* master:
tests: portability fixes in tests on amhello examples
remake: add test ensuring that slower remakes don't hang
remake: fix outdated comment in configure.am
* maint:
tests: portability fixes in tests on amhello examples
remake: add test ensuring that slower remakes don't hang
remake: fix outdated comment in configure.am
tests: portability fixes in tests on amhello examples
* tests/amhello-binpkg.test: Don't use "tar xzf too.tag.gz" to
extract a gzip-compressed tarball, that's unportable to some
tar implementations; use the "gzip -dc fo.tar.gz | tar xf -"
idiom instead.
* tests/amhello-cflags.test: Likewise.
* tests/amhello-cross-compile.test: Likewise.
docs, tests: synchronize examples from docs to tests
* tests/README (Writing test cases): Give suggestions on how to
keep test cases and examples in the documentation synchronized.
* doc/automake.texi: Improve or fix existing testcase-referencing
comments, and add many new ones.
* HACKING (Administrivia): Suggest to test complex examples and
idioms from the manual.
* tests/specflg8.test: Improve synchronization with the example
in the manual.
* tests/output11.test:Likewise.
* tests/txinfo21.test:Likewise.
* tests/interp.test: Likewise. Since we are at it, and enable
the `errexit' shell flag, do related changes, and add trailing
`:'command.
* tests/amhello-cflags.test: New test.
* tests/amhello-cross-compile.test: Likewise.
* tests/amhello-binpkg.test: Likewise.
* tests/tests-environment-backcompat: Likewise.
* tests/parallel-tests-log-compiler-example.test: Likewise.
* tests/Makefile.am (TESTS): Update.
coverage: new test on parallel-tests TESTS runtime overriding
* tests/parallel-tests-cmdline-override.test: New test, check that
we can use indirections when overriding TESTS and TEST_LOGS from
the command line.
* tests/Makefile.am (TESTS): Update.
* master:
cosmetics: fix typos in recent ChangeLog entries
docs: explain why AM_TESTS_ENVIRONMENT must be semicolon-terminated
docs: fix unportable example of AM_TESTS_ENVIRONMENT usage
docs: fix unportable example of AM_TESTS_ENVIRONMENT usage
* doc/automake.texi (Simple Tests using parallel-tests): The
old example on AM_TESTS_ENVIRONMENT relied on unportable shell
features, and in particular didn't work with various Korn
Shells (see also commit `v1.11-925-g29ca903'). Give another
example, simpler this time, but still inspired to real-world
usage (the GNU coreutils testsuite).
* master:
Revert "docs: parallel-tests is not experimental anymore"
parallel-tests: stricter checks on DISABLE_HARD_ERRORS support
docs: parallel-tests is not experimental anymore
check: document and test $(TEST_SUITE_LOG) overriding
* maint:
Revert "docs: parallel-tests is not experimental anymore"
parallel-tests: stricter checks on DISABLE_HARD_ERRORS support
docs: parallel-tests is not experimental anymore
check: document and test $(TEST_SUITE_LOG) overriding
* parallel-tests-maint:
Revert "docs: parallel-tests is not experimental anymore"
parallel-tests: stricter checks on DISABLE_HARD_ERRORS support
docs: parallel-tests is not experimental anymore
check: document and test $(TEST_SUITE_LOG) overriding
* master:
docs: avoid a footnote, some related rewordings and improvements
docs: minor cosmetic fixes
help: improve text about automatically-distributed files
refactor: split 'usage' subroutine in automake
tests: fix bug in 'autodist.test'
maintcheck: extend 'sc_tests_plain_*' checks
tests: interactions between TESTS_ENVIRONMENT and LOG_COMPILER
docs: avoid a footnote, some related rewordings and improvements
* doc/automake.texi (Dist): Reword the part about automatically
distributed files to avoid a footnote. Since we are at it, extend
a bit, and add an example and a reference to a relevant test case.
It turned out that the current work to introduce TAP and SubUnit
support in Automake-generated testsuite harnesses will probably
require the introduction of slight incompatibilities in the
'parallel-tests' behaviour, starting from release 1.12 onward.
