* t/remake-renamed-m4-macro-and-file.sh: Adjust to reflect to old
"acloca22 -> t/aclocal-deleted-header.sh" test rename.
* t/aclocal-pr450.sh (configure.ac): Use '$me' in the AC_INIT call,
instead of hard-coding the old name of this test, i.e., "acloca19".
tests: fix an old botched change to an aclocal test
* t/acloca10.sh (configure.ac): Here, invoke the m4 macro 'MACRO2'
before the macro 'MACRO1' (the related test 't/aclocal-I-order-2.sh'
does the opposite). This reverts a botched edit done (by myself,
oops) in commit 'v1.11-1335-gefdc3e1' of 2011-09-11, "tests: minor
optimizations/simplifications in some aclocal tests", and makes the
behaviour of the test match once again what is stated in the
heading comments. While at it, improve those same heading comments
a little.
coverage: compile rules used "-c -o" also with losing compilers
If the 'subdir-objects' option is used, Automake-generated rules for
C compilation pass both the "-c" and "-o" options to the C compiler,
*unconditionally*. There are some compilers that choke on such an
usage, but the AM_PROG_CC_C_O macro takes care of them (it does so by
redefining $CC to use the Automake-provided 'compile' wrapper script
automatically, if a losing compiler is detected at configure runtime).
Unfortunately, in case the 'subdir-objects' option is specified in a
Makefile.am, but all the source files resided anyway in the top-level
directory (relative to the Makefile.am), Automake do *not* complain
if AM_PROG_CC_C_O wasn't invoked in 'configure.ac' -- all the while
still passing "-c -o" to the compiler invocations. This could cause
compilation failures with losing compilers if the user forget to call
AM_PROG_CC_C_O in 'configure.ac' (and Automake would not warn him of
the issue).
Expose this bug in the testsuite.
Issue identified by Nick Bowler in the discussion on automake bug#13378:
<http://debbugs.gnu.org/cgi/bugreport.cgi?bug=13378#35>
<http://debbugs.gnu.org/cgi/bugreport.cgi?bug=13378#44>
* t/ccnoco4.sh: New test.
* t/list-of-tests.mk (XFAIL_TESTS, handwritten_TESTS): List it.
coverage: user can avoid distributing '.info' pages
Can be done like this:
AUTOMAKE_OPTIONS = info-in-builddir
dist-info:
@:
Note that this usage is not yet documented: we might decide to go
for a fully-fledged 'no-dist-info' flag, or something like that, in
future automake version (this is not yet decided); in which case,
it's better not to have people start to rely on the hack above.
Still, there's no good reason to break it gratuitously, hence this
test coverage.
* t/txinfo-nodist-info.sh: New test.
* t/list-of-tests.mk: Add it.
build: enable all warnings as fatal in our own build system
Automake should of course be able to bootstrap itself in a
warning-free manner w.r.t. the Autotools. So make any failure
to do so fatal. Not doing so caused the regression fixed by
previous commit 'v1.13.1-22-ga790fae' to go unnoticed.
* configure.ac (AM_INIT_AUTOMAKE): Add '-Werror' and '-Wall'.
* bootstrap.sh: Pass the '-Wall -Werror' options to aclocal,
automake and autoconf invocations.
texi: Texinfo sources and CLEANFILES definition should co-exist peacefully
But they don't now, due to a regression introduced in commit
'v1.13.1-4-gc1a8f56'. Fix it. The regression was hitting our
own build system!
* automake.in (handle_texinfo_helper): Only complain if the
'info-in-builddir' is not active and a '.info' file (not any
random file!) is listed in CLEANFILES or DISTCLEANFILES.
* t/ax/runtest.in: Accept options '-k' and '--keep-testdirs' (same
as exporting '$keep_testdirs' to "yes"). To improve compatibility
with the "make check" interface, allow environment variables to be
passes on the command line. Minor adjustments while at it.
Individual files or sub-directories about future and on-going
development plans in Automake will be added in follow-up commits.
This new set of documents is meant to help ensure a more controlled
and smooth development and evolution for Automake, in several ways.
- Having the plans clearly spelled out should will avoid messy
roadmaps with no clear way forward or with muddy or ill-defined
aims or purposes; a trap this is too easy to fall into.
