MASTER_SITES= site1 \
site2
style continuation lines to be simple repeated
MASTER_SITES+= site1
MASTER_SITES+= site2
lines. As previewed on tech-pkg. With thanks to rillig for fixing pkglint
accordingly.
Issues found with existing distfiles:
distfiles/eclipse-sourceBuild-srcIncluded-3.0.1.zip
distfiles/fortran-utils-1.1.tar.gz
distfiles/ivykis-0.39.tar.gz
distfiles/enum-1.11.tar.gz
distfiles/pvs-3.2-libraries.tgz
distfiles/pvs-3.2-linux.tgz
distfiles/pvs-3.2-solaris.tgz
distfiles/pvs-3.2-system.tgz
No changes made to these distinfo files.
Otherwise, existing SHA1 digests verified and found to be the same on
the machine holding the existing distfiles (morden). All existing
SHA1 digests retained for now as an audit trail.
New in 1.15:
* Improvements and refactorings in the install-sh script:
- It has been modernized, and now makes the following assumptions
*unconditionally*:
(1) a working 'dirname' program is available;
(2) the ${var:-value} shell parameters substitution works;
(3) the "set -f" and "set +f" shell commands work, and, respectively,
disable and enable shell globbing.
- The script implements stricter error checking, and now it complains
and bails out if any of the following expectations is not met:
(1) the options -d and -t are never used together;
(2) the argument passed to option -t is a directory;
(3) if there are two or more SOURCEFILE arguments, the
DESTINATION argument must be a directory.
* Automake-generated testsuites:
- The default test-driver used by the Automake-generates testsuites
now appends the result and exit status of each "plain" test to the
associated log file (automake bug#11814).
- The perl implementation of the TAP testsuite driver is no longer
installed in the Automake's scripts directory, and is instead just
distributed as a "contrib" addition. There should be no reason to
use this implementation anyway in real packages, since the awk+shell
implementation of the TAP driver (which is documented in the manual)
is more portable and has feature parity with the perl implementation.
- The rule generating 'test-suite.log' no longer risk incurring in an
extra useless "make all" recursive invocation in some corner cases
(automake bug#16302).
* Distribution:
- Automake bug#18286: "make distcheck" could sometimes fail to detect
files missing from the distribution tarball, especially in those cases
where both the generated files and their dependencies are explicitly
in $(srcdir). An important example of this are *generated* makefile
fragments included at Automake time in Makefile.am; e.g.:
...
$(srcdir)/fragment.am: $(srcdir)/data.txt $(srcdir)/preproc.sh
cd $(srcdir) && $(SHELL) preproc.sh <data.txt >fragment.am
include $(srcdir)/fragment.am
...
If the use forgot to add data.txt and/or preproc.sh in the distribution
tarball, "make distcheck" would have erroneously succeeded! This issue
is now fixed.
- As a consequence of the previous change, "make distcheck" will run
using '$(distdir)/_build/sub' as the build directory, rather than
simply '$(distdir)/_build' (as it was the case for Automake 1.14 and
earlier). Consequently, the './configure' and 'make' invocations
issued by the distcheck recipe now have $(srcdir) equal to '../..',
rather than to just '..'. Dependent and similar variables (e.g.,
'$(top_srcdir)') are also changed accordingly.
Thus, Makefiles that made assumptions about the exact values of the
build and source directories used by "make distcheck" will have to
be adjusted. Notice that making such assumptions was a bad and
unsupported practice anyway, since the exact locations of those
directories should be considered implementation details, and we
reserve the right to change them at any time.
* Miscellaneous bugs fixed:
- The expansion of AM_INIT_AUTOMAKE ends once again with a trailing
newline (bug#16841). Regression introduced in Automake 1.14.
- We no longer risk to use '$ac_aux_dir' before it's defined (see
automake bug#15981). Bug introduced in Automake 1.14.
- The code used to detect whether the currently used make is GNU make
or not (relying on the private macro 'am__is_gnu_make') no longer
risks causing "Arg list too long" for projects using automatic
dependency tracking and having a ton of source files (bug#18744).
- Automake tries to offer a more deterministic output for generated
Makefiles, in the face of the newly-introduced randomization for
hash keys order in Perl 5.18.
- In older Automake versions, if a user defined one single Makefile
fragment (say 'foo.am') to be included via Automake includes in
his main Makefile.am, and defined a custom make rule to generate that
file from other data, Automake used to spuriously complain with some
message like "... overrides Automake target '$(srcdir)/foo.am".
This bug is now fixed.
- The user can now extend the special .PRECIOUS target, the same way
he could already do with the .MAKE .and .PHONY targets.
- Some confusing typos have been fixed in the manual and in few warning
messages (automake bug#16827 and bug#16997).
Do it for all packages that
* mention perl, or
* have a directory name starting with p5-*, or
* depend on a package starting with p5-
like last time, for 5.18, where this didn't lead to complaints.
Let me know if you have any this time.
New in 1.14.1:
* Bugs fixed:
- The user is no longer allowed to override the --srcdir nor the --prefix
configure options used by "make distcheck" (bug#14991).
- Fixed a gross inefficiency in the recipes for installing byte-compiled
python files, that was causing an O(N^2) performance on the number N of
files, instead of the expected O(N) performance. Note that this bug
was only relevant when the number of python files was high (which is
unusual in practice).
- Automake try to offer a more reproducible output for warning messages,
in the face of the newly-introduced randomization for hash keys order
in Perl 5.18.
- The 'test-driver' script now actually error out with a clear error
message on the most common invalid usages.
- Several spurious failures/hangs in the testsuite (bugs #14706, #14707,
#14760, #14911, #15181, #15237).
* Documentation fixes:
- Fixed typos in the 'fix-timestamp.sh' example script that made it
nonsensical.
* WARNING: New versioning scheme for Automake.
- Beginning with the release 1.13.2, Automake has started to use a
more rational versioning scheme, that should allow users to know
which kind of changes can be expected from a new version, based
on its version number.
+ Micro releases (e.g., 1.13.3, 2.0.1, 3.2.8) introduce only bug
and regression fixes and documentation updates; they should not
introduce new features, nor any backward-incompatibility (any
such incompatibility would be considered a bug, to be fixed with
a further micro release).
+ Minor releases (e.g., 1.14, 2.1) can introduce new backward
compatible features; the only backward-incompatibilities allowed
in such a release are new *non-fatal* deprecations and warnings,
and possibly fixes for old or non-trivial bugs (or even inefficient
behaviours) that could unfortunately have been seen and used by
some as "corner case features". Possible disruptions caused by
this kind of fixes should hopefully be quite rare, and their
effects limited in scope.
+ Major versions (now expected to be released every 18 or 24 months,
and not more often) can introduce new big features (possibly with
rough edges and not-fully-stabilized APIs), removal of deprecated
features, backward-incompatible changes of behaviour, and possibly
major refactorings (that, while ideally transparent to the user,
could introduce new bugs). Incompatibilities should however not
be introduced gratuitously and abruptly; a proper deprecation path
should be duly implemented in the preceding minor releases.
- According to this new scheme, the next major version of Automake
(the one that had previously been labelled as "1.14") will actually
become "Automake 2.0". Automake 1.14 is *this* release (which is
a minor one). It introduces new features, deprecations and bug
fixes, but no serious backward incompatibility. A partial exception
is given by the behavioural changes in the AM_PROG_CC_C_O macro
(described in details below) but such changes can also be seen as a
fix for the old suboptimal and somewhat confusing behaviour.
- See discussion about automake bug#13578 for more details and
background: <http://debbugs.gnu.org/cgi/bugreport.cgi?bug=13578>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
* WARNING: Future backward-incompatibilities!
- Makefile recipes generated by Automake 2.0 will expect to use an
'rm' program that doesn't complain when called without any non-option
argument if the '-f' option is given (so that commands like "rm -f"
and "rm -rf" will act as a no-op, instead of raising usage errors).
This behavior of 'rm' is very widespread in the wild, and it will be
required in the next POSIX version:
<http://austingroupbugs.net/view.php?id=542>
Accordingly, AM_INIT_AUTOMAKE now expands some shell code that checks
that the default 'rm' program in PATH satisfies this requirement,
aborting the configure process if this is not the case. For the
moment, it's still possible to force the configuration process to
succeed even with a broken 'rm', that that will no longer be the case
for Automake 2.0.
- Automake 2.0 will require Autoconf 2.70 or later (which is still
unreleased at the moment of writing, but is planned to be released
before Automake 2.0 is).
- Automake 2.0 will drop support for the long-deprecated 'configure.in'
name for the Autoconf input file. You are advised to start using the
recommended name 'configure.ac' instead, ASAP.
- The ACLOCAL_AMFLAGS special make variable will be fully deprecated in
Automake 2.0: it will raise warnings in the "obsolete" category (but
still no hard error of course, for compatibilities with the many, many
packages that still relies on that variable). You are advised to
start relying on the new Automake support for AC_CONFIG_MACRO_DIRS
instead (which was introduced in Automake 1.13).
- Automake 2.0 will remove support for automatic dependency tracking
with the SGI C/C++ compilers on IRIX. The SGI depmode has been
reported broken "in the wild" already, and we don't think investing
time in debugging and fixing is worthwhile, especially considering
that SGI has last updated those compilers in 2006, and is expected
to retire support for them in December 2013:
<http://www.sgi.com/services/support/irix_mips_support.html>
- Automake 2.0 will remove support for MS-DOS and Windows 95/98/ME
(support for them was offered by relying on the DJGPP project).
Note however that both Cygwin and MSYS/MinGW on modern Windows
versions will continue to be fully supported.
- Automake-provided scripts and makefile recipes might (finally!)
start assuming a POSIX shell in Automake 2.0. There still is no
certainty about this though: we'd first like to wait and see
whether future Autoconf versions will be enhanced to guarantee
that such a shell is always found and provided by the checks in
./configure.
- Starting from Automake 2.0, third-party m4 files located in the
system-wide aclocal directory, as well as in any directory listed
in the ACLOCAL_PATH environment variable, will take precedence
over "built-in" Automake macros. For example (assuming Automake
is installed in the /usr/local hierarchy), a definition of the
AM_PROG_VALAC macro found in '/usr/local/share/aclocal/my-vala.m4'
should take precedence over the same-named automake-provided macro
(defined in '/usr/local/share/aclocal-2.0/vala.m4').
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
New in 1.14:
* C compilation, and the AC_PROG_CC and AM_PROG_CC_C_O macros:
- The 'compile' script is now unconditionally required for all packages
that perform C compilation (if you are using the '--add-missing'
option, automake will fetch that script for you, so you shouldn't
need any explicit adjustment). This new behaviour is needed to avoid
obscure errors when the 'subdir-objects' option is used, and the
compiler is an inferior one that doesn't grasp the combined use of
both the "-c -o" options; see discussion about automake bug#13378 for
more details:
<http://debbugs.gnu.org/cgi/bugreport.cgi?bug=13378#35>
<http://debbugs.gnu.org/cgi/bugreport.cgi?bug=13378#44>
- The next major Automake version (2.0) will unconditionally activate
the 'subdir-objects' option. In order to smooth out the transition,
we now give a warning (in the category 'unsupported') whenever a
source file is present in a subdirectory but the 'subdir-object' is
not enabled. For example, the following usage will trigger such a
warning:
bin_PROGRAMS = sub/foo
sub_foo_SOURCES = sub/main.c sub/bar.c
- Automake will automatically enhance the autoconf-provided macro
AC_PROG_CC to force it to check, at configure time, that the
C compiler supports the combined use of both the '-c' and '-o'
options. The result of this check is saved in the cache variable
'am_cv_prog_cc_c_o', and said result can be overridden by
pre-defining that variable.
