When built-in rules are disabled by adding -r to MAKEFLAGS in the
makefile, don't add suffix rules at all so that if suffixes are
added back via .SUFFIXES, the rules are still not there.
* src/main.c (main): Set default suffix rules after parsing makefiles.
* src/default.c (install_default_suffix_rules): Install a default
suffix rule only if there is no user defined rule.
* tests/scripts/features/suffixrules: Add tests.
Running regression tests on a single CPU system shows that if a child
exits while we're checking prerequisites we might not notice it, and
then we might block waiting for the next child to exit before we try
to do more work even though we could.
* src/remake.c (update_goal_chain): Remember the command_count before
we start checking prerequisites and if it is different when we restart
at the top of the goal chain then don't block.
The indentation shown in debug output was misleading. Increment the
depth when we are working on a prerequisite, not for the current
target.
* src/remake.c (check_dep): Increment depth only on recursion.
(update_file_1): Ditto. Don't show "Finished..." if prereqs are
still running.
* src/implicit.c (pattern_search): Increment depth for the entire
search.
* tests/scripts/features/output-sync: Support debug output when run
by hand.
Some systems that support mkfifo() don't support the usage that GNU
make wants. Provide a way to force using "pipe" jobserver mode even
when mkfifo() is available.
* src/makeint.h (MK_OS_HURD): Define if we're on Hurd.
(JOBSERVER_USE_FIFO): Define if we have mkfifo() and we're NOT on
Hurd.
* src/main.c (main): Test JOBSERVER_USE_FIFO not HAVE_MKFIFO.
* src/posixos.c (jobserver_setup): Ditto.
* maintMakefile: Create a config check test for forcing "pipe" mode.
* tests/run_make_tests.pl: Show discovered FEATURES in verbose mode.
* configure.ac: Add AC_PROG_CXX to search for a C++ compiler.
(MAKE_CXX): Add the discovered compiler to config.h.
* src/default.c (default_variables): Use MAKE_CXX as the C++ compiler.
* src/variable.c (lookup_variable_for_file): New function to retrieve
a variable assignment in a file context.
* src/variable.h (lookup_variable_for_file): Declare it.
* src/job.c (construct_command_argv): Look up .SHELLFLAGS. If .POSIX
is set and we're using the default value, choose -c if we're ignoring
errors else choose -ec.
(construct_command_argv_internal): Ditto.
* tests/scripts/targets/POSIX: Add tests.
Allow z/OS customizations to apply to regex output matching.
* tests/scripts/features/output-sync: Use a regex for Terminated.
* tests/scripts/features/temp_stdin: Ditto.
* NEWS: Update the date.
* README.git: Fix a typo.
* maintMakefile (check-alt-config): Report success.
(upload-*): Use the gnulib gnupload script for uploads.
The other issues related to whitespace reported in this bug are
not addressed by this change.
* src/functions.c (func_file): Strip whitespace from the start and
end of the filename provided to the $(file ...) function.
* tests/scripts/functions/file: Add tests for this.
* tests/test_driver.pl: Use 3-arg version of open().
If -e was given we weren't expanding MAKEFLAGS before passing it
through the environment to jobs: we don't expand variables we
receive from the environment and when -e is given we set the
origin of MAKEFLAGS to "environment override". Check for MAKEFLAGS
specifically, which seems like a hack but I don't have a better
idea offhand.
* src/main.c (main): Drive-by: use o_default for MAKEOVERRIDES.
* src/variable.c (target_environment): Always expand MAKEFLAGS
regardless of the origin type.
* tests/scripts/options/dash-e: Create a test.
Original patches provided by Igor Todorovski <itodorov@ca.ibm.com>
Reworked by Paul Smith <psmith@gnu.org>.
Thanks to IBM for providing a test system.
* NEWS: Announce support.
* AUTHORS: Ditto.
* README.zOS: Provide details on building GNU Make on z/OS.
* build.sh (get_mk_var): z/OS sh has a strange bug which causes it to
generate extra lines of output: rework the function to print output
as we compute it instead of collecting it into a variable, which
works around this bug.
* src/makeint.h: Declare MK_OS_ZOS if we're building for z/OS.
* src/arscan.c: Don't include <ar.h> on z/OS.
* src/job.c: We can't change environ in ASCII mode on z/OS.
* src/main.c: Ditto. Also we can't use pselect() on z/OS.
* src/posixos.c: pselect() seems to hang on z/OS: don't use it.
* tests/run_make_tests.pl: Handle different exit codes on z/OS.
* tests/test_driver.pl: Preserve some special z/OS env.vars.
Add special checks to output comparisons when on z/OS.
* tests/scripts/features/archives: Don't validate names. Don't
try to compile empty files as IBM compilers complain.
* tests/scripts/features/shell_assignment: Fix octal value of #.
* tests/scripts/features/temp_stdin: Don't print "term".
* tests/scripts/functions/shell: Handle shell exit codes.
* tests/scripts/targets/ONESHELL: Ditto.
* tests/scripts/targets/POSIX: sh -x prints differently.
