the
't/subobj-clean*-pr10697.sh' tests.
* t/subobj-clean-pr10697.sh: When redefining PATH,
use '$PATH_SEPARATOR', not hard-coded ':'.
* t/subobj-clean-lt-pr10697.sh: Likewise.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: merged some testsuite fixlets
Stefano Lattarini [Mon, 11 Jun 2012 17:48:08 +0000 (19:48 +0200)]
tests: merged some testsuite fixlets

* fix-cxx-libtool-demo:
  tests: avoid failure due to libtool quirks in C++ demo test

* subdir-objects-pr10697:
  tests: fix spurious failures due to missing '$sleep'

* subdirs-simplify:
  subdir tests: avoid an use of "make -j4", for portability

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agosubdir tests: avoid an use of "make -j4", for portability
Stefano Lattarini [Mon, 11 Jun 2012 17:46:46 +0000 (19:46 +0200)]
subdir tests: avoid an use of "make -j4", for portability

Otherwise, a spurious failure with Solaris CCS make can be triggered.
The coverage is not actually reduced, since the code path is still
covered when one runs the testsuite with AM_TESTSUITE_MAKE="make -jN"
(as should be done periodically).

* t/subdir-distclean.sh : Call simply "make maintainer-check", not
"make -j4 maintainer-check".

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: avoid failure due to libtool quirks in C++ demo test
Stefano Lattarini [Mon, 11 Jun 2012 17:38:54 +0000 (19:38 +0200)]
tests: avoid failure due to libtool quirks in C++ demo test

Revealed by failures on NetBSD 5.1.

* t/cxx-lt-demo.sh: In "make distcheck" invocation, don't define 'CC'
to "false", as that value would be exported and passed to the child
./configure invocation, and some configure checks generated by libtool
autoconf macros can still require a C preprocessor even for packages
using only C++, and bail out if it's not found.  The problem was not
apparent on Solaris and GNU/Linux because those systems have a
'/lib/cpp' program, and configure detected and used that as a fallback
C preprocessor.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: fix spurious failures due to missing '$sleep'
Stefano Lattarini [Mon, 11 Jun 2012 16:52:38 +0000 (18:52 +0200)]
tests: fix spurious failures due to missing '$sleep'

Revealed by failures on NetBSD 5.1.

* t/subobj-clean-pr10697.sh: Call '$sleep' before modifying the
files that should trigger an automatic remake.
* t/subobj-clean-lt-pr10697.sh: Likewise.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoMerge branches 'subdirs-simplify' and 'subdir-objects-pr10697' into maint
Stefano Lattarini [Mon, 11 Jun 2012 15:35:35 +0000 (17:35 +0200)]
Merge branches 'subdirs-simplify' and 'subdir-objects-pr10697' into maint

* subdirs-simplify:
  subdirs: unify rules for "cleaning" and "normal" recursive targets
  tests: add a "demo" test on C support

* subdir-objects-pr10697:
  subdir-objects: improve "make mostlyclean" efficiency and flexibility
  tests: look for '.lo' rather than '.o' object when using Libtool with C++
  cosmetics: few typofixes in older ChangeLogs, suggested by "codespell.py"

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agosubdirs: unify rules for "cleaning" and "normal" recursive targets
Stefano Lattarini [Sun, 10 Jun 2012 21:08:07 +0000 (23:08 +0200)]
subdirs: unify rules for "cleaning" and "normal" recursive targets

Before this change, the recursive invocation of cleaning targets in
the $(SUBDIRS) where done in inverse order, i.e., starting from the
last $(SUBDIRS) entry and proceeding towards the first.  According
to the code comments, this was done ...

    ... in an attempt to alleviate a problem that can happen when
    dependencies are enabled.  In this case, the .P file in one
    directory can depend on some automatically generated header
    in an earlier directory.  Since the dependencies are required
    before any target is examined, make bombs.

But this comment does not apply anymore to the current implementation
of automatic dependency tracking: the '.Po' and '.Plo' files does not
depend on any C header or source file, ever!