- The AM_PROG_CC_C_O macro can still be called, albeit that should no
longer be necessary. This macro is now just a thin wrapper around the
Automake-enhanced AC_PROG_CC. This means, among the other things,
that its behaviour is changed in three ways:
1. It no longer invokes the Autoconf-provided AC_PROG_CC_C_O
macro behind the scenes.
2. It caches the check result in the 'am_cv_prog_cc_c_o' variable,
and not in a 'ac_cv_prog_cc_*_c_o' variable whose exact name is
dynamically computed only at configure runtime (really!) from
the content of the '$CC' variable.
3. It no longer automatically AC_DEFINE the C preprocessor
symbol 'NO_MINUS_C_MINUS_O'.
* Texinfo support:
- Automake can now be instructed to place '.info' files generated from
Texinfo input in the builddir rather than in the srcdir; this is done
specifying the new 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.
- For quite a long time, Automake has been implementing an undocumented
hack which ensured that '.info' files which appeared to be cleaned
(by 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 package
such as Texinfo, which do 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 generated '.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 will
likely be removed altogether in Automake 2.0.
* Relative directory in Makefile fragments:
- The special Automake-time substitutions '%reldir%' and '%canon_reldir%'
(and their short versions, '%D%' and '%C%' respectively) can now be used
in an included Makefile fragment. The former is substituted with the
relative directory of the included fragment (compared to the top-level
including Makefile), and the latter with the canonicalized version of
the same relative directory.
# in 'Makefile.am':
bin_PROGRAMS = # will be updated by included Makefile fragments
include src/Makefile.inc
# in 'src/Makefile.inc':
bin_PROGRAMS += %reldir%/foo
%canon_reldir%_foo_SOURCES = %reldir%/bar.c
This should be especially useful for packages using a non-recursive
build system.
* Deprecated distribution formats:
- The 'shar' and 'compress' distribution formats are deprecated, and
scheduled for removal in Automake 2.0. Accordingly, the use of the
'dist-shar' and 'dist-tarZ' will cause warnings at automake runtime
(in the 'obsolete' category), and the recipes of the Automake-generated
targets 'dist-shar' and 'dist-tarZ' will unconditionally display
(non-fatal) warnings at make runtime.
* New configure runtime warnings about "rm -f" support:
- To simplify transition to Automake 2.0, the shell code expanded by
AM_INIT_AUTOMAKE now checks (at configure runtime) that the default
'rm' program in PATH doesn't complain when called without any
non-option argument if the '-f' option is given (so that commands like
"rm -f" and "rm -rf" act as a no-op, instead of raising usage errors).
If this is not the case, the configure script is aborted, to call the
attention of the user on the issue, and invite him to fix his PATH.
The checked 'rm' behavior is very widespread in the wild, and will be
required by future POSIX versions:
<http://austingroupbugs.net/view.php?id=542>
The user can still force the configure process to complete even in the
presence of a broken 'rm' by defining the ACCEPT_INFERIOR_RM_PROGRAM
environment variable to "yes". And the generated Makefiles should
still work correctly even when such broken 'rm' is used. But note
that this will no longer be the case with Automake 2.0 though, so, if
you encounter the warning, please report it to us ASAP (and try to fix
your environment as well).
New in 1.13.3:
* Documentation fixes:
- The documentation no longer mistakenly reports that the obsolete
'AM_MKDIR_PROG_P' macro and '$(mkdir_p)' make variable are going
to be removed in Automake 2.0.
* Bugs fixed:
- Byte-compilation of Emacs lisp files could fail spuriously on
Solaris, when /bin/ksh or /usr/xpg4/bin/sh were used as shell.
- If the same user-defined suffixes were transformed into different
Automake-known suffixes in different Makefile.am files in the same
project, automake could get confused and generate inconsistent
Makefiles (automake bug#14441).
For example, if 'Makefile.am' contained a ".ext.cc:" suffix rule,
and 'sub/Makefile.am' contained a ".ext.c:" suffix rule, automake
would have mistakenly placed into 'Makefile.in' rules to compile
"*.c" files into object files, and into 'sub/Makefile.in' rules to
compile "*.cc" files into object files --- rather than the other
way around. This is now fixed.
* Testsuite work:
- The test cases no longer have the executable bit set. This should
make it clear that they are not meant to be run directly; as
explained in t/README, they can only be run through the custom
'runtest' script, or by a "make check" invocation.
- The testsuite has seen the introduction of a new helper function
'run_make', and several related changes. These serve a two-fold
purpose:
1. Remove brittleness due to the use of "make -e" in test cases.
2. Seamlessly allow the use of parallel make ("make -j...") in the
test cases, even where redirection of make output is involved
(see automake bug#11413 for a description of the subtle issues
in this area).
- Several spurious failures have been fixed (they hit especially
MinGW/MSYS builds). See automake bugs #14493, #14494, #14495,
#14498, #14499, #14500, #14501, #14517 and #14528.
- Some other minor miscellaneous changes and fixlets.
a) refer 'perl' in their Makefile, or
b) have a directory name of p5-*, or
c) have any dependency on any p5-* package
Like last time, where this caused no complaints.
* WARNING: New versioning scheme for Automake.
- Starting with this version onward, Automake will use an update and
more rational versioning scheme, one that will allow users to know
which kind of changes can be expected from a new version, based on
its version number.
+ Micro versions (e.g., 1.13.3, 2.0.1, 3.2.8) will introduce only
documentation updates and bug and regression fixes; they will
not introduce new features, nor any backward-incompatibility (any
such incompatibility would be considered a bug, to be fixed with
a further micro release).
+ Minor versions (e.g., 1.14, 2.1) can introduce new backward
compatible features; the only backward-incompatibilities allowed
in such a release are new *non-fatal* deprecations and warnings,
and possibly fixes for old or non-trivial bugs (or even inefficient
behaviours) that could unfortunately have been seen, and used, by
some developers as "corner case features". Possible disruptions
caused by this kind of fixes should hopefully be quite rare.
+ Major versions (now expected to be released every 18 or 24 months,
and not more often) can introduce new big features (possibly with
rough edges and not-fully-stabilized APIs), removal of deprecated
features, backward-incompatible changes of behaviour, and possibly
major refactorings (that, while ideally transparent to the user,
could introduce new bugs). Incompatibilities should however not
be introduced gratuitously and abruptly; a proper deprecation path
should be duly implemented in the preceding minor releases.
- According to this new scheme, the next major version of Automake
(the one that has until now been labelled as '1.14') will actually
become "Automake 2.0". Automake 1.14 will be the next minor version,
which will introduce new features, deprecations and bug fixes, but
no real backward incompatibility.
- See discussion about automake bug#13578 for more details and
background: <http://debbugs.gnu.org/cgi/bugreport.cgi?bug=13578>
* WARNING: Future backward-incompatibilities!
- Automake 2.0 will require Autoconf 2.70 or later (which is still
unreleased at the moment of writing, but is planned to be released
before Automake 2.0 is).
- Automake 2.0 will drop support for the long-deprecated 'configure.in'
name for the Autoconf input file. You are advised to start using the
recommended name 'configure.ac' instead, ASAP.
- The ACLOCAL_AMFLAGS special make variable will be fully deprecated
in Automake 2.0 (where it will raise warnings in the "obsolete"
category). You are advised to start relying on the new Automake
support for AC_CONFIG_MACRO_DIRS instead (which was introduced in
Automake 1.13).
- Automake 2.0 will remove support for automatic dependency tracking
with the SGI C/C++ compilers on IRIX. The SGI depmode has been
reported broken "in the wild" already, and we don't think investing
time in debugging and fixing is worthwhile, especially considering
that SGI has last updated those compilers in 2006, and is expected
to retire support for them in December 2013:
<http://www.sgi.com/services/support/irix_mips_support.html>
- Future versions of Automake might remove support for MS-DOS and
Windows 95/98/ME (support for them was offered by relying on the
DJGPP project). Note however that both Cygwin and MSYS/MinGW on
modern Windows versions will continue to be fully supported.
- Automake-provided scripts and makefile recipes might (finally!)
start assuming a POSIX shell in Automake 2.0.
- Starting from Automake 2.0, third-party m4 files located in the
system-wide aclocal directory, as well as in any directory listed
in the ACLOCAL_PATH environment variable, will take precedence
over "built-in" Automake macros. For example (assuming Automake
is installed in the /usr/local hierarchy), a definition of the
AM_PROG_VALAC macro found in '/usr/local/share/aclocal/my-vala.m4'
should take precedence over the same-named automake-provided macro
(defined in '/usr/local/share/aclocal-2.0/vala.m4').
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
New in 1.13.2:
* Obsolescent features:
- Use of suffix-less info files (that can be specified through the
'@setfilename' macro in Texinfo input files) is discouraged, and
its use will raise warnings in the 'obsolete' category.
- Use of Texinfo input files with '.txi' or '.texinfo' extensions
is discouraged, and its use will raise warnings in the 'obsolete'
category. You are advised to simply use the '.texi' extension
instead.
* Documentation fixes:
- The long-deprecated but still supported two-arguments invocation form
of AM_INIT_AUTOMAKE is documented once again. This seems the sanest
thing to do, given that support for such an usage might need to remain
in place for a unspecified amount of time in order to cater for people
who want to define the version number for their package dynamically at
configure runtime (unfortunately, Autoconf does not yet support this
scenario, so we cannot delegate the work to it).
- The serial testsuite harness is no longer reported as "deprecated",
but as "discouraged". We have no plan to remove it, not to make its
use cause runtime warnings.
- The parallel testsuite is no longer reported as "experimental"; it
is well tested, and should be stable now.
- The 'shar' and 'tarZ' distribution formats and the 'dist-shar' and
'dist-tarZ' options are obsolescent, and their use is deprecated
in the documentation.
- Other minor miscellaneous fixes and improvements; in particular,
some improvements in cross-references.
* Bugs fixed:
- When the 'ustar' option is used, the generated configure script no
longer risks hanging during the tests for the availability of the
'pax' utility, even if the user running configure has a UID or GID
that requires more than 21 bits to be represented.
See automake bug#8343 and bug#13588.
- The obsolete macros AM_CONFIG_HEADER or AM_PROG_CC_STDC work once
again, as they did in Automake 1.12.x (albeit printing runtime
warnings in the 'obsolete' category). Removing them has turned
out to be a very bad idea, because it complicated distro packing
enormously. Making them issue fatal warnings, as we did in
Automake 1.13, has turned out to be a similarly very bad idea,
for exactly the same reason.
- aclocal will no longer error out if the first local m4 directory
(as specified by the '-I' option or the 'AC_CONFIG_MACRO_DIRS' or
'AC_CONFIG_MACRO_DIR' macros) doesn't exist; it will merely report
a warning in the 'unsupported' category. This is done to support
some pre-existing real-world usages. See automake bug#13514.
- aclocal will no longer consider directories for extra m4 files more
than once, even if they are specified multiple times. This ensures
packages that specify both
AC_CONFIG_MACRO_DIR([m4]) in configure.ac
ACLOCAL_AMFLAGS = -I m4 in Makefile.am
will work correctly, even when the 'm4' directory contains no
package-specific files, but is used only to install third-party
m4 files (as can happen with e.g., "libtoolize --install").