So it's advisable to continue to characterize the 'parallel-tests'
support as "experimental" in maintenance release 1.11.2.
* doc/automake.texi: Break few overly long lines, throughout the
file.
("Simple Tests"): Move @vindex for XFAIL_TESTS to the correct
position, i.e., before and not after the paragraph where it is
introduced.
("Options" @item ansi2knr): Use @pxref instead of @xref. This
fixes a texinfo warning.
("Other things Automake recognizes" @item AM_C_PROTOTYPES): Use
@pxref instead of @ref.
help: improve text about automatically-distributed files
This change fixes automake bug#7819.
* automake.in (usage): Distinguish between files that are always
automatically distributed when found, and those which are only
"under certain conditions".
* doc/automake.texi (Basics of Distribution): Update accordingly.
* tests/autodist-subdir.test: Update.
* tests/autodist-no-duplicate.test: Likewise.
* tests/autodist.test: Likewise.
(configure.in): Remove useless call to AM_MAINTAINER_MODE.
* automake.in (print_autodist_files): New subroutine,
extracted from ...
(usage): ... this, which now uses it.
* tests/autodist-no-duplicate.test: New test.
* tests/Makefile.am (TESTS): Update.
tests: interactions between TESTS_ENVIRONMENT and LOG_COMPILER
* tests/tests-environment-and-log-compiler.test: New test,
checking that we can use variables and functions set by
TESTS_ENVIRONMENT and AM_TESTS_ENVIRONMENT in LOG_COMPILER
and LOG_FLAGS (for tests both with and without registered
extensions).
* tests/Makefile.am (TESTS): Update.
* tests/depmod-data.test: Use creative quoting to avoid
spuriously triggering the `sc_tests_Exit_not_exit' maintainer
check. Commit `v1.11-900-g3453b8e' attempted to fix it, but
succeeded only partially.
* tests/cond33.test: Fix header comments, not to reference
obsolescent make variable `$(mkdir_p)'.
* tests/cond4.test: Rewrite "$MAKE exp=..." as "exp=.. $MAKE -e",
to please `sc_tests_overriding_macros_on_cmdline'
* tests/cond19.test: Likewise.
* tests/cond32.test: Likewise.
* tests/add-missing.test: Use "AUTOMAKE_fails" instead of
"$AUTOMAKE ... && Exit 1", to please `sc_tests_automake_fails'.
* Makefile.am (sc_tests_plain_aclocal, sc_tests_plain_perl,
sc_tests_plain_autoconf, sc_tests_plain_automake,
sc_tests_plain_autoupate): Be stricter in matching an erroneous
literal command, i.e., `aclocal', `automake', `perl', etc.
* maint:
check: don't use multi-line coloring for the report
ansi2knr: deprecate, it will go away in the next major release
docs: primary/prefix combination "pkglib_PROGRAMS" is now invalid
docs: replace obsolete @vindex entry with a useful one
docs: AM_DISTCHECK_CONFIGURE_FLAGS is for corner cases
* master:
check: don't use multi-line coloring for the report
ansi2knr: deprecate, it will go away in the next major release
docs: primary/prefix combination "pkglib_PROGRAMS" is now invalid
docs: replace obsolete @vindex entry with a useful one
tests: more checks on portable fd redirection in TESTS_ENVIRONMENT
docs: AM_DISTCHECK_CONFIGURE_FLAGS is for corner cases
* maint:
check: don't use multi-line coloring for the report
ansi2knr: deprecate, it will go away in the next major release
docs: primary/prefix combination "pkglib_PROGRAMS" is now invalid
docs: replace obsolete @vindex entry with a useful one
docs: AM_DISTCHECK_CONFIGURE_FLAGS is for corner cases
Bert Wesarg [Fri, 17 Jun 2011 19:59:52 +0000 (21:59 +0200)]
check: don't use multi-line coloring for the report
"less -R" can't handle multi-line coloring as it is done for the
check reports of the serial and parallel testsuite, because of
performance reasons. Thus, color each line of the check report
by its own.