So it seems that the distinction between "normal" and "cleaning"
recursive targets is a stale leftover of an older implementation of
the automatic dependency tracking.  In fact, the Automake History
manual seems to confirm this suspect; the section "First Take on
Dependency Tracking" reads:

    Because each .P file was a dependency of Makefile, this meant
    that dependency tracking was done eagerly by make.  For instance,
    "make clean" would cause all the dependency files to be updated,
    and then immediately removed.  This eagerness also caused problems
    with some configurations; if a certain source file could not be
    compiled on a given architecture for some reason, dependency
    tracking would fail, aborting the entire build.

and the following section "Dependencies As Side Effects" reads:

    In this approach, the .P files were included using the -include
    command, which let us create these files lazily.  This avoided
    the "make clean" problem.

So the distinction between "normal" and "cleaning" recursive targets
has likely been obsolete since by then already.  We can thus remove
such distinction, thus reducing some complications and duplication in
our rules.  Doing so, the whole testsuite still passes (both with GCC
and Sun C 5.9), even the test 'c-demo.sh', which, among the other
things, exercise the setup described in the obsolete code comment
referenced above.

Finally, note that we still keep '$(RECURSIVE_CLEAN_TARGETS)' and
'$(RECURSIVE_TARGETS)' as two distinct variables, to ensure a better
backward-compatibility for any user-defined rules that happen to use
those variables.

* NEWS: Update.
* lib/am/subdirs.am ($(RECURSIVE_CLEAN_TARGETS), $(CLEAN_TARGETS)):
Merge their recipes.
* t/subdir-distclean.sh: New test, check that "./configure && make
&& make distclean" is actually a no-op, even when conditional SUBDIRS
are involved.
* t/list-of-tests.mk: Add it.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: add a "demo" test on C support
Stefano Lattarini [Sun, 10 Jun 2012 22:49:18 +0000 (00:49 +0200)]
tests: add a "demo" test on C support

Showing and testing non-trivial use of C support, and its interaction
with other features.

* t/c-demo.sh: New test.
* t/list-of-tests.mk: Add it.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: look for '.lo' rather than '.o' object when using Libtool with C++
Stefano Lattarini [Sat, 9 Jun 2012 18:13:43 +0000 (20:13 +0200)]
tests: look for '.lo' rather than '.o' object when using Libtool with C++

* t/cxx-lt-demo.sh: ... in this test.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agocosmetics: few typofixes in older ChangeLogs, suggested by "codespell.py"
Stefano Lattarini [Sat, 9 Jun 2012 18:04:24 +0000 (20:04 +0200)]
cosmetics: few typofixes in older ChangeLogs, suggested by "codespell.py"

The typofixes offered by this patch have been suggested by the
"codespell.py" script.  Reference:
<http://git.profusion.mobi/cgit.cgi/lucas/codespell/>

* old/ChangeLog.01, old/ChangeLog.02, old/ChangeLog.03: Fix few typos.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agosilent: new $(AM_V_P) variable, tell if we're running in silent mode
Stefano Lattarini [Mon, 11 Jun 2012 14:12:01 +0000 (16:12 +0200)]
silent: new $(AM_V_P) variable, tell if we're running in silent mode

Addresses part of automake bug#8665.

* automake.in (handle_silent): Define a new make variable '$(AM_V_P)',
that expands to a shell conditional that can be used in make recipes to
determine whether they are being run in silent mode or not.  The choice
of the name derives from the LISP convention of appending the letter
'P' to denote a predicate (see also "the '-P' convention" in the Jargon
File); we do so for lack of a better convention.
* t/automake.texi, NEWS: Document the new variable.
* t/silent6.sh: Adjust and extend.  Move out the checks that didn't
actually deal with user extension of silent rules ...
* t/silent-obsolescent-warns.sh: ... into this test (bound to be
removed once 'maint' is merged into the 'master' branch).
* t/list-of-tests.mk: Add the new test.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agorefactor: silent rules handling (a little)
Stefano Lattarini [Mon, 11 Jun 2012 11:00:23 +0000 (13:00 +0200)]
refactor: silent rules handling (a little)

* automake.in (handle_languages): Move definition of $(AM_V_GEN)
variable ...
(define_verbose_tagvar): ... and of '$(AM_V_at)' variable ...
(handle_silent): ... in this new subroutine.
(generate_makefile): Call it.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agorefactor: &define_verbose_var: accept a third optional argument
Stefano Lattarini [Mon, 11 Jun 2012 10:48:29 +0000 (12:48 +0200)]
refactor: &define_verbose_var: accept a third optional argument

This is only required by future changes.