See automake bug#13514.
- Analysis of make flags in Automake-generated rules has been made more
robust, and more future-proof. For example, in presence of make that
(like '-I') take an argument, the characters in said argument will no
longer be spuriously considered as a set of additional make options.
In particular, automake-generated rules will no longer spuriously
believe to be running in dry mode ("make -n") if run with an invocation
like "make -I noob"; nor will they believe to be running in keep-going
mode ("make -k") if run with an invocation like "make -I kool"
(automake bug#12554).
New in 1.13.1:
* WARNING: Future backward-incompatibilities!
- Automake 1.14 will likely require Autoconf 2.70 or later (which is
still unreleased at the moment of writing, but is planned to be
released before Automake 1.14 is).
- Automake 1.14 will likely drop support for the long-deprecated
'configure.in' name for the Autoconf input file. You are advised
to use the recommended name 'configure.ac' instead.
- The long-obsolete (since automake 1.10) AM_PROG_MKDIR m4 macro will
be removed in Automake 1.14. The $(mkdir_p) make variable and the
@mkdir_p@ substitution will still remain available (as aliases of
$(MKDIR_P)) for the moment, for better backward compatibility; but
you are advised to stop using ASAP.
- The ACLOCAL_AMFLAGS special make variable will be fully deprecated
in Automake 1.14 (where it will raise warnings in the "obsolete"
category). You are advised to start relying on the new Automake
support for AC_CONFIG_MACRO_DIRS instead (which is introduced with
this release; see below for more information).
- Support for IRIX and the SGI C/C++ compilers will be removed in
Automake 1.14: they have seen their last release in 2006, and SGI
is expected to retire support from them in December 2013; see
<http://www.sgi.com/services/support/irix_mips_support.html> for
more information.
- Future versions of Automake might remove support for MS-DOS and
Windows 95/98/ME (support for them was offered by relying on the
DJGPP project). Note however that both Cygwin and MSYS/MinGW on
modern Windows versions will continue to be fully supported.
- Support for the long-deprecated INCLUDES variable will be removed
altogether in Automake 1.14. The AM_CPPFLAGS variable should be
used instead.
- Automake-provided scripts and makefile recipes might (finally!)
start assuming a POSIX shell in Automake 1.14.
- Starting from Automake 1.14, third-party m4 files located in the
system-wide aclocal directory, as well as in any directory listed
in the ACLOCAL_PATH environment variable, will take precedence
over "built-in" Automake macros. For example (assuming Automake
is installed in the /usr/local hierarchy), a definition of the
AM_PROG_VALAC macro found in '/usr/local/share/aclocal/my-vala.m4'
should take precedence over the same-named automake-provided macro
(defined in '/usr/local/share/aclocal-1.14/vala.m4').
* Bugs fixed:
- Use of the obsolete macros AM_CONFIG_HEADER or AM_PROG_CC_STDC now
causes a clear and helpful error message, instead of obscure ones
(issue introduced in Automake 1.13).
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
New in 1.13:
* Bugs fixed:
- ylwrap renames properly header guards in generated header files
(*.h), instead of leaving Y_TAB_H.
- ylwrap now also converts header guards in implementation files
(*.c). 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.
* Version requirements:
- Autoconf 2.65 or greater is now required.
- The rules to build PDF and DVI output from Texinfo input now
require Texinfo 4.9 or later.
* Obsolete features:
- Support for the "Cygnus-style" trees (once enabled by the 'cygnus'
option) has been removed. See discussion about automake bug#11034
for more background: <http://debbugs.gnu.org/11034>.
- The deprecated aclocal option '--acdir' has been removed. You
should use the options '--automake-acdir' and '--system-acdir'
instead (which have been introduced in Automake 1.11.2).
- The following long-obsolete m4 macros have been removed:
AM_PROG_CC_STDC: superseded by AC_PROG_CC since October 2002
fp_PROG_CC_STDC: broken alias for AM_PROG_CC_STDC
fp_WITH_DMALLOC: old alias for AM_WITH_DMALLOC
AM_CONFIG_HEADER: superseded by AC_CONFIG_HEADERS since July 2002
ud_PATH_LISPDIR: old alias for AM_PATH_LISPDIR
jm_MAINTAINER_MODE: old alias for AM_MAINTAINER_MODE
ud_GNU_GETTEXT: old alias for AM_GNU_GETTEXT
gm_PROG_LIBTOOL: old alias for AC_PROG_LIBTOOL
fp_C_PROTOTYPES: old alias for AM_C_PROTOTYPES (which was part
of the now-removed automatic de-ANSI-fication
support of Automake)
- All the "old alias" macros in 'm4/obsolete.m4' have been removed.
- Use of the long-deprecated two- and three-arguments invocation forms
of the AM_INIT_AUTOMAKE is no longer documented. It's still supported
though (albeit with a warning in the 'obsolete' category), to cater
for people who want to define the version number for their package
dynamically (e.g., from the current VCS revision). We'll have to
continue this support until Autoconf itself is fixed to allow better
support for such dynamic version numbers.
* Elisp byte-compilation:
- The byte compilation of '.el' files into '.elc' files is now done
with a suffix rule. This has simplified the compilation process, and
more importantly made it less brittle. The downside is that emacs is
now invoked once for each '.el' files, which cause some noticeable
slowdowns. These should however be mitigated on multicore machines
(which are becoming the norm today) if concurrent make ("make -j")
is used.
- Elisp files placed in a subdirectory are now byte-compiled to '.elc'
files in the same subdirectory; for example, byte-compiling of file
'sub/foo.el' file will result in 'sub/foo.elc' rather than in
'foo.elc'. This behaviour is backward-incompatible with older
Automake versions, but it is more natural and more sane. See also
automake bug#7441.
- The Emacs invocation performing byte-compilation of '.el' files honors
the $(AM_ELCFLAGS) and $(ELCFLAGS) variables; as typical, the former
one is developer-reserved and the latter one user-reserved.
- The 'elisp-comp' script, once provided by Automake, has been rendered
obsoleted by the just-described changes, and thus removed.
* Changes to Automake-generated testsuite harnesses:
- The parallel testsuite harness (previously only enabled by the
'parallel-tests' option) is the default one; the older serial
testsuite harness will still be available through the use of the
'serial-tests' option (introduced in Automake 1.12).
- The 'color-tests' option is now unconditionally activated by default.
In particular, this means that testsuite output is now colorized by
default if the attached terminal seems to support ANSI escapes, and
that the user can force output colorization by setting the variable
AM_COLOR_TESTS to "always". The 'color-tests' is still recognized
for backward-compatibility, although it's a handled as a no-op now.
* Silent rules support:
- Support for silent rules is now always active in Automake-generated
Makefiles. So, although the verbose output is still the default,
the user can now always use "./configure --enable-silent-rules" or
"make V=0" to enable quieter output in the package he's building.
- The 'silent-rules' option has now become a no-op, preserved for
backward-compatibility only. In particular, its use no longer
disables the warnings in the 'portability-recursive' category.
* Texinfo Support:
- The rules to build PDF and DVI files from Texinfo input now require
Texinfo 4.9 or later.
- The rules to build PDF and DVI files from Texinfo input now use the
'--build-dir' option, to keep the auxiliary files used by texi2dvi
and texi2pdf around without cluttering the build directory, and to
make it possible to run the "dvi" and "pdf" recipes in parallel.
* Automatic remake rules and 'missing' script:
- The 'missing' script no longer tries to update the timestamp of
out-of-date files that require a maintainer-specific tool to be
remade, in case the user lacks such a tool (or has a too-old version
of it). It just gives a useful warning, and in some cases also a
tip about how to obtain such a tool.
- The missing script has thus become useless as a (poor) way to work
around the sketched-timestamps issues that can happen for projects
that keep generated files committed in their VCS repository. Such
projects are now encouraged to write a custom "fix-timestamps.sh"
script to avoid such issues; a simple example is provided in the
"CVS and generated files" chapter of the automake manual.
* Recursive targets:
- The user can now define his own recursive targets that recurse
in the directories specified in $(SUBDIRS). This can be done by
specifying the name of such targets in invocations of the new
'AM_EXTRA_RECURSIVE_TARGETS' m4 macro.
* Tags:
- Any failure in the recipe of the "tags", "ctags", "cscope" or
"cscopelist" targets in a subdirectory is now propagated to the
top-level make invocation.
- Tags are correctly computed also for files in _SOURCES variables that
only list files with non-standard suffixes (see automake bug#12372).
* Improvements to aclocal and related rebuilds rules:
- Autoconf-provided macros AC_CONFIG_MACRO_DIR and AC_CONFIG_MACRO_DIRS
are now traced by aclocal, and can be used to declare the local m4
include directories. Formerly, one had to specify it with an explicit
'-I' option to the 'aclocal' invocation.
- The special make variable ACLOCAL_AMFLAGS is deprecated; future
Automake versions will warn about its use, and later version will
remove support for it altogether.
* The depcomp script:
- Dropped support for libtool 1.4.
- Various internal refactorings. They should cause no visible change,
but the chance for regression is there anyway, so please report any
unexpected or suspicious behaviour.
- Support for pre-8.0 versions of the Intel C Compiler has been dropped.
This should cause no problem, since icc 8.0 has been released in
December 2003 -- almost nine years ago.
- Support for tcc (the Tiny C Compiler) has been improved, and is now
handled through a dedicated 'tcc' mode.
* The ylwrap script:
- ylwrap generates header guards with a single '_' for series of non
alphabetic characters, instead of several. This is what Bison >=
2.5.1 does.
(same backwards compat warnings related to 1.13 as in 1.12.5)
Bugs fixed in 1.12.6:
* Python-related bugs:
- The default installation location for python modules has been improved
for Python 3 on Debian and Ubuntu systems, changing from:
${prefix}/lib/python3/dist-packages
to
${prefix}/lib/python3.x/site-packages
This change should ensure modules installed using the default ${prefix}
"/usr/local" are found by default by system python 3.x installations.
See automake bug#10227.
- Python byte-compilation supports the new layout mandated by PEP-3147,
with its __pycache__ directory (automake bug#8847).
* Build system issues:
- The maintainer rebuild rules for Makefiles and aclocal.m4 in
Automake's own build system works correctly again (bug introduced
in Automake 1.12.5).
* Testsuite issues:
- The Vala-related tests has been changed to adjust to the removal of
the 'posix' profile in the valac compiler. See automake bug#12934
a.k.a. bug#12522.
- Some spurious testsuite failures related to older tools and systems
have been fixed.
New in 1.12.5:
* WARNING: Future backward-incompatibilities!
- Future versions of Automake will likely drop support for the
long-deprecated 'configure.in' name for the Autoconf input file.
You are advised to use the recommended name 'configure.ac' instead.
- Support for the "Cygnus-style" trees (as enabled by the 'cygnus'
option) will be removed in the next major Automake release (1.13).
- The long-obsolete (since automake 1.10) AM_PROG_MKDIR m4 macro will
be removed in Automake 1.13. The $(mkdir_p) make variable and the
@mkdir_p@ substitution will still remain available (as aliases of
$(MKDIR_P)) for the moment, for better backward compatibility.
- Autoconf 2.65 or later will be required by the next major Automake
version (1.13). Until now, Automake has required Autoconf version
2.62 or later.
- Starting from the next major Automake version (1.13), the rules
to build pdf, ps and dvi output from Texinfo input will use the
'--build-dir' option by default. Since such an option was only
introduced in Texinfo 4.9, this means that Makefiles generated by
future Automake versions will require at least that version of
Texinfo.