* lib/am/check.am (am__text_box): Accept colors for lines, and
color each line by its own.
[%?PARALLEL_TESTS%] $(TEST_SUITE_LOG): Let am__text_box handle
the line coloring.
[!%?PARALLEL_TESTS%] $(check-TESTS): Color each report line by
its own.
* THANKS: Update.
ansi2knr: deprecate, it will go away in the next major release
* doc/automake.texi: Loudly and repeatedly state that the old
de-ANSI-fication features are now deprecated and will be removed
in the next major Automake release. Other related adjustments.
* lib/Automake/Options.pm (_process_option_list ): Give a warning
in the `obsolete' category when the `ansi2knr' option is used.
* m4/protos.m4 (AM_C_PROTOTYPES): Deprecate this macro: a warning
in the `obsolete' category will be emitted it if is used.
* tests/ansi2knr-deprecation.test: New test.
* tests/Makefile.am (TESTS): Update.
* tests/ansi.test: Adjust, by calling autoconf and/or automake
with the `-Wno-obsolete' flag.
* tests/ansi10.test: Likewise.
* tests/ansi2.test: Likewise.
* tests/ansi3.test: Likewise.
* tests/ansi3b.test: Likewise.
* tests/ansi4.test: Likewise.
* tests/ansi5.test: Likewise.
* tests/ansi6.test: Likewise.
* tests/ansi7.test: Likewise.
* tests/ansi8.test: Likewise.
* tests/ansi9.test: Likewise.
* tests/cxxansi.test: Likewise.
* tests/libobj8.test: Likewise.
* NEWS: Update about the future planned backward-incompatibility
due to the removal of de-ANSI-fication feature.
docs: primary/prefix combination "pkglib_PROGRAMS" is now invalid
* doc/automake.texi (Program Sources): pkglib_PROGRAMS is not a
valid combination anymore, so don't document it. Inconsistency
introduced in commit `v1.11-373-g9ca6326'.
tests: more checks on portable fd redirection in TESTS_ENVIRONMENT
* tests/tests-environment-fd-redirect.test: Extend by also using
a perl script among the tests. Run the test shell script with
the `errexit' flag active. Export `VERBOSE' to yes when running
"make check", to give more debugging information in case of
failures. Look for a Korn Shell also in `/usr/bin', not on only
in `/bin'.
docs: AM_DISTCHECK_CONFIGURE_FLAGS is for corner cases
* doc/automake.texi (Checking the Distribution): Explain that the
developers should take care of making their code buildable without
requiring any special configure options, so that in general
AM_DISTCHECK_CONFIGURE_FLAGS shouldn't be used. Give an example
of where its use is legitimate.
* maint:
news: update w.r.t. introduction of AM_DISTCHECK_CONFIGURE_FLAGS
tests: optimize tests on primary/prefix mismatch for speed
Warnings about primary/prefix mismatch fixed and extended.
maintcheck: DISTCHECK_CONFIGURE_FLAGS can be defined on make cmdline
distcheck: add support for AM_DISTCHECK_CONFIGURE_FLAGS
docs: better documentation for silent make rules
* master:
tests: check portable fd redirection in TESTS_ENVIRONMENT
news: update w.r.t. introduction of AM_DISTCHECK_CONFIGURE_FLAGS
tests: optimize tests on primary/prefix mismatch for speed
Warnings about primary/prefix mismatch fixed and extended.
maintcheck: DISTCHECK_CONFIGURE_FLAGS can be defined on make cmdline
distcheck: add support for AM_DISTCHECK_CONFIGURE_FLAGS
docs: better documentation for silent make rules
* maint:
news: update w.r.t. introduction of AM_DISTCHECK_CONFIGURE_FLAGS
tests: optimize tests on primary/prefix mismatch for speed
Warnings about primary/prefix mismatch fixed and extended.
maintcheck: DISTCHECK_CONFIGURE_FLAGS can be defined on make cmdline
distcheck: add support for AM_DISTCHECK_CONFIGURE_FLAGS
docs: better documentation for silent make rules
tests: make test 'self-check-reexec.test' more portable
* tests/self-check-reexec.test: Rewrite not to require a Korn
Shell able to grok variable expansions such as `${.sh.version}';
Solaris 10 /bin/ksh fails on this for example. Instead, just
require bash and a shell that is not bash.
tests: check portable fd redirection in TESTS_ENVIRONMENT
* tests/tests-environment-fd-redirect.test: New test.
