path: root/ChangeLog

2002-07-08  Paul D. Smith  <psmith@gnu.org>

	* maintMakefile: Add "update" targets for wget'ing the latest
	versions of various external files.  Taken from Makefile.maint in
	autoconf, etc.

	* dosbuild.bat: Somehow we got _extra_ ^M's.  Remove them.
	Reported by Eli Zaretskii  <eliz@is.elta.co.il>.

2002-07-07  Paul D. Smith  <psmith@gnu.org>

	* po/*.po: Remove.  We'll use wget to retrieve them at release
	time.

	* variable.c (do_variable_definition) [W32]: On W32 using cmd
	rather than a shell you get an exception.  Make sure we look up
	the variable.  Patch provided by Eli Zaretskii <eliz@is.elta.co.il>.

	* remake.c (notice_finished_file): Fix handling of -t flag.
	Patch provided by Henning Makholm <henning@makholm.net>.

	* implicit.c (pattern_search): Some systems apparently run short
	of stack space, and using alloca() in this function caused an
	overrun.  I modified it to use xmalloc() on the two variables
	which seemed like they might get large.  Bug #476.

	* main.c (print_version): Update copyright notice to conform with
	GNU standards.
	(print_usage): Update help output.

	* function.c (func_eval): Create a new make function, $(eval
	...).  Expand the arguments, put them into a buffer, then invoke
	eval_buffer() on the resulting string.
	(func_quote): Create a new function, $(quote VARNAME).  Inserts
	the value of the variable VARNAME without expanding it any
	further.

	* read.c (struct ebuffer): Change the linebuffer structure to an
	"eval buffer", which can be either a file or a buffer.
	(eval_makefile): Move the code in the old read_makefile() which
	located a makefile into here: create a struct ebuffer with that
	information.  Have it invoke the new function eval() with that
	ebuffer.
	(eval_buffer): Create a new function that creates a struct ebuffer
	that holds a string buffer instead of a file.  Have it invoke
	eval() with that ebuffer.
	(eval): New function that contains the guts of the old
	read_makefile() function: this function parses makefiles.  Obtains
	data to parse from the provided ebuffer.  Some modifications to
	make the flow of the function cleaner and clearer.  Still could
	use some work here...
	(do_define): Takes a struct ebuffer instead of a FILE*.  Read the
	contents of the define/endef variable from the ebuffer.
	(readstring): Read the next line from a string-style ebuffer.
	(readline): Read the next line from an ebuffer.  If it's a string
	ebuffer, invoke readstring().  If it's a FILE* ebuffer, read it
	from the file.

	* dep.h (eval_buffer): Prototype eval_buffer();

	* variable.c (do_variable_definition): Make sure that all
	non-target-specific variables are registered in the global set.
	If we're invoked from an $(eval ...) we might be inside a $(call
	...) or other function which has pushed a variable scope; we still
	want to define our variables from evaluated makefile code in the
	global scope.

2002-07-03  Greg McGary  <greg@mcgary.org>

	* dep.h (struct dep) [ignore_mtime]: New member.
	[changed]: convert to a bitfield.
	* implicit.c (pattern_search): Zero ignore_mtime.
	* main.c (main, handle_non_switch_argument): Likewise.
	* rule.c (convert_suffix_rule): Likewise.
	* read.c (read_all_makefiles, read_makefile, multi_glob): Likewise.
	(read_makefile): Parse '|' in prerequisite list.
        (uniquize_deps): Consider ignore_mtime when comparing deps.
	* remake.c (update_file_1, check_dep): Don't force remake for
	dependencies that have d->ignore_mtime.
	* commands.c (FILE_LIST_SEPARATOR): New constant.
        (set_file_variables): Don't include a
	prerequisite in $+, $^ or $? if d->ignore_mtime.
        Define $|.

2002-06-18  Paul D. Smith  <psmith@gnu.org>

	* make.texinfo: Updates for next revision.  New date/rev/etc.
	Recreate all Info menus.  Change license on the manual to the GNU
	Free Documentation License.  A number of typos.
	(Variables Simplify): Don't use "-" before it's defined.
	(Automatic Prerequisites): Rewrite the target example to work
	properly if the compile fails.  Remove incorrect comments about
	how "set -e" behaves.
	(Text Functions): Move the "word", "wordlist", "words", and
	"firstword" functions here, from "File Name Functions".
	* make-stds.texi: Update from latest GNU version.
	* fdl.texi: (created) Import the latest GNU version.