- Starting from the next major Automake version (1.13), the parallel
testsuite harness (previously only enabled by the 'parallel-tests'
option) will become the default one; the older serial testsuite
harness will still be available through the use of the 'serial-tests'
option.
- The following long-obsolete m4 macros will be removed in the
next major Automake version (1.13):
AM_PROG_CC_STDC: superseded by AC_PROG_CC since October 2002
fp_PROG_CC_STDC: broken alias for AM_PROG_CC_STDC
fp_WITH_DMALLOC: old alias for AM_WITH_DMALLOC
AM_CONFIG_HEADER: superseded by AC_CONFIG_HEADERS since July 2002
ud_PATH_LISPDIR: old alias for AM_PATH_LISPDIR
jm_MAINTAINER_MODE: old alias for AM_MAINTAINER_MODE
ud_GNU_GETTEXT: old alias for AM_GNU_GETTEXT
gm_PROG_LIBTOOL: old alias for AC_PROG_LIBTOOL
fp_C_PROTOTYPES: old alias for AM_C_PROTOTYPES (which was part
of the now-removed automatic de-ANSI-fication
support of Automake)
- All the "old alias" macros in 'm4/obsolete.m4' will be removed in
the next major Automake version (1.13).
- The '--acdir' option of aclocal is deprecated, and will probably
be removed in the next major Automake release (1.13). You should
use the options '--automake-acdir' and '--system-acdir' instead
(which have been introduced in Automake 1.11.2).
- The 'missing' script will no longer try to update the timestamp
of out-of-date files that require a maintainer-specific tool to be
remade, in case the user lacks such a tool (or has a too-old version
of it). In fact, starting from Automake 1.13, all it'll do will be
giving more useful warnings than a bare "command not found" from a
make recipe would.
* Vala support:
- The AM_PROG_VALAC macro has been enhanced to takes two further
optional arguments; it's signature now being
AM_PROG_VALAC([MINIMUM-VERSION], [ACTION-IF-FOUND],
[ACTION-IF-NOT-FOUND])
- By default, AM_PROG_VALAC no longer aborts the configure invocation
if the Vala compiler found is too old, but simply prints a warning
messages (as it did when the Vala compiler was not found). This
should avoid unnecessary difficulties for end users that just want
to compile the unmodified, distributed Vala-generated C sources,
but happens to have an old Vala compiler in their PATH. This fixes
automake bug#12688.
- If no proper Vala compiler is found at configure runtime, AM_PROG_VALAC
will set the AC_SUBST'd variable 'VALAC' to 'valac' rather than to ':'.
This is a better default, because with it a triggered makefile rule
invoking a Vala compilation will clearly fail with an informative error
message like "valac: command not found", rather than silently, with
the error possibly going unnoticed or triggering harder-to-diagnose
fallout failures in later steps.
* Miscellaneous changes:
- automake and aclocal no longer honours the 'perllibdir' environment
variable. That had always been intended only as an hack required in
the testsuite, not meant for any use beyond that.
Bugs fixed in 1.12.5:
* Long-standing bugs:
- Automake no longer generates spurious remake rules invoking autoheader
to regenerate the template corresponding to header files specified after
the first one in AC_CONFIG_HEADERS (automake bug#12495).
- When wrapping Microsoft tools, the 'compile' script falls back to
finding classic 'libname.a' style libraries when 'name.lib' and
'name.dll.lib' aren't available.
New in 1.12.4:
* WARNING: Future backward-incompatibilities!
- Future versions of Automake will likely drop support for the
long-deprecated 'configure.in' name for the Autoconf input file.
You are advised to use the recommended name 'configure.ac' instead.
- Support for the "Cygnus-style" trees (as enabled by the 'cygnus'
option) will be removed in the next major Automake release (1.13).
- The long-obsolete (since automake 1.10) AM_PROG_MKDIR m4 macro will
be removed in Automake 1.13. The $(mkdir_p) make variable and the
@mkdir_p@ substitution will still remain available (as aliases of
$(MKDIR_P)) for the moment, for better backward compatibility.
- Autoconf 2.65 or later will be required by the next major Automake
version (1.13). Until now, Automake has required Autoconf version
2.62 or later.
- Starting from the next major Automake version (1.13), the rules
to build pdf, ps and dvi output from Texinfo input will use the
'--build-dir' option by default. Since such an option was only
introduced in Texinfo 4.9, this means that Makefiles generated by
future Automake versions will require at least that version of
Texinfo.
- Starting from the next major Automake version (1.13), the parallel
testsuite harness (previously only enabled by the 'parallel-tests'
option) will become the default one; the older serial testsuite
harness will still be available through the use of the 'serial-tests'
option.
- The following long-obsolete m4 macros will be removed in the
next major Automake version (1.13):
AM_PROG_CC_STDC: superseded by AC_PROG_CC since October 2002
fp_PROG_CC_STDC: broken alias for AM_PROG_CC_STDC
fp_WITH_DMALLOC: old alias for AM_WITH_DMALLOC
AM_CONFIG_HEADER: superseded by AC_CONFIG_HEADERS since July 2002
ud_PATH_LISPDIR: old alias for AM_PATH_LISPDIR
jm_MAINTAINER_MODE: old alias for AM_MAINTAINER_MODE
ud_GNU_GETTEXT: old alias for AM_GNU_GETTEXT
gm_PROG_LIBTOOL: old alias for AC_PROG_LIBTOOL
fp_C_PROTOTYPES: old alias for AM_C_PROTOTYPES (which was part
of the now-removed automatic de-ANSI-fication
support of Automake)
- All the "old alias" macros in 'm4/obsolete.m4' will be removed in
the next major Automake version (1.13).
- The '--acdir' option of aclocal is deprecated, and will probably
be removed in the next major Automake release (1.13). You should
use the options '--automake-acdir' and '--system-acdir' instead
(which have been introduced in Automake 1.11.2).
- The 'missing' script will not try anymore to update the timestamp
of out-of-date files that require a maintainer-specific tool to be
remade, in case the user lacks such a tool (or has a too-old version
of it). In fact, starting from Automake 1.13, all it'll do will be
giving more useful warnings than a bare "command not found" from a
make recipe would.
* Warnings and deprecations:
- Warnings in the 'obsolete' category are enabled by default both in
automake and aclocal.
* Miscellaneous changes:
- Some testsuite weaknesses and spurious failures have been fixed.
New in 1.12.3:
* WARNING: Future backward-incompatibilities!
- Future versions of Automake will likely drop support for the
long-deprecated 'configure.in' name for the Autoconf input file.
You are advised to use the recommended name 'configure.ac' instead.
- The long-obsolete (since automake 1.10) AM_PROG_MKDIR m4 macro will
be removed in Automake 1.13. The $(mkdir_p) make variable and the
@mkdir_p@ substitution will still remain available (as aliases of
$(MKDIR_P)) for the moment, for better backward compatibility.
- Autoconf 2.65 or later will be required by the next major Automake
version (1.13). Until now, Automake has required Autoconf version
2.62 or later.
- Starting from the next major Automake version (1.13), the rules
to build pdf, ps and dvi output from Texinfo input will use the
'--build-dir' option by default. Since such an option was only
introduced in Texinfo 4.9, this means that Makefiles generated by
future Automake versions will require at least that version of
Texinfo.
- Starting from the next major Automake version (1.13), the parallel
testsuite harness (previously only enabled by the 'parallel-tests'
option) will become the default one; the older serial testsuite
harness will still be available through the use of the 'serial-tests'
option.
- The following long-obsolete m4 macros will be removed in the
next major Automake version (1.13):
AM_PROG_CC_STDC: superseded by AC_PROG_CC since October 2002
fp_PROG_CC_STDC: broken alias for AM_PROG_CC_STDC
fp_WITH_DMALLOC: old alias for AM_WITH_DMALLOC
AM_CONFIG_HEADER: superseded by AC_CONFIG_HEADERS since July 2002
ud_PATH_LISPDIR: old alias for AM_PATH_LISPDIR
jm_MAINTAINER_MODE: old alias for AM_MAINTAINER_MODE
ud_GNU_GETTEXT: old alias for AM_GNU_GETTEXT
gm_PROG_LIBTOOL: old alias for AC_PROG_LIBTOOL
fp_C_PROTOTYPES: old alias for AM_C_PROTOTYPES (which was part
of the now-removed automatic de-ANSI-fication
support of Automake)
- All the "old alias" macros in 'm4/obsolete.m4' will be removed in
the next major Automake version (1.13).
- Support for the two- and three-arguments invocation forms of the
AM_INIT_AUTOMAKE macro is deprecated, and will be removed in the
next major Automake version (1.13).
- The '--acdir' option of aclocal is deprecated, and will probably
be removed in the next major Automake release (1.13). You should
use the options '--automake-acdir' and '--system-acdir' instead
(which have been introduced in Automake 1.11.2).
- The exact order in which the directories in the aclocal macro
search path are looked up is probably going to be changed in the
next Automake release (1.13).
- The 'missing' script will not try anymore to update the timestamp
of out-of-date files that require a maintainer-specific tool to be
remade, in case the user lacks such a tool (or has a too-old version
of it). In fact, starting from Automake 1.13, all it'll do will be
giving more useful warnings than a bare "command not found" from a
make recipe would.
* Miscellaneous changes:
- The '.m4' files provided by Automake does not define serial numbers
anymore. This should cause no difference in the behaviour of aclocal
though.
- Some testsuite weaknesses and spurious failures have been fixed.
- There is initial support for automatic dependency tracking with the
Portland Group C/C++ compilers, thanks to the new new depmode 'pgcc'.
Bugs fixed in 1.12.3:
* Long-standing bugs:
- Instead of renaming only self-references of files (typically for
#lines), ylwrap now also renames references to the other generated
files. This fixes support for GLR and C++ parsers from Bison (PR
automake/491 and automake bug#7648): 'parser.c' now properly
#includes 'parser.h' instead of 'y.tab.h'.
- Generated files unknown to ylwrap are now preserved. This fixes
C++ support for Bison (automake bug#7648): location.hh and the
like are no longer discarded.
New in 1.12.1:
* WARNING: Future backward-incompatibilities!
- Starting from either the next minor version (1.12.2) or the next major
version (1.13), Automake will start warning if 'configure.in' is used
instead of 'configure.ac' as the Autoconf input. Future versions of
Automake will drop support for 'configure.in' altogether.
- Autoconf 2.65 or later will be required by the next major Automake
version (1.13). Until now, Automake has required Autoconf version
2.62 or later.
- Starting from the next major Automake version (1.13), the rules to
build pdf, ps and dvi output from Texinfo input will use the '--tidy'
option by default. Since such an option was introduced in Texinfo
4.9, this means that Makefiles generated by future Automake versions
will require at least that version of Texinfo.
- Starting from the next major Automake version (1.13), the parallel
testsuite harness (previously only enabled by the 'parallel-tests'
option) will become the default one; the older serial testsuite
harness will still be available through the use of the 'serial-tests'
option.