- Keeping planned changes cooking and re-hashed for a while should
ensure rough edges are smoothed up, transitions are planned in a
proper way (hopefully avoiding debacles like the AM_MKDIR_PROG_P
deprecation and the AM_CONFIG_HEADER too-abrupt removal), and
"power users" have more chances of getting informed in due time,
thus having all the time to prepare for the changes or raise
objections against them.
- Having the plans clearly stated and registered in a "centralized"
location should make it more difficult to them to slip through
the cracks, getting forgotten or (worse) only half-implemented.
- Even for discussions and plans registered on the Bug Tracker
as well, a corresponding entry in the PLANS directory can help
in keeping main ideas summarized, and consensus and/or objections
registered and easily compared.
Motivation:
<http://blog.flameeyes.eu/2013/01/autotools-mythbuster-automake-pains>
Not a flatting picture for us (and maybe a little too harsh), but
basically true and even spot-on in some regards.
They are likely not general enough for widespread use, but they
are useful nonetheless.
In the best-case scenario, they will start to be used by other
people, and thus accordingly improved and made more general and
flexible.
In the worst case scenario, well, I still get to keep them in a
centralized, blessed place, simplifying the deployment and use
of them; so still a win for me :-)
tests: tweak tests on obsolete EXTRA_DATA variable
* t/extra3.sh, t/extra4.sh: Merge ....
* t/extra-data.sh: ... into this, with updated comments.
* t/list-of-tests.mk: Adjust. Also Tweak the order in
which some other tests are listed.
docs: re-introduce mention of two-args AM_INIT_AUTOMAKE invocation
Albeit obsolescent and raising warnings in the 'obsolete' category,
that usage is still supported, and will need to be until Autoconf
improves its handling of configure-time-generated package version
numbers. So it's better to explicitly document it again, stating
that it is obsoleted but still working (and why), rather then leaving
it as Yet Another Undocumented Feature (that will mysteriously and
suddenly break some random day in the future).
It's worth giving some background about how we ended up in the
situation that this patch fixes.
We had originally removed support for the long-deprecated two-args
AM_INIT_AUTOMAKE invocation (see commit v1.12-67-ge186355). Before
that removal could land in a released Automake version, Bob Friesenhahn
made a quite compelling point that the two-args AM_INIT_AUTOMAKE
invocation could still be useful for modern, maintained packages like
GraphicsMagick, at least until Autoconf is fixed to offer better support
for "dynamic" package versions (see commit v1.12.2-245-g2abe183 for more
in-depth rationales and references). However, in that commit we didn't
revert the removal of the *documentation* for this two-arguments
AM_INIT_AUTOMAKE invocation (and no rationale for not doing so was given
in the commit message). Time to remedy that.
Indirectly suggested by Diego Elio Pattenò:
<http://blog.flameeyes.eu/2013/01/autotools-mythbuster-automake-pains>
texi: warn against '.txi' and '.texinfo' input suffixes
The warning being in the 'obsolete' category. This is mostly to
ease transition to Automake-NG (see commit v1.12.1-416-gd5459b9),
and to discourage use of seldom-tested setups.
* automake.in (handle_texinfo_helper): Warn against Texinfo input
files with '.txi' or '.texinfo' suffixes.
* NEWS: Update.
* t/txinfo-other-suffixes.sh: Adjust and enhance.
* t/txinfo-no-repeated-targets.sh: No longer use '.txi' and
'.texinfo' extensions.
The warning being in the 'obsolete' category. This is mostly to
ease transition to Automake-NG (see commit v1.12.1-392-ga0c7b6a),
and to discourage use of seldom-tested setups.
* automake.in (scan_texinfo_file): Warn against '@setfilename'
directives that specify suffix-less output info files.
* t/txinfo-without-info-suffix.sh: Adjust and enhance.
* t/txinfo-makeinfo-error-no-clobber.sh: No longer use suffix-less
info files in '@setfilename' directives.
* t/primary-prefix-valid-couples.sh: Likewise.
* t/txinfo-setfilename-repeated.sh: Likewise.
* t/txinfo-vtexi2.sh : Likewise.
* t/mdate2.sh: Likewise.