* tests/Makefile.am (TESTS): Update.
Motivated by coreutils bug#8846:
<http://debbugs.gnu.org/cgi/bugreport.cgi?bug=8846>
See also following CC:ed thread on bug-autoconf list:
<http://lists.gnu.org/archive/html/bug-autoconf/2011-06/msg00002.html>
* tests/cond31.test ($required): Remove `cc', it's not really
needed.
* tests/confh.test: Call autoheader too. The lack of this call
wasn't causing spurious failures because, when make was called,
the automatic remake rules somehow ended up invoking it on our
behalf.
* tests/fn99subdir.test: Use $subdirname throughout, instead of
${subdirname}, for consistency with the rest of the testsuite.
Avoid an unnecessary subshell, which could also cause spurious
passes, being guarded by a trailing `|| Exit 1', which neutralize
the `errexit' flag. Remove an unnecessary `|| Exit 1' guard.
* tests/insh2.test: Rewrite to avoid hackish Makefile.in munging,
and to also run configure and make.
test defs: fix ksh-related portability bug in warning messages
Running "make check" normally prints a diagnostic to the outermost
stderr (usually a tty) to explain why a test is skipped, thus
giving better and faster feedback to the user. It used to do
so by redirecting file descriptor 9 to stderr (via "exec 9>&2")
before invoking the test scripts, which then would write any skip
explanation to file descriptor 9 via the `skip_' function defined
in `tests/defs'.
However, various Korn Shells (at least Solaris 10's /bin/ksh and
Debian GNU/Linux's /bin/ksh) and the HP-UX's /bin/sh close open
file descriptors > 2 upon an `exec' system call; thus the effects
of "exec 9>&2" are cancelled upon fork-and-exec, so we would get
a "Bad file number" diagnostic and no skip explanation with those
shells.
The present change remedies this situation.
* tests/Makefile.am (AM_TESTS_ENVIRONMENT): Redirect more portably,
via a trailing "9>&2", rather than the prior "exec 9>&2; ...". Add
explanatory comments.
* tests/defs (stderr_fileno_): Update the advice in comments.
Based on commit v8.12-82-g6b68745 "tests: accommodate HP-UX and
ksh-derived shells" in GNU coreutils.
Further references, with lots of discussion:
<http://lists.gnu.org/archive/html/bug-autoconf/2011-06/msg00002.html>
<http://thread.gmane.org/gmane.comp.gnu.coreutils.bugs/22488>
<http://debbugs.gnu.org/cgi/bugreport.cgi?bug=8846>
Merge branch 'am-distcheck-configure-flags' into maint
* am-distcheck-configure-flags:
maintcheck: DISTCHECK_CONFIGURE_FLAGS can be defined on make cmdline
distcheck: add support for AM_DISTCHECK_CONFIGURE_FLAGS
tests: extend tests on `--add-missing' and `--copy' a bit
* tests/add-missing.test: Fix typo in heading comments. Try with
another testcase that install many (but not all) the auxiliary
scripts at once, and uses non-standard (but valid and documented)
setups (e.g., defining YACC in Makefile.am instead of calling
AC_PROG_YACC from configure.in).
* tests/copy.test: Reference `add-missing.test' in heading
comments. Try few more test scenarios.
tests: remove 'test_prefer_config_shell' from the environment
Since commit `v1.11-910-g7df1a9b', the once user-overridable
variable `$test_prefer_config_shell' has become an internal
detail, and the test scripts now complain and bail out if it is
set in the environment.