- The following long-obsolete m4 macros will be removed in the
next major Automake version (1.13):
AM_PROG_CC_STDC: superseded by AC_PROG_CC since October 2002
fp_PROG_CC_STDC: broken alias for AM_PROG_CC_STDC
fp_WITH_DMALLOC: old alias for AM_WITH_DMALLOC
AM_CONFIG_HEADER: superseded by AC_CONFIG_HEADERS since July 2002
ud_PATH_LISPDIR: old alias for AM_PATH_LISPDIR
jm_MAINTAINER_MODE: old alias for AM_MAINTAINER_MODE
ud_GNU_GETTEXT: old alias for AM_GNU_GETTEXT
gm_PROG_LIBTOOL: old alias for AC_PROG_LIBTOOL
fp_C_PROTOTYPES: old alias for AM_C_PROTOTYPES (which was part
of the now-removed automatic de-ANSI-fication
support of Automake)
- All the "old alias" macros in 'm4/obsolete.m4' will be removed in
the next major Automake version (1.13).
- Support for the two- and three-arguments invocation forms of the
AM_INIT_AUTOMAKE macro will be deprecated in the next minor version
of Automake (1.12.1) and removed in the next major version (1.13).
- The '--acdir' option of aclocal is deprecated, and will probably
be removed in the next major Automake release (1.13). You should
use the options '--automake-acdir' and '--system-acdir' instead
(which have been introduced in Automake 1.11.2).
- The exact order in which the directories in the aclocal macro
search path are looked up is probably going to be changed in the
next Automake release (1.13).
* New supported languages:
- Support for Objective C++ has been added; it should work similarly to
the support for Objective C.
* Deprecated obsolescent features:
- Use of the long-deprecated two- and three-arguments invocation forms
of the AM_INIT_AUTOMAKE macro now elicits a warning in the 'obsolete'
category. Starting from the next major Automake release (1.13), such
usages won't be allowed anymore.
- Support for the "Cygnus-style" trees (enabled by the 'cygnus' option) is
now deprecated (its use triggers a warning in the 'obsolete' category).
It will be removed in the next major Automake release (1.13).
- The long-obsolete (since 1.10) automake-provided $(mkdir_p) make
variable, @mkdir_p@ configure-time substitution and AM_PROG_MKDIR
m4 macro are deprecated, eliciting a warning in the 'obsolete'
category. They will be removed in the next major version (1.13).
* Miscellaneous changes:
- The Automake test cases now require a proper POSIX-conforming shell.
Older non-POSIX Bourne shells (like Solaris 10 /bin/sh) won't be
accepted anymore. In most cases, the user shouldn't have to specify
such POSIX shell explicitly, since it will be looked up at configure
time. Still, when this lookup fails, or when the user wants to
override its conclusion, the variable 'AM_TEST_RUNNER_SHELL' can be
used (pointing to the shell that will be used to run the Automake
test cases).
Bugs fixed in 1.12.1:
* Bugs introduced by 1.12:
- Several weaknesses in Automake's own build system and test suite
have been fixed.
* Bugs introduced by 1.11.3:
- When given non-option arguments, aclocal rejects them, instead of
silently ignoring them.
* Long-standing bugs:
- When the 'color-tests' option is in use, forcing of colored testsuite
output through "AM_COLOR_TESTS=always" works even if the terminal is
a non-ANSI one, i.e., if the TERM environment variable has a value of
"dumb".
- Several inefficiencies and poor performances in the implementation
of the parallel-tests 'check' and 'recheck' targets have been fixed.
- The post-processing of output "#line" directives done the ylwrap
script is more faithful w.r.t. files in a subdirectory; for example,
if the processed file is "src/grammar.y", ylwrap will correctly
produce directives like:
#line 7 "src/grammar.y"
rather than like
#line 7 "grammar.y"
as it did before.
* Bugs with new Perl versions:
- Aclocal works correctly with perl 5.16.0 (automake bug#11543).
* WARNING: Future backward-incompatibilities!
- Starting from the next major Automake version (1.13), the rules to
build pdf, ps and dvi output from Texinfo input will use the '--tidy'
option by default. Since such an option was introduced in Texinfo
4.9, this means that Makefiles generated by future Automake versions
will require at least that version of Texinfo.
- Starting from the next major Automake version (1.13), the parallel
testsuite harness (previously only enabled by the 'parallel-tests'
option) will become the default one; the older serial testsuite
harness will still be available through the use of the 'serial-tests'
option.
- The following long-obsolete m4 macros will be removed in the
next major Automake version (1.13):
AM_PROG_CC_STDC: superseded by AC_PROG_CC since October 2002
fp_PROG_CC_STDC: broken alias for AM_PROG_CC_STDC
fp_WITH_DMALLOC: old alias for AM_WITH_DMALLOC
AM_CONFIG_HEADER: superseded by AC_CONFIG_HEADERS since July 2002
ud_PATH_LISPDIR: old alias for AM_PATH_LISPDIR
jm_MAINTAINER_MODE: old alias for AM_MAINTAINER_MODE
ud_GNU_GETTEXT: old alias for AM_GNU_GETTEXT
gm_PROG_LIBTOOL: old alias for AC_PROG_LIBTOOL
fp_C_PROTOTYPES: old alias for AM_C_PROTOTYPES (which was part
of the now-removed automatic de-ANSI-fication
support of Automake)
- All the "old alias" macros in 'm4/obsolete.m4' will be removed in
the next major Automake version (1.13).
- Support for the "Cygnus-style" trees (enabled by the 'cygnus' option)
will be deprecated in the next minor version of Automake (1.12.1) and
removed in the next major version (1.13).
- Support for the two- and three-arguments invocation forms of the
AM_INIT_AUTOMAKE macro will be deprecated in the next minor version
of Automake (1.12.1) and removed in the next major version (1.13).
- The long-obsolete (since 1.10) automake-provided $(mkdir_p) make
variable, @mkdir_p@ substitution and AM_PROG_MKDIR m4 macro will
all be deprecated in the next minor version of Automake (1.12.1)
and removed in the next major version (1.13).
- The '--acdir' option of aclocal is deprecated, and will probably
be removed in the next major Automake release (1.13). You should
use the options '--automake-acdir' and '--system-acdir' instead
(which have been introduced in Automake 1.11.2).
- The exact order in which the directories in the aclocal macro
search path are looked up is probably going to be changed in the
next Automake release (1.13).
* Obsolete features removed:
- The never documented nor truly used script 'acinstall' has been
removed.
- Support for automatic de-ANSI-fication has been removed.
- The support for the "obscure" multilib feature has been removed
from Automake core (but remains available in the 'contrib/'
directory of the Automake distribution).
- Support for ".log -> .html" conversion and the check-html and
recheck-html targets has been removed from Automake core (but
remains available in the 'contrib/' directory of the Automake
distribution).
- The deprecated 'lzma' compression format for distribution archives
has been removed, in favor of 'xz' and 'lzip'.
- The obsolete AM_WITH_REGEX macro has been removed.
- The long-deprecated options '--output-dir', '--Werror' and
'--Wno-error' have been removed.
- The chapter on the history of Automake has been moved out of the
reference manual, into a new dedicated Texinfo file.
* New targets:
- New 'cscope' target to build a cscope database for the source tree.
* Changes to Automake-generated testsuite harnesses:
- The new automake option 'serial-tests' has been introduced. It can
be used to explicitly instruct automake to use the older serial
testsuite harness. This is still the default at the moment, but it
might change in future versions.
- The 'recheck' target (provided by the parallel testsuite harness) now
depends on the 'all' target. This allows for a better user-experience
in test-driven development. See automake bug#11252.
- Test scripts that exit with status 99 to signal an "hard error" (e.g.,
and unexpected or internal error, or a failure to set up the test case
scenario) have their outcome reported as an 'ERROR' now. Previous
versions of automake reported such an outcome as a 'FAIL' (the only
difference with normal failures being that hard errors were counted
as failures even when the test originating them was listed in
XFAIL_TESTS).
- The testsuite summary displayed by the parallel-test harness has a
completely new format, that always list the numbers of passed, failed,
xfailed, xpassed, skipped and errored tests, even when these numbers
are zero (but using smart coloring when the color-tests option is in
effect).
- The default testsuite driver offered by the 'parallel-tests' option is
now implemented (partly at least) with the help of automake-provided
auxiliary scripts (e.g., 'test-driver'), instead of relying entirely
on code in the generated Makefile.in.
This has two noteworthy implications. The first one is that projects
using the 'parallel-tests' option should now either run automake with
the '--add-missing' option, or manually copy the 'test-driver' script
into their tree. The second, and more important, implication is that
now, when the 'parallel-tests' option is in use, TESTS_ENVIRONMENT can
not be used anymore to define a test runner, and the command specified
in LOG_COMPILER (and <ext>_LOG_COMPILER) must be a *real* executable
program or script. For example, this is still a valid usage (albeit
a little contorted):
TESTS_ENVIRONMENT = \
if test -n '$(STRICT_TESTS)'; then \
maybe_errexit='-e'; \
else \
maybe_errexit=''; \
fi;
LOG_COMPILER = $(SHELL) $$maybe_errexit
while this is not anymore:
TESTS_ENVIRONMENT = \
$(SHELL) `test -n '$(STRICT_TESTS_CHECKING)' && echo ' -e'`
neither is this:
TESTS_ENVIRONMENT = \
run_with_perl_or_shell () \
{ \
if grep -q '^#!.*perl' $$1; then
$(PERL) $$1; \
else \
$(SHELL) $$1; \
fi; \
}
LOG_COMPILER = run_with_perl_or_shell
- The package authors can now use customary testsuite drivers within
the framework provided by the 'parallel-tests' testsuite harness.
Consistently with the existing syntax, this can be done by defining
special makefile variables 'LOG_DRIVER' and '<ext>_LOG_DRIVER'.
- A new developer-reserved variable 'AM_TESTS_FD_REDIRECT' can be used
to redirect/define file descriptors used by the test scripts.
- The parallel-tests harness generates now, in addition the '.log' files
holding the output produced by the test scripts, a new set of '.trs'
files, holding "metadata" derived by the execution of the test scripts;
among such metadata are the outcomes of the test cases run by a script.
- Initial and still experimental support for the TAP test protocol is
now provided.
* Changes to Yacc and Lex support:
- C source and header files derived from non-distributed Yacc and/or
Lex sources are now removed by a simple "make clean" (while they were
previously removed only by "make maintainer-clean").
- Slightly backward-incompatible change, relevant only for use of Yacc
with C++: the extensions of the header files produced by the Yacc
rules are now modelled after the extension of the corresponding
sources. For example, yacc files named "foo.y++" and "bar.yy" will
produce header files named "foo.h++" and "bar.hh" respectively, where
they would have previously produced header files named simply "foo.h"
and "bar.h". This change offers better compatibility with 'bison -o'.
* Miscellaneous changes:
- The AM_PROG_VALAC macro now causes configure to exit with status 77,
rather than 1, if the vala compiler found is too old.
- The build system of Automake itself now avoids the use of make
recursion as much as possible.
- Automake now prefers to quote 'like this' or "like this", rather
than `like this', in diagnostic message and generated Makefiles,
to accommodate the new GNU Coding Standards recommendations.
- Automake has a new option '--print-libdir' that prints the path of the
directory containing the Automake-provided scripts and data files.
- The 'dist' and 'dist-all' targets now can run compressors in parallel.
- The rules to create pdf, dvi and ps output from Texinfo files now
works better with modern 'texi2dvi' script, by explicitly passing
it the '--clean' option to ensure stray auxiliary files are not
left to clutter the build directory.
- Automake can now generate silenced rules for texinfo outputs.
- Some auxiliary files that are automatically distributed by Automake
(e.g., 'install-sh', or the 'depcomp' script for packages compiling
C sources) might now be listed in the DIST_COMMON variable in many
Makefile.in files, rather than in the top-level one.