* NEWS: Update.
texi: deprecate hack about info files in CLEANFILES variables
For quite a long time, Automake has been implementing an undocumented
hack which ensured that '.info' files which appeared to be cleaned (by
e.g. being listed in the CLEANFILES or DISTCLEANFILES variables) were
built in the builddir rather than in the srcdir; this hack was introduced
to ensure better backward-compatibility with packages such as Texinfo,
which did things like:
info_TEXINFOS = texinfo.txi info-stnd.texi info.texi
DISTCLEANFILES = texinfo texinfo-* info*.info*
# Do not create info files for distribution.
dist-info:
@:
in order not to distribute .info files.
Now that we have the 'info-in-builddir' option that explicitly causes
generated '.info' files to be placed in the builddir, this hack should
be longer necessary, so we deprecate it with runtime warnings. It is
scheduled to be removed altogether in Automake 1.14.
* automake.in (handle_texinfo_helper): Raise proper runtime warnings
if the hack is triggered.
* NEWS: Update.
* t/txinfo28.sh: Adjust.
* t/txinfo23.sh: Likewise.
* t/txinfo25.sh: Adjust and extend.
* t/txinfo24.sh: Likewise.
User can now ask info files to be built in the $(builddir), rather than
the $(srcdir), by specifying the Automake option 'info-in-builddir'.
This feature was requested by the developers of GCC, GDB, GNU binutils
and the GNU bfd library. See the extensive discussion about automake
bug#11034 for more details.
OK, to be honest, having '.info' files built in the builddir was
*already* possible, but only using ugly and undocumented hacks involving
definition of the CLEANFILES and/or DISTCLEANFILES. For example, the
binutils project did something like this in the relevant 'Makefile.am':
# Automake 1.9 will only build info files in the objdir if they are
# mentioned in DISTCLEANFILES. It doesn't have to be unconditional,
# though, so we use a bogus condition.
if GENINSRC_NEVER
DISTCLEANFILES = binutils.info
endif
See also the extensive discussion about automake bug#11034; in
particular, the following messages:
<http://debbugs.gnu.org/cgi/bugreport.cgi?bug=11034#65>
<http://debbugs.gnu.org/cgi/bugreport.cgi?bug=11034#80>
<http://debbugs.gnu.org/cgi/bugreport.cgi?bug=11034#86>
<http://debbugs.gnu.org/cgi/bugreport.cgi?bug=11034#101>
* lib/Automake/Options.pm (_is_valid_easy_option): Recognize the
new 'info-in-builddir' option.
* automake.in (handle_texinfo_helper): If that option is set,
initialize '$insrc' to '0', so that info files will be generated
in the builddir. Adjust comments to match.
* t/txinfo-builddir.sh: New test.
* t/list-of-tests.mk: Add it.
* NEWS: Update.
* doc/automake.texi: Document the new options.
tests: don't always look for a C++ compiler named 'RCC'
On MacOS X (10.8), since the file system is case-insensitive, RCC
can point to the "Resource Compiler" of the Qt4 Toolkit:
<http://doc.qt.digia.com/4.2/rcc.html>
That mismatch causes our configure script to erroneously think that
no working C++ compiler is present, and that is thus necessary to
skip all the test cases requiring such a compiler.
So only look for a compiler named 'RCC' if the file system is
case-sensible.
Issue spotted analyzing the testsuite logs reported in bug#13317.
tests: fix bug in pkg-config-macros.sh, could cause spurious SKIPs
Issue spotted perusing the testsuite logs reported in bug#13317.
* t/pkg-config-macros.sh: Don't use (uninitialized) '$dir' where '$d'
should have been used instead. Set IFS to ':' before looping on the
$PATH expansion. Fix typo: 'alocal' instead of 'aclocal'. These
issues were causing the location in PATH of the 'pkg-config' program
not to be found even when the program was present.
* THANKS: Update.
docs: autom4te cache can break CONFIGURE_DEPENDENCIES
In fact, the new test 'remake-configure-dependencies.sh' only
works because it disables the autom4te cache :-( That is more
of an AUtotools design issue than an Automake bug, so better
just document it ...
tests: some enhancements to texinfo tests (from Automake-NG)
* t/txinfo-many-output-formats.sh: Backport improvements, almost verbatim,
from Automake-NG commit 'v1.13-768-gb434acc' (plus minor tweaks for the
sake of non-GNU make implementation).