* automake.in (define_verbose_var): Accept a third optional argument,
specifying the value to assign to the given make variable when silent
rules are disabled.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agomaint: grammar fixes: s/all these/all of these/
Jim Meyering [Sun, 10 Jun 2012 20:03:49 +0000 (22:03 +0200)]
maint: grammar fixes: s/all these/all of these/

Run this command:
  git grep -li '\<all.these\>' \
    |xargs perl -pi -e 's/\b([Aa])ll these\b/${1}ll of these/'

11 years agosubdir-objects: improve "make mostlyclean" efficiency and flexibility
Stefano Lattarini [Sun, 10 Jun 2012 11:38:58 +0000 (13:38 +0200)]
subdir-objects: improve "make mostlyclean" efficiency and flexibility

Fixes automake bug#10697.

Before this change, the generated Makefile issued one 'rm' invocation
for each subdir object file.  Not only was this very inefficient when
there were several such files, but it also caused stale object files
to be left behind when a source file was renamed or removed.

* automake.in (handle_single_transform): When a subdir object is seen,
update '%compile_clean_files' to clean all the compiled objects in its
same subdirectory, and all the libtool compiled objects ('.lo') there
as well is that subdir object is a libtool one.
* t/subobj-clean-pr10697.sh: New test.
* t/subobj-clean-lt-pr10697.sh: Likewise.
* t/list-of-tests.mk: Add them.
* NEWS: Update.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoMerge branch 'maint'
Stefano Lattarini [Sat, 9 Jun 2012 18:15:18 +0000 (20:15 +0200)]
Merge branch 'maint'

* maint:
  tests: look for '.lo' rather than '.o' object when using Libtool with C++
  cosmetics: few typofixes in older ChangeLogs, suggested by "codespell.py"
  tests: add basic semantic tests on C++ support
  tests: minor tweak to 't/objc-megademo.sh'

11 years agotests: look for '.lo' rather than '.o' object when using Libtool with C++
Stefano Lattarini [Sat, 9 Jun 2012 18:13:43 +0000 (20:13 +0200)]
tests: look for '.lo' rather than '.o' object when using Libtool with C++

* t/cxx-lt-demo.sh: ... in this test.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agocosmetics: few typofixes in older ChangeLogs, suggested by "codespell.py"
Stefano Lattarini [Sat, 9 Jun 2012 18:04:24 +0000 (20:04 +0200)]
cosmetics: few typofixes in older ChangeLogs, suggested by "codespell.py"

The typofixes offered by this patch have been suggested by the
"codespell.py" script.  Reference:
<http://git.profusion.mobi/cgit.cgi/lucas/codespell/>

* old/ChangeLog.01, old/ChangeLog.02, old/ChangeLog.03: Fix few typos.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: add basic semantic tests on C++ support
Stefano Lattarini [Sat, 9 Jun 2012 13:11:10 +0000 (15:11 +0200)]
tests: add basic semantic tests on C++ support

Strange as it might seem, we were still missing them (yikes).

* t/cxx-demo.sh: New test.
* t/cxx-lt-demo.sh: Likewise (using libtool).
* t/list-of-tests.mk: Add them.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: minor tweak to 't/objc-megademo.sh'
Stefano Lattarini [Sat, 9 Jun 2012 11:50:12 +0000 (13:50 +0200)]
tests: minor tweak to 't/objc-megademo.sh'

* t/objc-megademo.sh (am_create_testdir): Define to "empty" before
including ./defs, because this test doesn't rely on the files usually
pre-set by the setup in there.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agocleanup: simplify subroutine '&saw_extension'
Stefano Lattarini [Thu, 7 Jun 2012 17:14:30 +0000 (19:14 +0200)]
cleanup: simplify subroutine '&saw_extension'

* automake.in (saw_extension): Now that we use the '%extension_seen'
hash only as a set (it's keys being its elements), we can simplify
the implementation of this function accordingly.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agocleanup: remove subroutine '&saw_sources_p'
Stefano Lattarini [Thu, 7 Jun 2012 17:14:12 +0000 (19:14 +0200)]
cleanup: remove subroutine '&saw_sources_p'