* tests/Makefile.am (AM_TESTS_ENVIRONMENT): Unset the variable
`test_prefer_config_shell' if it is set in the environment.
tests: autogenerate list of wrapped tests for `lib/' shell scripts
* tests/gen-config-shell-tests: New script, generates distributed
makefile snippet `tests/config-shell-tests.am' to list all tests
that use the `get_shell_script' function, with names mangled to
use suffix `-w.shtst', in ...
* tests/Makefile.am (config_shell_tests): ... this macro, whose
definition has been consequently removed from Makefile.am.
(EXTRA_DIST): Distribute the new script.
($(srcdir)/config-shell-tests.am): Generate using the new script.
(include): Include the `config-shell-tests.am' fragment.
* bootstrap: Invoke `tests/gen-config-shell-tests' to generate
`tests/config-shell-tests.am'.
* tests/.gitignore: Ignore `config-shell-tests.am'.
* tests/gen-parallel-tests: Fixlet in heading comments.
* tests/mdate5.test: Fetch the `mdate-sh' script using the
`get_shell_script' function, and run it directly instead of
using `$SHELL'.
* tests/mdate6.test: Likewise. Since we are at it, make checks
on the `mdate-sh' output stricter, remove now unneeded calls to
aclocal and automake and creation/extension of `configure.in',
`Makefile.am' and `textutils.tex' files, and add a trailing `:'
command.
* tests/Makefile.am (config_shell_tests): Add `mdate5-w.shtst'
and `mdate6-w.shtst'.
tests: optimize tests on primary/prefix mismatch for speed
* tests/primary-prefix-invalid-couples.test: Partial rewrite, in
order to use just a single automake invocation rather than one
invocation for each invalid primary/prefix couple. This improves
the test script execution time by an order of magnitude.
Since we are at it, throw in some other improvements to avoid
unrelated automake warnings and failures that could potentially
cause false positives w.r.t. the automake exit status.
Warnings about primary/prefix mismatch fixed and extended.
* automake.in (%standard_prefix): Add `doc' and `locale'.
Rename `pkgdatadir' to `pkgdata'. Similarly for`pkglibdir',
`pkgincludedir' and `pkglibexecdir'.
(handle_programs): List `pkglibexec', not `pkglib', among the
prefixes valid for the `PROGRAMS' primary.
(handle_data): List also `doc' among the prefixes valid for
the `DATA' primary. This is required by automake's own build
system.
* tests/dirforbid.test: Test removed, superseded by ...
* tests/primary-prefix-invalid-couples.test: ... this new test.
* tests/primary-prefix-valid-couples.test: New test.
* tests/primary-prefix-documented-valid.test: Likewise.
* tests/primary-prefix-force-valid.test: Likewise.
* tests/java3.test: Adjusted, and extended a bit.
* tests/Makefile.am (TESTS): Updated.
* NEWS: Updated.
testsuite: use 'fatal_' and 'framework_failure_' for hard errors
* tests/defs (require_xsi): Use `fatal_', not `framework_failure',
to report an invalid usage.
* tests/remake-gnulib-remove-header.test: Prefer using `fatal_'
with a proper error message over a direct call to `Exit 99'.
* tests/pr8365-remake-timing.test: Likewise.
* tests/cygnus-imply-foreign.test: Likewise.
* tests/missing6.test: Likewise.
* tests/cond8.test: Likewise.
* tests/cond33.test: Likewise.
* tests/python-virtualenv.test: Prefer using `framework_failure_'
with a proper error message over a direct call to `Exit 99'.
* tests/instspc-tests.sh: Prefer using `framework_failure_' and
`fatal_' over direct calls to `Exit 99'.
(fatal_): Define this (which is a simplified version of the one
in `tests/defs') for early uses (i.e., before `tests/defs'
gets sourced).