- Messages of types warning or error from 'automake' and 'aclocal'
are now prefixed with the respective type, and presence of -Werror
is noted.
- Automake's early configure-time sanity check now tries to avoid
sleeping for a second, which slowed down cached configure runs
noticeably. In that case, it will check back at the end of the
configure script to ensure that at least one second has passed, to
avoid time stamp issues with makefile rules rerunning autotools
programs.
- The warnings in the category 'extra-portability' are now enabled by
'-Wall'. In previous versions, one has to use '-Wextra-portability'
to enable them.
Bugs fixed in 1.12:
- Various minor bugfixes for recent or long-standing bugs.
* Bugs introduced by 1.11:
- The AM_COND_IF macro also works if the shell expression for the
conditional is no longer valid for the condition.
- The automake-provided parallel testsuite harness does not fail anymore
with BSD make used in parallel mode when there are test scripts in a
subdirectory, like in:
TESTS = sub/foo.test sub/bar.test
* Long-standing bugs:
- Automake's own build system finally have a real "installcheck" target.
- Vala-related cleanup rules are now more complete, and work better in
a VPATH setup.
- Files listed with the AC_REQUIRE_AUX_FILE macro in configure.ac are
now automatically distributed also if the directory of the auxiliary
files coincides with the top-level directory.
- Automake now detects the presence of the '-d' flag in the various
'*YFLAGS' variables even when their definitions involve indirections
through other variables, such as in:
foo_opts = -d
AM_YFLAGS = $(foo_opts)
- Automake now complains if a '*YFLAGS' variable has any conditional
content, not only a conditional definition.
- Explicit enabling and/or disabling of Automake warning categories
through the '-W...' options now always takes precedence over the
implicit warning level implied by Automake strictness (foreign, gnu
or gnits), regardless of the order in which such strictness and
warning flags appear. For example, a setting like:
AUTOMAKE_OPTIONS = -Wall --foreign
will cause the warnings in category 'portability' to be enabled, even
if those warnings are by default disabled in 'foreign' strictness.
Bugs fixed in 1.11.5:
* Bugs introduced by 1.11.3:
- Vala files with '.vapi' extension are now recognized and handled
correctly again. See automake bug#11222.
- Vala support work again for projects that contain some program
built from '.vala' (and possibly '.c') sources and some other
program built from '.c' sources *only*. See automake bug#11229.
New in 1.11.4:
* WARNING: Future backward-incompatibilities!
- The support for the "obscure" multilib feature has been deprecated,
and will be moved out of the automake core in the next major Automake
release (1.12).
- The support for ".log -> .html" conversion and the check-html and
recheck-html targets will be removed in the next major Automake
release (1.12).
- The obsolescent AM_WITH_REGEX macro has been deprecated (since the
GNU rx library has been decommissioned), and will be removed in the
next major Automake release (1.12).
- The `lzma' compression format for distribution archives has been
deprecated in favor of `xz' and `lzip', and will be removed in the
next major Automake release (1.12).
- The `--acdir' option of aclocal is deprecated, and will probably be
removed in the next major Automake release (1.12).
- The exact order in which the directories in the aclocal macro
search path are looked up is probably going to be changed in the
next Automake release (1.12).
- The Automake support for automatic de-ANSI-fication will be removed
in the next major Automake release (1.12).
- Starting from the next Automake release (1.12), warnings in the
`extra-portability' category will be enabled by `-Wall' (right now,
one has to use `-Wextra-portability' explicitly).
* Miscellaneous changes:
- The 'ar-lib' script now ignores the "s" (symbol index) and "S" (no
symbol index) modifiers as well as the "s" action, as the symbol index
is created unconditionally by Microsoft lib. Also, the "q" (quick)
action is now a synonym for "r" (replace). Also, the script has been
ignoring the "v" (verbose) modifier already since Automake 1.11.3.
- When the 'compile' script is used to wrap MSVC, it now accepts an
optional space between the -I, -L and -l options and their respective
arguments, for better POSIX compliance.
- There is an initial, experimental support for automatic dependency
tracking with tcc (the Tiny C Compiler). Its associated depmode is
currently recognized as "icc" (but this and other details are likely
to change in future versions).
- Automatic dependency tracking now works also with the IBM XL C/C++
compilers, thanks to the new new depmode 'xlc'.
Bugs fixed in 1.11.4:
* Bugs introduced by 1.11.2:
- A definition of 'noinst_PYTHON' before 'python_PYTHON' (or similar)
don't cause spurious failures upon "make install" anymore.
- The user can now instruct the 'uninstall-info' rule not to update
the '${infodir}/dir' file by exporting the environment variable
'AM_UPDATE_INFO_DIR' to the value "no". This is done for consistency
with how the 'install-info' rule operates since automake 1.11.2.
* Long-standing bugs:
- It is now possible for a foo_SOURCES variable to hold Vala sources
together with C header files, as well as with sources and headers for
other supported languages (e.g., C++). Previously, only mixing C and
Vala sources was supported.
- If "aclocal --install" is used, and the first directory specified with
'-I' is non-existent, aclocal will now create it before trying to copy
files in it.
- An empty declaration of a "foo_PRIMARY" don't cause anymore the
generated install rules to create an empty $(foodir) directory;
for example, if Makefile.am contains something like:
pkglibexec_SCRIPTS =
if FALSE
pkglibexec_SCRIPTS += bar.sh
endif
the $(pkglibexec) directory will not be created upon "make install".
New in 1.11.3:
* WARNING: Future backward-incompatibilities!
- The support for the "obscure" multilib feature has been deprecated,
and will be moved out of the automake core in the next major Automake
release (1.12).
- The support for ".log -> .html" conversion and the check-html and
recheck-html targets will be removed in the next major Automake
release (1.12).
- The obsolescent AM_WITH_REGEX macro has been deprecated (since the
GNU rx library has been decommissioned), and will be removed in the
next major Automake release (1.12).
- The `lzma' compression format for distribution archives has been
deprecated in favor of `xz' and `lzip', and will be removed in the
next major Automake release (1.12).
- The `--acdir' option of aclocal is deprecated, and will probably be
removed in the next major Automake release (1.12).
- The exact order in which the directories in the aclocal macro
search path are looked up is probably going to be changed in the
next Automake release (1.12).
- The Automake support for automatic de-ANSI-fication will be removed
in the next major Automake release (1.12).
- Starting from the next Automake release (1.12), warnings in the
`extra-portability' category will be enabled by `-Wall' (right now,
one has to use `-Wextra-portability' explicitly).
* Miscellaneous changes:
- Automake's own build system is more silent by default, making use of
the 'silent-rules' option.
- The master copy of the `gnupload' script is now maintained in gnulib,
not in automake.
- The `missing' script doesn't try to wrap calls to `tar' anymore.
- "make dist" doesn't wrap `tar' invocations with the `missing' script
anymore. Similarly, the obsolescent variable `$(AMTAR)' (which you
shouldn't be using BTW ;-) does not invoke the missing script anymore
to wrap tar, but simply invokes the `tar' program itself.
- "make dist" can now create lzip-compressed tarballs.
- In the Automake info documentation, the Top node and the nodes about
the invocation of the automake and aclocal programs have been renamed;
now, calling "info automake" will open the Top node, while calling
"info automake-invocation" and "info aclocal-invocation" will access
the nodes about the invocation of respectively automake and aclocal.
- Automake is now distributed as a gzip-compressed and an xz-compressed
tarball. Previously, bzip2 was used instead of xz.
- The last relics of Python 1.5 support have been removed from the
AM_PATH_PYTHON macro.
- For programs and libraries, automake now detects EXTRA_foo_DEPENDENCIES
and adds them to the normal list of dependencies, but without
overwriting the foo_DEPENDENCIES variable, which is normally computed
by automake.
Bugs fixed in 1.11.3:
* Bugs introduced by 1.11.2:
- Automake now correctly recognizes the prefix/primary combination
`pkglibexec_SCRIPTS' as valid.
- The parallel-tests harness doesn't trip anymore on sed implementations
with stricter limits on the length of input lines (problem seen at
least on Solaris 8).
* Long-standing bugs:
- The "deleted header file problem" for *.am files is avoided by stub
rules. This allows `make' to trigger a rerun of `automake' also if
some previously needed `.am' file has been removed.
- The `silent-rules' option now generates working makefiles even
for the uncommon `make' implementations that do not support the
nested-variables extension to POSIX 2008. For such `make'
implementations, whether a build is silent is determined at
configure time, and cannot be overridden at make time with
`make V=0' or `make V=1'.
- Vala support now works better in VPATH setups.
New in 1.11.2:
* Changes to aclocal:
- The `--acdir' option is deprecated. Now you should use the new options
`--automake-acdir' and `--system-acdir' instead.
- The `ACLOCAL_PATH' environment variable is now interpreted as a
colon-separated list of additional directories to search after the
automake internal acdir (by default ${prefix}/share/aclocal-APIVERSION)
and before the system acdir (by default ${prefix}/share/aclocal).
* Miscellaneous changes:
- The Automake support for automatic de-ANSI-fication has been
deprecated. It will probably be removed in the next major Automake
release (1.12).
- The `lzma' compression scheme and associated automake option `dist-lzma'
is obsoleted by `xz' and `dist-xz' due to upstream changes.
- You may adjust the compression options used in dist-xz and dist-bzip2.
The default is now merely -e for xz, but still -9 for bzip; you may
specify a different level via the XZ_OPT and BZIP2 envvars respectively.
E.g., "make dist-xz XZ_OPT=-7" or "make dist-bzip2 BZIP2=-5"
- The `compile' script now converts some options for MSVC for a better
user experience. Similarly, the new `ar-lib' script wraps Microsoft lib.
- The py-compile script now accepts empty arguments passed to the options
`--destdir' and `--basedir', and complains about unrecognized options.
Moreover, a non-option argument or a special `--' argument terminates
the list of options.
- A developer that needs to pass specific flags to configure at "make
distcheck" time can now, and indeed is advised to, do so by defining
the developer-reserved makefile variable AM_DISTCHECK_CONFIGURE_FLAGS,
instead of the old DISTCHECK_CONFIGURE_FLAGS.
The DISTCHECK_CONFIGURE_FLAGS variable should now be reserved for the
user; still, the old Makefile.am files that used to define it will
still continue to work as before.
- New macro AM_PROG_AR that looks for an archiver and wraps it in the new
'ar-lib' auxiliary script if the selected archiver is Microsoft lib.
This new macro is required for LIBRARIES and LTLIBRARIES when automake
is run with -Wextra-portability and -Werror.
- When using DejaGnu-based testsuites, the user can extend the `site.exp'
file generated by automake-provided rules by defining the special make
variable `$(EXTRA_DEJAGNU_SITE_CONFIG)'.
- The `install-info' rule can now be instructed not to create/update
the `${infodir}/dir' file, by exporting the new environment variable
`AM_UPDATE_INFO_DIR' to the value "no".
Bugs fixed in 1.11.2:
* Bugs introduced by 1.11:
- The parallel-tests driver no longer produces erroneous results with
Tru64/OSF 5.1 sh upon unreadable log files.
- The `parallel-tests' test driver does not report spurious successes
when used with concurrent FreeBSD make (e.g., "make check -j3").