* automake.in (saw_sources_p): Remove, its implementation is now so
trivial that it's easier to inline it into ...
(handle_languages): ... it's only caller.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agocleanup: remove unused '&count_files_for_language' subroutine
Stefano Lattarini [Thu, 7 Jun 2012 17:13:39 +0000 (19:13 +0200)]
cleanup: remove unused '&count_files_for_language' subroutine

* automake.in (count_files_for_language): Remove.  Its only remaining
caller (saw_sources_p) has stopped calling it in the previous commit.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoautomake: simplify '&saw_sources_p'
Stefano Lattarini [Thu, 7 Jun 2012 17:12:43 +0000 (19:12 +0200)]
automake: simplify '&saw_sources_p'

This is just a minor cleanup.  No semantic change is intended.

* automake.in (saw_sources_p): Since its now-only caller calls it with
the '0' argument, remove the code paths that assumed the argument could
be '1', and assume no arguments.
(handle_languages): Drop the arguments '0' in the '&saw_sources_p'
invocation.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotags: look at $(am__tagged_files) to decide whether to activate
Stefano Lattarini [Thu, 7 Jun 2012 17:12:11 +0000 (19:12 +0200)]
tags: look at $(am__tagged_files) to decide whether to activate

This is just a preparatory refactoring in view of future changes.

* automake.in (handle_tags): To decide whether to include the tags rules,
rely on whether the automake-defined variable $(SOURCES) is non-empty,
rather than on the fact that '&saw_sources(1)' returns true.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotags: new var $(am__tagged_files) to list all files to run taggers on
Stefano Lattarini [Thu, 7 Jun 2012 18:13:37 +0000 (20:13 +0200)]
tags: new var $(am__tagged_files) to list all files to run taggers on

This change reduce code duplication a little, and will be needed by
future simplifications.

* automake.in (handle_tags): Define a new 'am__tagged_files' private
make variable that lists all files to run taggers (mkid, cscope, ctags,
etc) on.  Drop the now-useless transform '%CONFIG%' when including
'tags.am'.
* lib/am/tags.am: Use it in several recipes and dependencies list to
reduce code duplication.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotags (cscope): also process $(TAGS_FILES)
Stefano Lattarini [Thu, 7 Jun 2012 17:41:37 +0000 (19:41 +0200)]
tags (cscope): also process $(TAGS_FILES)

* lib/am/tags.am (cscopelist): Also scan the files (if any) listed in
the '$(TAGS_FILES) variable, for consistency by what is done by the
'ID', 'TAGS' and 'CTAGS' rules.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotags (ID, cscope): also process config header (if any)
Stefano Lattarini [Thu, 7 Jun 2012 17:19:15 +0000 (19:19 +0200)]
tags (ID, cscope): also process config header (if any)

* lib/am/tags.am (cscopelist, ID): Also scan the file given by the
'%CONFIG%' transform, for consistency by what is done by the 'TAGS'
and 'CTAGS' rules.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agoMerge branch 'maint'
Stefano Lattarini [Sat, 2 Jun 2012 07:54:17 +0000 (09:54 +0200)]
Merge branch 'maint'

* maint:
  news: document deprecation of configure.in
  news: put planned backward compatibilities early
  sync: update files from upstream with "make fetch"
  maint: deprecate 'configure.in' as autoconf input
  maint: post-release minor version bump
  release: stable release 1.12.1
  copyright: update copyright years in a couple of files
  tests: avoid spurious failures with Solaris 9 cscope program
  maintcheck: fix a spurious failure
  perf: beginning of a performance testsuite
  aclocal: declare function prototypes, do not use '&' in function calls
  news: support for configure.in will be dropped in future automake versions
  tests: fix botched heading comments in 'lex-clean-cxx.sh'
  [ng] maintcheck: some tweaks and fixlets
  help: fix a typo in the list of warning categories
  maint: version bump after beta release
  release: beta release 1.12.0b (will become 1.12.1)
  release: remove overly picky check