* tests/depmode-tests.sh: Likewise. Also, simplify the
'get_depmodes' function and calls to it accordingly.
maintcheck: DISTCHECK_CONFIGURE_FLAGS can be defined on make cmdline
* Makefile.am (sc_tests_overriding_macros_on_cmdline): It's now
acceptable that the test scripts override DISTCHECK_CONFIGURE_FLAGS
on the make command line. Update comments accordingly. Since we
are at it, make the relevant grepping rules slightly tighter.
distcheck: add support for AM_DISTCHECK_CONFIGURE_FLAGS
* doc/automake.texi (Checking the Distribution): Suggest to use
AM_DISTCHECK_CONFIGURE_FLAGS, not DISTCHECK_CONFIGURE_FLAGS, to
define (in the top-level Makefile.am) extra flags to be passed
to configure at "make distcheck" time; DISTCHECK_CONFIGURE_FLAGS
should be reserved for the user. Add proper `@vindex' directive.
Document that AM_DISTCHECK_CONFIGURE_FLAGS is not honoured in a
subpackage Makefile.am, but the flags in it are passed down to
the configure script of the subpackage.
* lib/am/distdir.am (distcheck): Also pass the flags in
$(AM_DISTCHECK_CONFIGURE_FLAGS) to the configure invocation.
Update comments.
* tests/defs.in.test (AM_DISTCHECK_CONFIGURE_FLAGS,
DISTCHECK_CONFIGURE_FLAGS): Unset in case they are exported in
the environment, they might improperly influence our testsuite.
* tests/distcheck-configure-flags.test: New test.
* tests/distcheck-configure-flags-am.test: Likewise.
* tests/distcheck-configure-flags-subpkg.test: Likewise.
* distcheck-hook.test: Likewise.
* distcheck-hook2.test: Likewise.
* tests/Makefile.am (TESTS): Update.
* doc/automake.texi (Options): Detailed description of the
automake option `silent-rules' moved from here ...
(Silent Make): ... into this new chapter, expanded, improved,
and subdivided into ...
(Make verbosity, Tricks For Silencing Make,
Automake silent-rules Option): ... these new sections.
(@menu, @detailmenu): Update.
* tests/silent-configsite.test: New test, checking that the
user can control default mode of silent-rules from config.site,
as is documented in the manual.
* tests/Makefile.am (TESTS): Updated.
Before this patch, the only way offered by tests/defs to
properly signal a hard error was the `framework_failure_'
function. But the error message issued by that function,
as its name would suggest, refers to a set-up failure in the
testsuite, while hard errors can obviously also be due to
other reasons. The best way to fix this inconsistency is to
introduce a new function with a more general error message.
Inspired by a recent similar change to Gnulib's tests/init.sh.
* tests/defs.in (fatal_): New function.
* tests/README (Section "Writing test cases" subsection "Do"):
Suggest the use of `fatal_', not of `framework_failure_', for
generic hard errors. The latter should be reserved for "real"
set-up failures.
* tests/self-check-me.test: Check that `$me' can be overridden
before sourcing ./defs, with or without sourcing ./defs-static
beforehand, and that this override is honored. Update heading
comments.
tests: `lib/' shell scripts transparently tested also with $SHELL
With the previous commit, the user could prefer the use of $SHELL
over /bin/sh in some tests checking the Automake-provided shell
scripts by manually exporting `test_prefer_config_shell' to "yes"
in the environment. With this commit, we ensure that such tests
*always* and *transparently* run using both $SHELL and /bin/sh
to execute the checked scripts. The `test_prefer_config_shell'
variable becomes an internal detail, and is no more meant to be
manually defined or overridden.
* tests/defs-static.in: Check that `test_prefer_config_shell' is
not exported in the environment. Error out if this is the case.
* tests/config-shell-tests.sh: New file, driver script to run
checks on the shell scripts in `lib/' using the $SHELL determined
at configure time instead of the default system shell /bin/sh.
* Makefile.am (TESTS_EXTENSIONS): Add `.shtst'.
(SHTST_LOG_COMPILER): Define, it calls `config-shell-tests.sh'.
(config_shell_tests): Define to a list of tests that wraps other
`*.test' tests using `config-shell-tests.sh'.
($(config_shell_tests)): Dummy dependency declaration required
in order to have make actually produce expected log files from
the `.shtst.log' suffix rule.