* t/txinfo-many-output-formats-vpath.sh: New test, backported almost
verbatim from that same Automake-NG commit (again, with minor tweaks for
the sake of non-GNU make implementations).
* t/lsit-of-tests.mk: Adjust.
* t/txinfo-subdir-pr343.sh: Move the tests checking that '.info.bak'
files in subdirs are not unduly distributed into ...
* t/txinfo-no-extra-dist.sh: ... in here. Enhance a little while
at it.
* t/txinfo-other-suffixes.sh: Also check that the '.texinfo' suffix
is accepted and works.
* t/txinfo-setfilename-suffix-match.sh: Remove useless call to
autoconf, enhance a little.
* t/txinfo-add-missing.sh: This one, its checks being already covered
by the more extensive 't/add-missing.tap' test.
* t/txinfo-add-missing2.sh: Rename ...
* t/txinfo-add-missing-and-dist.sh: Like this.
* t/list-of-tests.mk: Adjust.
Do so for several Texinfo-related tests. Note that some of such tests
with suboptimal names are *not* renamed; that's because they are going
to be removed in the master branch anyway (since they check from some
soon-to-be-removed features), so renaming them would be wasted work,
and could even create useless merge conflicts.
* t/txinfo2.sh: Rename ...
* t/txinfo-no-extra-dist.sh: ... like this.
* t/txinfo3.sh: Rename ...
* t/txinfo-suffix-less-info.sh: ... like this.
* t/txinfo4.sh: Rename ...
* t/txinfo-unrecognized-info-suffix.sh: ... like this.
* t/txinfo6.sh: Rename ...
* t/txinfo-other-suffixes.sh: ... like this.
* t/txinfo7.sh: Rename ...
* t/txinfo-add-missing.sh: ... like this.
* t/txinfo8.sh: Rename ...
* t/txinfo-add-missing2.sh: ... like this.
* t/txinfo9.sh: Rename ...
* t/txinfo-no-repeated-targets.sh: ... like this.
* t/txinfo17.sh: Rename ...
* t/txinfo-setfilename-repeated.sh: ... like this.
* t/txinfo13.sh: Rename ...
* t/txinfo-subdir-pr343.sh: ... like this.
* t/txinfo16.sh: Rename ...
* t/txinfo-info-in-srcdir.sh: ... like this.
* t/txinfo21.sh: Rename ...
* t/txinfo-many-output-formats.sh: ... like this.
* t/txinfo22.sh: Rename ...
* t/txinfo-override-texinfo-tex.sh: ... like this.
* t/txinfo26.sh: Rename ...
* t/txinfo-absolute-srcdir-pr408.sh: ... like this.
* t/txinfo27.sh: Rename ...
* t/txinfo-no-installinfo.sh: ... like this.
* t/txinfo29.sh: Rename ...
* t/txinfo-override-infodeps.sh: ... like this.
* t/txinfo31.sh: Rename ...
* t/txinfo-setfilename-suffix-match.sh: ... like this.
* t/txinfo32.sh: Rename ...
* t/txinfo-bsd-make-recurs.sh: ... like this.
* t/txinfo33.sh: Rename ...
* t/txinfo-clean.sh: ... like this.
* t/list-of-tests.mk: Adjust.
obsolete: better error message if AM_PROG_CC_STDC is used
In Automake 1.13, the long-deprecated (since 2002) macro AM_PROG_CC_STDC
has been removed. Such a removal, albeit sensible and justified, was
probably done in a too-abrupt way, since it didn't turn the pre-existing
warning messages into fatal error messages, but simply dropped the macro
definition, so that remaining usages of it would cause unclear error
messages, e.g.:
configure.ac:4: warning: macro 'AM_PROG_CC_STDC' not found in library
from aclocal, and:
configure.ac:4: error: possibly undefined macro: AM_PROG_CC_STDC
If this token and others are legitimate, please use m4_pattern_allow.
See the Autoconf documentation.
from autoconf.
In an attempt to mitigate this issue, we re-add an AM_PROG_CC_STDC
definition that simply raises a *clear* error message when the macro
is used.