- When the parallel-tests driver is in use, automake now explicitly
rejects invalid entries and conditional contents in TEST_EXTENSIONS,
instead of issuing confusing and apparently unrelated error messages
(e.g., "non-POSIX variable name", "bad characters in variable name",
or "redefinition of TEST_EXTENSIONS), or even, in some situations,
silently producing broken `Makefile.in' files.
- The `silent-rules' option now truly silences all compile rules, even
when dependency tracking is disabled. Also, when `silent-rules' is
not used, `make' output no longer contains spurious backslash-only
lines, thus once again matching what Automake did before 1.11.
- The AM_COND_IF macro also works if the shell expression for the
conditional is no longer valid for the condition.
* Long-standing bugs:
- The order of Yacc and Lex flags is fixed to be consistent with other
languages: $(AM_YFLAGS) comes before $(YFLAGS), and $(AM_LFLAGS) before
$(LFLAGS), so that the user variables override the developer variables.
- "make distcheck" now correctly complains also when "make uninstall"
leaves one and only one file installed in $(prefix).
- A "make uninstall" issued before a "make install", or after a mere
"make install-data" or a mere "make install-exec" does not spuriously
fail anymore.
- Automake now warns about more primary/directory invalid combinations,
such as "doc_LIBRARIES" or "pkglib_PROGRAMS".
- Rules generated by Automake now try harder to not change any files when
`make -n' is invoked. Fixes include compilation of Emacs Lisp, Vala, or
Yacc source files and the rule to update config.h.
- Several scripts and the parallel-tests testsuite driver now exit with
the right exit status upon receiving a signal.
- A per-Makefile.am setting of -Werror does not erroneously carry over
to the handling of other Makefile.am files.
- The code for automatic dependency tracking works around a Solaris
make bug triggered by sources containing repeated slashes when the
`subdir-objects' option was used.
- The makedepend and hp depmodes now work better with VPATH builds.
- Java sources specified with check_JAVA are no longer compiled for
"make all", but only for "make check".
- An usage like "java_JAVA = foo.java" will now cause Automake to warn
and error out if `javadir' is undefined, instead of silently producing
a broken Makefile.in.
- aclocal and automake now honour the configure-time definitions of
AUTOCONF and AUTOM4TE when they spawn autoconf or autom4te processes.
- The `install-info' recipe no longer tries to guess whether the
`install-info' program is from Debian or from GNU, and adaptively
change its behaviour; this has proven to be frail and easy to
regress.
Bugs fixed in 1.11.1:
- Lots of minor bugfixes.
* Bugs introduced by 1.11:
- The `parallel-tests' test driver works around a GNU make 3.80 bug with
trailing white space in the test list (`TESTS = foo $(EMPTY)').
* Long standing bugs:
- On Darwin 9, `pythondir' and `pyexecdir' pointed below `/Library/Python'
even if the `--prefix' argument pointed outside of a system directory.
AM_PATH_PYTHON has been fixed to ignore the value returned from python's
`get_python_lib' function if it points outside the configured prefix,
unless the `--prefix' argument was either `/usr' or below `/System'.
- The testsuite does not try to change the mode of `ltmain.sh' files from
a Libtool installation (symlinked to test directories) any more.
- AM_PROG_GCJ uses AC_CHECK_TOOLS to look for `gcj' now, so that prefixed
tools are preferred in a cross-compile setup.
- The distribution is tarred up with mode 755 now by the `dist*' targets.
This fixes a race condition where untrusted users could modify files
in the $(PACKAGE)-$(VERSION) distdir before packing if the toplevel
build directory was world-searchable. This is CVE-2009-4029.
New in 1.11:
* Version requirements:
- Autoconf 2.62 or greater is required.
* Changes to aclocal:
- The autoconf version check implemented by aclocal in aclocal.m4
(and new in Automake 1.10) is degraded to a warning. This helps
in the common case where the Autoconf versions used are compatible.
* Changes to automake:
- The automake program can run multiple threads for creating most
Makefile.in files concurrently, if at least Perl 5.7.2 is available
with interpreter-based threads enabled. Set the environment variable
AUTOMAKE_JOBS to the maximum number of threads to use, in order to
enable this experimental feature.
* Changes to Libtool support:
- Libtool generic flags are now passed to the install and uninstall
modes as well.
- distcheck works with Libtool 2.x even when LT_OUTPUT is used, as
config.lt is removed correctly now.
* Languages changes:
- subdir-object mode works now with Fortran (F77, FC, preprocessed
Fortran, and Ratfor).
- For files with extension .f90, .f95, .f03, or .f08, the flag
$(FCFLAGS_f[09]x) computed by AC_FC_SRCEXT is now used in compile rules.
- Files with extension .sx are also treated as preprocessed assembler.
- The default source file extension (.c) can be overridden with
AM_DEFAULT_SOURCE_EXT now.
- Python 3.0 is supported now, Python releases prior to 2.0 are no
longer supported.
- AM_PATH_PYTHON honors python's idea about the site directory.
- There is initial support for the Vala programming language, when using
Vala 0.7.0 or later.
* Miscellaneous changes:
- Automake development is done in a git repository on Savannah now, see
http://git.sv.gnu.org/gitweb/?p=automake.git
A read-only CVS mirror is provided at
cvs -d :pserver:anonymous@pserver.git.sv.gnu.org:/automake.git \
checkout -d automake HEAD
- "make dist" can now create xz-compressed tarballs,
as well as (deprecated?) lzma-compressed tarballs.
- `automake --add-missing' will by default install the GPLv3 file as
COPYING if it is missing. It will also warn that the license file
should be added to source control. Note that Automake will never
overwrite an existing COPYING file, even when the `--force-missing'
option is used.
- The manual is now distributed under the terms of the GNU FDL 1.3.
- Automake ships and installs man pages for automake and aclocal now.
- New shorthand `$(pkglibexecdir)' for `$(libexecdir)/@PACKAGE@'.
- install-sh supports -C, which does not update the installed file
(and its time stamps) if the contents did not change.
- The `gnupload' script has been revamped.
- The `depcomp' and `compile' scripts now work with MSVC under MSYS.
- The targets `install' and `uninstall' are more efficient now, in that
for example multiple files from one Automake variable such as
`bin_SCRIPTS' are copied in one `install' (or `libtool --mode=install')
invocation if they do not have to be renamed.
Both install and uninstall may sometimes enter (`cd' into) the target
installation directory now, when no build-local scripts are used.
Both install and uninstall do not fail anymore but do nothing if an
installation directory variable like `bindir' is set to the empty string.
For built-in rules, `make install' now fails reliably if installation
of a file failed. Conversely, `make uninstall' even succeeds when
issued multiple times.
These changes may need some adjustments from users: For example,
some `install' programs refuse to install multiple copies of the
same file in one invocation, so you may need to remove duplicate
entries from file lists.
Also, within one set of files, say, nobase_data_DATA, the order of
installation may be changed, or even unstable among different hosts,
due to the use of associative arrays in awk. The increased use of
awk matches a similar move in Autoconf to provide for better scaling.
Further, most undocumented per-rule install command variables such as
binSCRIPT_INSTALL have been removed because they are not needed any
more. Packages which use them should be using the appropriate one of
INSTALL_{DATA,PROGRAM,SCRIPT} or their install_sh_{DATA,PROGRAM,SCRIPT}
counterpart, depending on the type of files and the need for automatic
target directory creation.
- The "deleted header file problem" for *.m4 files is avoided by
stub rules. This allows `make' to trigger a rerun of `aclocal'
also if some previously needed macro file has been removed.
- Rebuild rules now also work for a removed `subdir/Makefile.in' in
an otherwise up to date tree.
- The `color-tests' option causes colored test result output on terminals.
- The `parallel-tests' option enables a new test driver that allows for
parallel test execution, inter-test dependencies, lazy test execution
for unit-testing, re-testing only failed tests, and formatted result output
as RST (reStructuredText) and HTML. Enabling this option may require some
changes to your test suite setup; see the manual for details.
- The `silent-rules' option enables Linux kernel-style silent build output.
This option requires the widely supported but non-POSIX `make' feature
of recursive variable expansion, so do not use it if your package needs
to build with `make' implementations that do not support it.
To enable less verbose build output, the developer has to use the Automake
option `silent-rules' in `AM_INIT_AUTOMAKE', or call the `AM_SILENT_RULES'
macro. The user may then set the default verbosity by passing the
`--enable-silent-rules' option to `configure'. At `make' run time, this
default may be overridden using `make V=0' for less verbose, and `make V=1'
for backward-compatible verbose output.
- New prefix `notrans_' for manpages which should not be transformed
by --program-transform.
- New macro AM_COND_IF for conditional evaluation and conditional
config files.
- For AC_CONFIG_LINKS, if source and destination are equal, do not
remove the file in a non-VPATH build. Such setups work with Autoconf
2.62 or newer.
- AM_MAINTAINER_MODE now allows for an optional argument specifying
the default setting.
- AM_SUBST_NOTMAKE may prevent substitution of AC_SUBSTed variables,
useful especially for multi-line values.
- Automake's early configure-time sanity check now diagnoses an
unsafe absolute source directory name and makes configure fail.
- The Automake macros and rules cope better with whitespace in the
current directory name, as long as the relative path to `configure'
does not contain whitespace. To this end, the values of `$(MISSING)'
and `$(install_sh)' may contain suitable quoting, and their expansion
might need `eval'uation if used outside of a makefile. These
undocumented variables may be used in several documented macros such
as $(AUTOCONF) or $(MAKEINFO).
Bugs fixed in 1.11:
* Long standing bugs:
- Fix aix dependency tracking for libtool objects.
- Work around AIX sh quoting issue in AC_PROG_CC_C_O, leading to
unnecessary use of the `compile' script.
- For nobase_*_LTLIBRARIES with nonempty directory components, the
correct `-rpath' argument is used now.
- `config.status --file=Makefile depfiles' now also works with the
extra quoting used internally by Autoconf 2.62 and newer
(it used to work only without the `--file=' bit).
- The `missing' script works better with versioned tool names.
- Semantics for `missing help2man' have been revamped:
Previously, if `help2man' was not present, `missing help2man' would have
the following semantics: if some man page was out of date but present, then
a warning would be printed, but the exit status was 0. If the man page was
not present at all, then `missing' would create a replacement man page
containing an error message, and exit with a status of 2. This does not play
well with `make': the next run will see this particular man page as being up
to date, and will only error out on the next generated man page, if any;
repeat until all pages are done. This was not desirable.
These are the new semantics: if some man page is not present, and help2man
is not either, then `missing' will warn and generate the replacement page
containing the error message, but exit successfully. However, `make dist'
will ensure that no such bogus man pages are packaged into a tarball.
- Targets provided by automake behave better with `make -n', in that they
take care not to create files.
- `config.status Makefile... depfiles' works fine again in the presence of
disabled dependency tracking.
- The default no-op recursive rules for these targets also work with BSD make
now: html, install-html, install-dvi, install-pdf, install-pdf, install-info.
- `make distcheck' works also when both a directory and some file below it
have been added to a distribution variable, such as EXTRA_DIST or *_SOURCES.
- Texinfo dvi, ps, pdf, and html output files are not removed upon
`make mostlyclean' any more; only the LaTeX by-products are.
- Renamed objects also work with the `subdir-objects' option and
source file languages which Automake does not know itself.
- `automake' now correctly complains about variable assignments which are
preceded by a comment, extend over multiple lines with backslash-escaped
newlines, and end in a comment sign. Previous versions would silently
and wrongly ignore such assignments completely.