(EXTRA_DIST): Distribute `config-shell-tests.sh'.
(TESTS): Add `$(config_shell_tests)'.
* tests/self-check-env-sanitize.test: Update, by checking that
`test_prefer_config_shell' isn't exported in the environment.
tests: can use also $SHELL to check shell scripts from `lib/'
* tests/ar-lib.test: If the variable `$test_prefer_config_shell'
is set to "yes", run the script under test with configure-time
determined $SHELL, rather than with /bin/sh.
The `$test_prefer_config_shell' variable defaults to empty, but
can be overridden at runtime by the user, thus allowing more
coverage.
* tests/compile.test: Likewise.
* tests/compile2.test: Likewise.
* tests/compile3.test: Likewise.
* tests/compile4.test: Likewise.
* tests/compile5.test: Likewise.
* tests/compile6.test: Likewise.
* tests/instsh2.test: Likewise.
* tests/instsh3.test: Likewise.
* tests/mkinst3.test: Likewise.
* tests/missing.test: Likewise.
* tests/missing2.test: Likewise.
* tests/missing3.test: Likewise.
* tests/missing5.test: Likewise.
* tests/defs (get_shell_script): New subroutine, factoring out
code common to the tests above.
(xsi-lib-shell): If `$test_prefer_config_shell' is set to "yes",
check that $SHELL, not /bin/sh, supports XSI constructs, as we
expect the test will use $SHELL and not /bin/sh to run the
script being tested.
This change avoids potential spurious failures with tests using
the requirement 'xsi-shell' to mean that they want */bin/sh* (not
$SHELL) to be XSI-conforming. This idiom used to work before
commit `v1.11-874-g1321be7' (as back then the test scripts were
unconditionally run with /bin/sh), but has become inconsistent
now that the test scripts re-execute themselves with configure
determined $SHELL.
The described spurious failures have already occurred in practice,
for examples on Solaris systems which also had GNU Bash installed.
From a suggestion by Peter Rosin. See discussion at:
<http://lists.gnu.org/archive/html/automake-patches/2011-06/msg00016.html>
* tests/defs (xsi-shell): Now check that $SHELL, rather than the
shell currently running the test script, is an XSI shell.
(xsi-bin-sh): New requirement, checking that /bin/sh (which can
differ from $SHELL) is an XSI shell.
(xsi-lib-shell): New requirement, checking that the shell that
should be used to test the Automake-provided scripts from `lib/'
is an XSI shell. For the moment, this is just an alias for
`xsi-bin-sh'.
(require_xsi): New subroutine, used to factor out code common to
the requirements above.
($xsi_shell_code): New variable, contains shell code supposed to
work only with XSI shells. Used by the new subroutine above.
* tests/ar-lib.test ($required): Require 'xsi-lib-shell' instead
of 'xsi-shell', since the script we test here is run with /bin/sh,
not with $SHELL.
* tests/compile3.test: Likewise.
* tests/compile6.test: Likewise.
* tests/auxdir2.test (configure.in): Close m4 quoting in the
argument to AC_CONFIG_AUX_DIR. Without this, aclocal fails with
"ERROR: end of file in string". This problem hasn't been exposed
by the testsuite before because this test is in XFAIL_TESTS, so
its failure went unnoticed, even if it was due to a wrong cause.
* maint:
maintcheck: fix some failures, extend some checks
automake, aclocal: honour configure-time AUTOCONF and AUTOM4TE
build: the user can override AUTOM4TE, AUTORECONF and AUTOUPDATE too
remake: behave better with non-GNU make in subdirectories
tests/README: fix example about `make -e' usage
Our ad-hoc usage of `tests/defs' in the testsuite's self tests
stopped working properly when we made the test scripts re-execute
themselves with the configure-time $SHELL. Fix this.
* tests/self-check-exit.test: Export `AM_TESTS_REEXEC' to "no"
before running the self tests. This fixes a spurious failure
present only when the test was run by hand.
projects / thirdparty / automake.git / log