* t/am-prog-cc-stdc-no-more.sh: New test.
* t/list-of-tests.mk: Add it.
* m4/obsolete-err.m4: Add the new "error-raising" definition for
AM_PROG_CC_STDC; the error message is a variation of the one already
present in the older version of this macro, before it got removed in
commit 'v1.12-15-gd2ca168'.
* NEWS: Update.
obsolete: better error message if AM_CONFIG_HEADER is used
In Automake 1.13, the long-deprecated macro AM_CONFIG_HEADER (deprecated
since 2002) has been removed in favour of AC_CONFIG_HEADERS. However,
the removal was done without a proper deprecation period, and that
caused packages upgrading to Automake 1.13 to fail with very unclear
error messages, e.g.:
configure.ac:4: warning: macro 'AM_CONFIG_HEADER' not found in library
from aclocal, and:
configure.ac:4: error: possibly undefined macro: AM_CONFIG_HEADER
If this token and others are legitimate, please use m4_pattern_allow.
See the Autoconf documentation.
from autoconf.
In an attempt to mitigate this issue, we re-add an AM_CONFIG_HEADER
definition that simply raises a *clear* error message when the macro
is used.
Report by Paolo Bonzini:
<http://lists.gnu.org/archive/html/automake/2012-12/msg00039.html>
* t/am-config-header-no-more.sh: New test.
* t/list-of-tests.mk: Add it.
* m4/obsolete-err.m4: New file, contain the new AM_CONFIG_HEADER
"error-raising" definition, as well as the definition of the
similarly obsolete macros 'AM_C_PROTOTYPES' and 'fp_C_PROTOTYPES',
moved in from ...
* m4/protos.m4: ... this file, which has thus been removed.
* Makefile.am (dist_automake_ac_DATA): Adjust.
* t/ansi2knr-no-more.sh: Likewise.
* NEWS: Update.
tests: better names for temporary configure.ac files
When editing a pre-existent 'configure.ac' file in a test script, name
the temporary file 'configure.tmp' rather than 'configure.int'; the
latter was a relict of the times when we used 'configure.in' as the
default name for the Autoconf input files throughout the testsuite.
* t/cond43.sh: Adjust.
* t/depcomp8a.sh: Likewise.
* t/depcomp8b.sh: Likewise.
* t/gettext.sh: Likewise.
* t/mmode.sh: Likewise.
* t/pr401.sh: Likewise.
* t/pr401b.sh: Likewise.
* t/pr401c.sh: Likewise.
* t/python11.sh: Likewise.
* t/cond42.sh: Likewise. Also, while at it, rename ...
(edit_configure_in): ... this function ...
(edit_configure_ac): ... like this.
Akim Demaille [Wed, 19 Dec 2012 13:51:58 +0000 (14:51 +0100)]
ylwrap: various fixes
Rename properly header guards in generated header files, instead of
leaving Y_TAB_H.
Convert header guards in implementation files. Because ylwrap failed
to rename properly #include in the implementation files, current
versions of Bison (e.g., 2.7) duplicate the generated header file in
the implementation file. The header guard then protects the
implementation file from duplicate definitions from the header file.
Generate header guards with a single '_' for series of non alphabetic
characters, instead of several. This is what Bison does.
Makes the test t/yacc-d-basic.sh pass again.
* lib/ylwrap (guard): Properly honor $1 to rename properly the
header guards.
Keep a single _ instead of several.
(rename_sed): Rename as...
(sed_fix_filenames): this.
Suggested by Stefano Lattarini.
(sed_fix_header_guards): New.
Use it.
* t/tags-pr12372.sh (configure.ac): AC_SUBST the LINK variable to a
dummy invocation, to avoid possible errors from make or the linker;
errors we do not care about in the least in this test.
(Makefile.am, sub/Makefile.am): Remove LINK definitions; simply
inherit that in configure.ac.
tests: re-enable some checks disabled by mistake ...