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agonews: document deprecation of configure.in
Stefano Lattarini [Sat, 2 Jun 2012 07:34:29 +0000 (09:34 +0200)]
news: document deprecation of configure.in

* NEWS (Warnings and deprecations): We now warn if 'configure.in' is
used instead of 'configure.ac' as autoconf input.
(Future backward-incompatibilities): Adjust.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agonews: put planned backward compatibilities early
Stefano Lattarini [Sat, 2 Jun 2012 07:24:29 +0000 (09:24 +0200)]
news: put planned backward compatibilities early

* NEWS (Future backward-incompatibilities): Put them in the news
entries for the future 1.12.2 release.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agosync: update files from upstream with "make fetch"
Stefano Lattarini [Fri, 1 Jun 2012 18:24:20 +0000 (20:24 +0200)]
sync: update files from upstream with "make fetch"

* lib/config.sub, lib/gitlog-to-changelog, lib/texinfo.tex: Update.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agomaint: deprecate 'configure.in' as autoconf input
Stefano Lattarini [Fri, 1 Jun 2012 17:15:29 +0000 (19:15 +0200)]
maint: deprecate 'configure.in' as autoconf input

It has been years since that has been deprecated in the documentation,
in favour of 'configure.ac':

    Previous versions of Autoconf promoted the name configure.in, which
    is somewhat ambiguous (the tool needed to process this file is not
    described by its extension), and introduces a slight confusion with
    config.h.in and so on (for which '.in' means "to be processed by
    configure"). Using configure.ac is now preferred.

It's now time to start giving runtime warning about the use of
'configure.in', so that support for it can be removed in future
versions of autoconf/automake.

See also, in the Autoconf repository, commit 'v2.69-4-g560f16b' of
2012-05-23, "general: deprecate 'configure.in' as autoconf input".

* lib/Automake/Configure_ac.pm: Issue a warning in the 'obsolete'
category if 'configure.in' is detected.  Since this module is synced
from Automake, this change is to be backported there (and will be
soon).
* t/help.sh: Adjust.
* t/configure.sh: Adjust and enhance.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agomaint: post-release minor version bump
Stefano Lattarini [Fri, 1 Jun 2012 16:55:14 +0000 (18:55 +0200)]
maint: post-release minor version bump

* configure.ac (AC_INIT): Bump version number to 1.12.1a.
* m4/amversion.m4: Likewise (automatically regenerated by
"make bootstrap").

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agorelease: stable release 1.12.1 v1.12.1
Stefano Lattarini [Fri, 1 Jun 2012 13:50:03 +0000 (15:50 +0200)]
release: stable release 1.12.1

* configure.ac (AC_INIT): Bump version number to 1.12.1.
* m4/amversion.m4: Likewise (auto-updated by "./bootstrap").

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agocopyright: update copyright years in a couple of files
Stefano Lattarini [Fri, 1 Jun 2012 13:47:17 +0000 (15:47 +0200)]
copyright: update copyright years in a couple of files

* doc/help2man, t/aclocal-verbose-install.sh: Update copyright,
thanks to "make update-copyright".

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agotests: avoid spurious failures with Solaris 9 cscope program
Stefano Lattarini [Fri, 1 Jun 2012 10:37:41 +0000 (12:37 +0200)]
tests: avoid spurious failures with Solaris 9 cscope program

* t/cscope.tap: Also check that the cscope program in use supports the
'-q' option; if this is not the case, some cscope usages in the generated
Makefiles would fail, so we must skip the checks involving those usages.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
11 years agomaintcheck: fix a spurious failure
Stefano Lattarini [Fri, 1 Jun 2012 09:49:16 +0000 (11:49 +0200)]
maintcheck: fix a spurious failure

* Makefile.am (test_subdirs): Add 't/perf'.  Fix spurious failure
of the 'maintainer-check-list-of-tests' target.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoperf: beginning of a performance testsuite
Stefano Lattarini [Tue, 29 May 2012 09:58:02 +0000 (11:58 +0200)]
perf: beginning of a performance testsuite

Some tests in the Automake testsuite already aims only at verifying
the performance, rather than the correctness, of some operations.
Still, they are somewhat shoehorned and forced into the PASS/FAIL
framework (say, with the 'ulimit' shell builtin used to verify some
operation doesn't take up too much time or memory), but that is
conceptually a stretch, and has already caused problems in practice
(see automake bug#11512 for an example).