* tests/scripts/variables/SHELL: Ditto.
Put declarations for missing functions which we create in src/misc.c
into a file which is included by config.h via AH_BOTTOM(). This
ensures those prototypes are available, even in files added to
lib/... from gnulib.
* src/mkcustom.h: Add a new file with function declarations.
* configure.ac: Include src/mkcustom.h in config.h with AH_BOTTOM()
* Makefile.am: Add the header to the SRCS list.
* src/makeint.h: Remove content that we added to src/mkcustom.h.
* src/config.ami: Add #include "mkcustom.h" to specialized config.h.
* src/config.h-vms: Ditto.
* src/config.h.W32: Ditto.
* src/configh.dos: Ditto.
Various code uses str{,n}casecmp() so include <strings.h>, if it
exists, everywhere.
* configure.ac: Check for <strings.h> explicitly.
* src/makeint.h: Include it if HAVE_STRINGS_H, for str{,n}casecmp().
* src/job.c: Remove include of <strings.h>.
* src/main.c: Ditto.
Reported by Dmitry Goncharov <dgoncharov@users.sf.net>
* src/variable.c (struct defined_vars): Create a struct that holds the
name and length of each variable name.
(warn_undefined): Check the lengths before comparing the contents.
* tests/scripts/options/warn-undefined-variables: Add a test.
Compare the timestamp of the object file (if it exists) with the
archived object and if the object file is newer, ensure it's updated
in the archive.
* NEWS: Announce the new capability.
* doc/make.texi (Dangers When Using Archives): Explain how to enable
parallel builds with archives.
* src/remake.c (f_mtime): For archive element files check the mod
time of the object file (if it exists) against the archive object
(if it exists).
* tests/scripts/features/archives: Add tests for this capability.
If a parent target has an exported variable that is private, don't
export it in child targets.
* NEWS: Mention this change.
* src/variable.c (target_environment): Ignore private inherited
variables.
* tests/thelp.pl: Add a new "env" operation to show env.var. values.
* tests/scripts/variables/private: Verify this new behavior.
I looked again at trying to use the latest gnulib implemenentations
of GNU glob and fnmatch, and the effort required to extract them
from gnulib and make them portable to systems which don't support
configure is simply far too daunting for me. However it's clear
that the previous implementations are growing too long on the tooth
to continue to be used without some maintenance, so perform some
upkeep on them.
- Remove support for pre-ANSI function definitions.
- Remove the obsolete "register" keyword.
- Assume standard ISO C90/C99 header file support.
- Assume standard ISO C "void" and "const" support.
- Avoid symbols prefixed with "__" as they're reserved.
* maintMakefile: Add a rule to verify lib has the latest content.
* src/dir.c: Use void* not __ptr_t which was removed.
* gl/lib/glob.c: See above.
* gl/lib/fnmatch.in.h: See above.
* gl/lib/glob.in.h: See above.
* gl/lib/fnmatch.c: See above. Remove __strchrnul(): it is not
checked anywhere and is only used in one place anyway.
* src/dir.c (clear_directory_contents): We use ht_vec to mark when
we have initialized the hash_table: don't free if it's NULL.
(everywhere): Use NULL instead of 0 when working with pointers.
* bootstrap.conf: Request the latest gnulib stable branch.
* README.git: Recommend the argument form of autogen/autopull.
* bootstrap: Import the latest version of bootstrap.
* bootstrap-funclib.sh: Ditto.
* autogen.sh: Ditto.
* autopull.sh: Ditto.
The POSIX standard only requires false(1) to return a "non-zero" exit
code; almost all systems return 1 but some (Solaris!!!) return 255 or
possibly even other values. Use our helper "fail" instead.
* tests/thelp.pl: Have the "fail" command obey -q.
* tests/scripts/features/parallelism: Helper -q no longer prints fail.
* tests/scripts/targets/POSIX: Replace false with #HELPER# -q fail 1.
* tests/scripts/variables/MAKEFLAGS: Ditto.
* tests/scripts/variables/SHELL: Ditto.
Using backslashes in a sed command line is tricky as different programs
use them differently as escape sequences. Eli Zaretskii points out
that Windows "echo" doesn't do any processing, so rework all our sed
invocations to use script files created by echo.
* bootstrap.bat: Use echo to create sed script files instead of -e.
* Basic.mk.template: Fix typo in the comments.
* .gitignore: Ignore any .sed scripts.
Don't generate undefined variable warnings for variables that are
internal / special to make and where the empty string is valid.
Rather than defining them to empty, which could introduce unwanted
behavior, keep a list of variable names which we should never warn
about.
* src/variable.h (warn_undefined): Convert the macro to a function.
* src/variable.c (defined_vars): Always "defined" variable names.
(warn_undefined): Implement as a function and check against the
defined_vars before generating a warning.
* src/read.c (read_all_makefiles): No need to reset warning flag.
* src/vpath.c (build_vpath_lists): Ditto.
* tests/scripts/options/warn-undefined-variables: Expand all the
pre-defined variables to ensure warnings are not generated.