* t/subpkg-macrodir.sh: ... in this test. And fix some typos that would
have caused the newly re-enabled check to spuriously fail. Issue revealed
by the lack of '.PHONY' support in Solaris CCS make (yes, so we've been
actually *helped* by that make implementation once, at last. Scary ;-)
tests: avoid a spurious failure due to a Clang bug
This version of clang:
clang version 3.2 (trunk 163574)
Target: powerpc64-unknown-linux-gnu
Thread model: posix
caused the test 't/ltcond2.sh' to spuriously fail due to what
appeared like a clang bug. Here is a part of the diagnostic (trimmed
down for better clarity):
clang: .../cfarm/llvm/lib/MC/MCAsmStreamer.cpp:338: \
virtual void {anonymous}::MCAsmStreamer::EmitLabel(llvm::MCSymbol*): \
Assertion `Symbol->isUndefined() && "Cannot define a symbol twice!"' \
failed.
...
7 clang 0x0000000012a459c4 llvm::AsmPrinter::EmitGlobalVariable\
(llvm::GlobalVariable const*) + 18446744073680468044
8 clang 0x0000000012a490a8 llvm::AsmPrinter::doFinalization\
(llvm::Module&) + 18446744073680481840
...
Stack dump:
0. Program arguments: .../opt/cfarm/clang-2012.09.10/bin/clang \
-cc1 -triple powerpc64-unknown-linux-gnu -S -disable-free \
...
1. <eof> parser at end of file
2. Code generation
3. Running pass 'Function Pass Manager' on module 'hello-generic.c'.
clang: error: unable to execute command: Aborted
clang: error: clang frontend command failed due to signal (use -v to \
see invocation)
clang version 3.2 (trunk 163574)
Target: powerpc64-unknown-linux-gnu
Thread model: posix
So tweak the affected test case to avoid triggering this bug. This is
the easiest way for us to keep the testsuite result clean and meaningful
on our main Clang test bed.
* t/ltcond2.sh: Prefer using "extern const char *" variables rather
than functions returning a statically allocated "const char *" variable.
* maint:
tests: typofix in the shebang line of 't/py-compile-destdir.sh'
HACKING: release process: fixlets about the announcement
maint: post-release minor version bump
release: stable release 1.12.5
tests: avoid errors due to underquoted third-party macros
maintcheck: also consider contrib tests for syntax checks
cosmetics: whitespace fixlets in syntax-checks.mk
maint: prefer $(MKDIR_P) over $(mkinstalldirs) in maintainer rules
HACKING: release process: fixlets about the announcement
* HACKING: The announcement should be generate *before* bumping
the version number to the next alpha number. Also, only an
abridged version of the announcement should be copied on the
Savannah news feed (a link to the complete announcement should
be linked from there, though).
* maint:
sync: update files from upstream with "make fetch"
gitignore: align with recent changes
fixup: add dummy rule for ChangeLog generation in Makefile.am
maint: no longer use AM_MAKEFLAGS in maintainer rules
maint: use more GNU make features in maintainer rules
maint: move maintainer make rules in maint.mk
1. It will enable us to take advantage of more GNU make
features in our maintainer-specific recipes, if we
want to (and we probably will).
2. The Makefile.am won't have to be modified each time a
maintainer rule needs to be changed or tweaked; this
will prevent useless remaking of (in particular) the
't/testsuite-part.am' file.
* maint.mk: New, several recipes, rules and variables for
maintainer-specific tasks, extracted (with some tweaks)
from ...
* Makefile.am: ... this file.
(EXTRA_DIST): Add 'maint.mk'.
* GNUmakefile: Include 'maint.mk' after 'Makefile'.
* maint:
docs: fixlet about xz compression levels and command line options
maint: more files excluded in 'update-copyright'
maint: adapt 'update-copyright' recipe to the new $(FETCHFILES) format
HACKING: update instructions to update the manuals at www.gnu.org
maint: add an explicative comment in Makefile
maint: factor out the name of the dir where web manuals are generated
maint: rename rules to build/upload web manuals
cosmetic: minor variable and rules reordering in Makefile.am
docs: build the web manuals for CVS in the builddir, not in the srcdir
release: automate uploading of web manuals in CVS
docs: AM_CFLAGS: remove reference to non-existing "more detailed" desc
tests: avoid two spurious failures on FreeBSD
tests: improve 'pkg-config' requirement
projects / thirdparty / automake.git / log