So we start moving the "performance tests" out of the testsuite proper,
and make them run only "on demand" (when the user exports the variable
'AM_TESTSUITE_PERF' to "yes").  Ideally, we should provide those tests
with a custom runner/driver that measures and displays the relevant
performance information, but doing that correctly and with the right
APIs is definitely more difficult, so we leave it for a later step
(an hope we'll take such a step eventually).

* t/cond29.sh: Move ...
* t/perf/cond.sh: ... here, and adjust.
* t/testsuite-recheck-speed.sh: Move ...
* t/perf/testsuite-recheck.sh: ... here.
* t/testsuite-summary-speed.sh: Move ...
* t/perf/testsuite-summary.sh: ... here.
* t/list-of-tests.mk (perf_TESTS): New variable, listing the tests in
the 't/perf' directory.
(handwritten_TESTS): Adjust.
* defs: Skip any tests in the 't/perf/' subdirectory unless the
'AM_TESTSUITE_PERF' variable is set to "yes" or "y".
* .gitignore: Update.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agoMerge branch 'fix-pr11543' into maint
Stefano Lattarini [Tue, 29 May 2012 07:36:26 +0000 (09:36 +0200)]
Merge branch 'fix-pr11543' into maint

* fix-pr11543:
  aclocal: declare function prototypes, do not use '&' in function calls

12 years agoaclocal: declare function prototypes, do not use '&' in function calls
Stefano Lattarini [Mon, 28 May 2012 11:32:03 +0000 (13:32 +0200)]
aclocal: declare function prototypes, do not use '&' in function calls

This change will also fix automake bug#11543 (from a report by Matt
Burgess).

* aclocal.in: Declare prototypes for almost all functions early, before
any actual function definition (but omit the prototype for the dynamically
generated '&search' function).  Add prototypes to any function definition.
Remove '&' from function invocations (i.e., simply use "func(ARGS..)"
instead of "&func(ARGS...)").
* THANKS, NEWS: Update.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agonews: support for configure.in will be dropped in future automake versions
Stefano Lattarini [Mon, 28 May 2012 11:52:34 +0000 (13:52 +0200)]
news: support for configure.in will be dropped in future automake versions

The use of configure.in as Autoconf input has been deprecated for a
very long time in the Autoconf documentation, and the next version of
Autoconf (2.70) will start warning about it ar runtime as well (see
commit 'v2.69-4-g560f16b' or 2012-05-23, "general: deprecate
'configure.in' as autoconf input", in the Autoconf's git repository).

* NEWS (Future backward-incompatibilities): Update.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agomissing: support AC_CONFIG_HEADERS in addition to obsolete AC_CONFIG_HEADER
Stefano Lattarini [Sun, 27 May 2012 17:33:26 +0000 (19:33 +0200)]
missing: support AC_CONFIG_HEADERS in addition to obsolete AC_CONFIG_HEADER

* missing (autoheader*): Do not look only for AC_CONFIG_HEADER when
grepping configure.ac, but also for AC_CONFIG_HEADERS: the latter has
been preferred to the former by several years now in Autoconf.

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agomissing: don't bother checking for AM_CONFIG_HEADER: it is gone now
Stefano Lattarini [Sun, 27 May 2012 17:05:14 +0000 (19:05 +0200)]
missing: don't bother checking for AM_CONFIG_HEADER: it is gone now

* missing (autoheader*): Do not look for AM_CONFIG_HEADER when
grepping configure.ac: it is not supported anymore (has been in
fact deleted in commit v1.12-15-gd2ca168 of 2012-04-27, "m4: delete
several obsolete macros").

Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
12 years agodocs: remove mention of AM_CONFIG_HEADER: it is obsolete and removed now
Stefano Lattarini [Sun, 27 May 2012 16:47:46 +0000 (18:47 +0200)]
docs: remove mention of AM_CONFIG_HEADER: it is obsolete and removed now

* doc/automake.texi (Variables used when building a program): Remove
obsolete mention of AM_CONFIG_HEADER as an alias to AC_CONFIG_HEADERS.

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