* Bugs introduced by 1.10:
- Fix output of dummy dependency files in presence of post-processed
Makefile.in's again, but also cope with long lines.
- $(EXEEXT) is automatically appended to filenames of XFAIL_TESTS
that have been declared as programs in the same Makefile.
This is for consistency with the analogous change to TESTS in 1.10.
- Fix order of standard includes to again be `-I. -I$(srcdir)',
followed by directories containing config headers.
New in 1.10.1:
- Automake development is done in a git repository on Savannah now, see
http://git.sv.gnu.org/gitweb/?p=automake.git
A read-only CVS mirror is provided at
cvs -d :pserver:anonymous@pserver.git.sv.gnu.org:/automake.git \
checkout -d automake HEAD
- "make dist" can now create lzma-compressed tarballs.
- `automake --add-missing' will by default install the GPLv3 file as
COPYING if it is missing. Note that Automake will never overwrite
an existing COPYING file, even when the `--force-missing' option is
used. Further note that Automake 1.10.1 is still licensed under GPLv2+.
- Libtool generic flags are now passed to the install and uninstall
modes as well.
- Files with extension .sx are also treated as preprocessed assembler.
- install-sh now has an BSD-like option `-C' to preserve modification
times of unchanged files upon installation.
Bugs fixed in 1.10.1:
* Long standing bugs:
- Fix aix dependency tracking for libtool objects.
- The signal handling of aclocal has been improved.
- Targets beginning with a digit are now recognized correctly.
- All directories `.libs'/`_libs' used by libtool are cleaned now,
not only those in which libraries are built.
* Bugs introduced by 1.10:
- Fix output of dummy dependency files in presence of post-processed
Makefile.in's again, but also cope with long lines.
- $(EXEEXT) is automatically appended to filenames of XFAIL_TESTS
that have been declared as programs in the same Makefile.
This is for consistency with the analogous change to TESTS in 1.10.
- The autoconf version check implemented by aclocal in aclocal.m4
(and new in Automake 1.10) is degraded to a warning. This helps
in the common case where the Autoconf versions used are compatible.
- Fix order of standard includes to again be `-I. -I$(srcdir)',
followed by directories containing config headers.
New in 1.10:
* Version requirements:
- Autoconf 2.60 or greater is required.
- Perl 5.6 or greater is required.
* Changes to aclocal:
- aclocal now also supports -Wmumble and -Wno-mumble options.
- `dirlist' entries (for the aclocal search path) may use shell
wildcards such as `*', `?', or `[...]'.
- aclocal supports an --install option that will cause system-wide
third-party macros to be installed in the local directory
specified with the first -I flag. This option also uses #serial
lines in M4 files to upgrade local macros.
The new aclocal options --dry-run and --diff help to review changes
before they are installed.
- aclocal now outputs an autoconf version check in aclocal.m4 in
projects using automake.
For a few years, automake and aclocal have been calling autoconf
(or its underlying engine autom4te) to accurately retrieve the
data they need from configure.ac and its siblings. Doing so can
only work if all autotools use the same version of autoconf. For
instance a Makefile.in generated by automake for one version of
autoconf may stop working if configure is regenerated with another
version of autoconf, and vice versa.
This new version check ensures that the whole build system has
been generated using the same autoconf version.
* Support for new Autoconf macros:
- The new AC_REQUIRE_AUX_FILE Autoconf macro is supported.
- If `subdir-objects' is set, and AC_CONFIG_LIBOBJ_DIR is specified,
$(LIBOBJS), $(LTLIBOBJS), $(ALLOCA), and $(LTALLOCA) can be used
in different directories. However, only one instance of such a
library objects directory is supported.
* Change to Libtool support:
- Libtool generic flags (those that go before the --mode=MODE option)
can be specified using AM_LIBTOOLFLAGS and target_LIBTOOLFLAGS.
* Yacc and Lex changes:
- The rebuild rules for distributed Yacc and Lex output will avoid
overwriting existing files if AM_MAINTAINER_MODE and maintainer-mode
is not enabled.
- ylwrap is now always used for lex and yacc source files,
regardless of whether there is more than one source per directory.
* Languages changes:
- Preprocessed assembler (*.S) compilation now honors CPPFLAGS,
AM_CPPFLAGS and per-target _CPPFLAGS, and supports dependency
tracking, unlike non-preprocessed assembler (*.s).
- subdir-object mode works now with Assembler. Automake assumes
that the compiler understands `-c -o'.
- Preprocessed assembler (*.S) compilation now also honors
$(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES).
- Improved support for Objective C:
- Autoconf's new AC_PROG_OBJC will enable automatic dependency tracking.
- A new section of the manual documents the support.
- New support for Unified Parallel C:
- AM_PROG_UPC looks for a UPC compiler.
- A new section of the manual documents the support.
- Per-target flags are now correctly handled in link rules.
For instance maude_CFLAGS correctly overrides AM_CFLAGS; likewise
for maude_LDFLAGS and AM_LDFLAGS. Previous versions bogusly
preferred AM_CFLAGS over maude_CFLAGS while linking, and they
used both AM_LDFLAGS and maude_LDFLAGS on the same link command.
The fix for compiler flags (i.e., using maude_CFLAGS instead of
AM_CFLAGS) should not hurt any package since that is how _CFLAGS
is expected to work (and actually works during compilation).
However using maude_LDFLAGS "instead of" AM_LDFLAGS rather than
"in addition to" breaks backward compatibility with older versions.
If your package used both variables, as in
AM_LDFLAGS = common flags
bin_PROGRAMS = a b c
a_LDFLAGS = more flags
...
and assumed *_LDFLAGS would sum up, you should rewrite it as
AM_LDFLAGS = common flags
bin_PROGRAMS = a b c
a_LDFLAGS = $(AM_LDFLAGS) more flags
...
This new behavior of *_LDFLAGS is more coherent with other
per-target variables, and the way *_LDFLAGS variables were
considered internally.
* New installation targets:
- New targets mandated by GNU Coding Standards:
install-dvi
install-html
install-ps
install-pdf
By default they will only install Texinfo manuals.
You can customize them with *-local variants:
install-dvi-local
install-html-local
install-ps-local
install-pdf-local
- The undocumented recursive target `uninstall-info' no longer exists.
(`uninstall' is in charge of removing all possible documentation
flavors, including optional formats such as dvi, ps, or info even
when `no-installinfo' is used.)
* Miscellaneous changes:
- Automake no longer complains if input files for AC_CONFIG_FILES
are specified using shell variables.
- clean, distribution, or rebuild rules are normally disabled for
inputs and outputs of AC_CONFIG_FILES, AC_CONFIG_HEADERS, and
AC_CONFIG_LINK specified using shell variables. However, if these
variables are used as ${VAR}, and AC_SUBSTed, then Automake will
be able to output rules anyway.
(See the Automake documentation for AC_CONFIG_FILES.)
- $(EXEEXT) is automatically appended to filenames of TESTS
that have been declared as programs in the same Makefile.
This is mostly useful when some check_PROGRAMS are listed in TESTS.
- `-Wportability' has finally been turned on by default for `gnu' and
`gnits' strictness. This means, automake will complain about %-rules
or $(GNU Make functions) unless you switch to `foreign' strictness or
use `-Wno-portability'.
- Automake now uses AC_PROG_MKDIR_P (new in Autoconf 2.60), and uses
$(MKDIR_P) instead of $(mkdir_p) to create directories. The
$(mkdir_p) variable is still defined (to the same value as
$(MKDIR_P)) but should be considered obsolete. If you are using
$(mkdir_p) in some of your rules, please plan to update them to
$(MKDIR_P) at some point.
- AM_C_PROTOTYPES and ansi2knr are now documented as being obsolete.
They still work in this release, but may be withdrawn in a future one.
- Inline compilation rules for gcc3-style dependency tracking are
more readable.
- Automake installs a "Hello World!" example package in $(docdir).
This example is used throughout the new "Autotools Introduction"
chapter of the manual.
run-time dependency (DEPENDS) on a tool is to append a ":run" modifier
to the tool name, e.g.,
USE_TOOLS+= perl:run
Tools without modifiers or with an explicit ":build" modifier will
cause build dependencies (BUILD_DEPENDS) on those tools to be added.
This makes the notation a bit more compact.
Bugs fixed in 1.9.6:
* Longstanding bugs:
- Correctly diagnose `#' comments following trailing backslash.
- Preserve backslashes preceding `##' lines.
- Preserve the order of items conditionally appended (+=) to variables.
- Fix support of installation directory names with spaces on systems
where mkinstalldirs cannot use `mkdir -p'.
- Avoid infinite loop in mdate-sh when TIME_STYLE is set.
- Do not output two definitions of SOURCES (an internal variable).
One was not formatted and could exceed the maximum line length of
some third-party tools (e.g., AIX 5.1 grep, breaking dependency
tracking).
- Do not empty info files when attempting to rebuild them without makeinfo.
- Be smarter when a Makefile.am references files in both "./dir" and
"dir": do not output two rules to create these directories.
(PR/461)
- Do not attempt to recover a missing *.elc file if it cannot be
created because emacs does not exist.
- Several aclocal fixes for issues occuring when configure.ac
includes some other m4 files explicitely with m4_include or m4_sinclude.
(PR/450)
- depcomp's cpp mode now understands preprocessors that output either
`#line 42 file' or `# 42 file'. (Only the latter was supported.)
* Other miscellaneous changes:
- Update the GPL, and the FSF postal address.
- Anticipate for python2.5 in AM_PATH_PYTHON.
- The manual should now compile without any warning from TeX.
Bugs fixed in 1.9.5:
* Longstanding bugs:
- All m4 files have been relicensed under an all-permissive license.
Previously they used a GPL license, and an all-permissive license
was prepended when they were copied into aclocal.m4, leading to
some confusion.
- aclocal now recognizes AU_ALIAS.
- Improve support for `make -k', it didn't work on cygwin.
- Fix the definition of FCLINK for preprocessed Fortran.
- Fixes for auxiliary scripts:
- depcomp's tru64 mode did not work while compiling libtool libraries
with static libraries disabled.
- mdate-sh now copes with Darwin's non-standard ls output.
- missing properly emulates makeinfo when neither -o nor
@setfilename are used.
- Don't output long variable definitions. Automake normally tries
to output variable definitions as they are input, bug very long
lines in Makefile.in can break some sed and make implementations
with a limited line length. If a line in a variable definition is
longer than 1000 characters, automake will wrap the definition
around 80 columns. Other definitions are still output untouched.
* Noteworthy manual updates:
- Hard-Coded Install Paths: New FAQ entry.
- How the Linker is Chosen: Rewritten.
Bugs fixed in 1.9.4:
* Longstanding bugs:
- Portability nits in install-sh and mdata-sh.
- Don't let `make install' fails if a _JAVA primary becomes empty
because of conditionals.
- Do not confuse CHANGELOG with ChangeLog on case-insensitive
case-preserving file systems (likewise for all automatically
distributed files).
- Do not embed $DESTDIR in Python's byte-code files.
- Work around programs that read stdin when checking for --version
and --help options (when the `std-options' is used).
- Fix AM_PATH_PYTHON to correctly define PYTHON as `:' when no minimum
version was supplied and no interpreter is found.
* Noteworthy manual updates:
- Conditional Subdirectories: more comments about non-distributed
subdirectories.
- Flag Variables Ordering: new FAQ entry.
- Per-Object Flags: new FAQ entry.
agc
wiz
asau
bsiegert
dholland
sbd
joerg
jlam
rillig
tv