2002-06-06  Paul D. Smith  <psmith@gnu.org>

	* variable.c (do_variable_definition): New function: extract the
	part of try_variable_definition() that actually sets the value
	into a separate function.
	(try_variable_definition): Call do_variable_definition() after
	parsing the variable definition string.
	(define_variable_in_set): Make the name argument const.

	* variable.h (enum variable_flavor): Make public.
	(do_variable_definition): Create prototype.

	* read.c (read_all_makefiles): Create a new built-in variable,
	MAKEFILE_LIST.
	(read_makefile): Add each makefile read in to this variable value.

2002-05-18  Eli Zaretskii  <eliz@is.elta.co.il>

        * Makefile.DOS.template: Tweak according to changes in the
        distribution.  Add back the dependencies of *.o files.

        * configh.dos.template: Synchronize with config.h.in.

2002-05-09  Paul D. Smith  <psmith@gnu.org>

	* file.c (file_timestamp_now): Use K&R function declaration.

	* getloadavg.c (getloadavg): Merge setlocale() fix from sh-utils
	getloadavg.c.  Autoconf thinks QNX is SVR4-like, but it isn't, so
	#undef it.  Remove predefined setup of NLIST_STRUCT.  Decide
	whether to include nlist.h based on HAVE_NLIST_H.  Change obsolete
	NLIST_NAME_UNION to new HAVE_STRUCT_NLIST_N_UN_N_NAME.
	* configure.in (NLIST_STRUCT): Define this if we have nlist.h and
	nlist.n_name is a pointer rather than an array.

	* acinclude.m4 (make_FUNC_SETVBUF_REVERSED): Grab the latest
	version of AC_FUNC_SETVBUF_REVERSED from autoconf CVS.
	* configure.in: Use it instead of the old version.

	* main.c (main): Prefer setvbuf() to setlinebuf().

2002-05-08  Paul D. Smith  <psmith@gnu.org>

	* Makefile.am (make_LDADD): Add GETLOADAVG_LIBS.
	(loadavg_LDADD): Ditto.

2002-04-29  Paul D. Smith  <psmith@gnu.org>

	* expand.c (recursively_expand_for_file): Rename
	recursively_expand() to recursively_expand_for_file() and provide
	an extra argument, struct file.  If the argument is provided, set
	the variable scope to that of the file before expanding.
	* variable.h (recursively_expand): Make this a macro that invokes
	recursively_expand_for_file() with a NULL file pointer.
	* variable.c (target_environment): Call the renamed function and
	provide the current file context.
	Fixes Debian bug #144306.

2002-04-28  Paul D. Smith  <psmith@gnu.org>

	Allow $(call ...) user-defined variables to be self-referencing
	without throwing an error.  Allows implementation of transitive
	closures, among other possibly useful things.
	Requested by: Philip Guenther <guenther@sendmail.com>

	* variable.h (struct variable): Add a new field: exp_count, and
	new macros to hold its size and maximum value.
	(warn_undefined): Make this a macro.
	* variable.c (define_variable_in_set): Initialize it.
	* expand.c (recursively_expand): If we detect recursive expansion
	of a variable, check the exp_count field.  If it's greater than 0
	allow the recursion and decrement the count.
	(warn_undefined): Remove this (now a macro in variable.h).
	* function.c (func_call): Before we expand the user-defined
	function, modify its exp_count field to contain the maximum
	number of recursive calls we'll allow.  After the call, reset it
	to 0.

2002-04-21  Paul D. Smith  <psmith@gnu.org>

	Modified to use latest autoconf (2.53), automake (1.6.1), and
	gettext (0.11.1).  We're using gettext's new "external" support,
	to avoid including libintl source with GNU make.

	* README.cvs: New file.  Explain how to build GNU make from CVS.

	* configure.in: Modify checking for the system glob library.
	Use AC_EGREP_CPP instead of AC_TRY_CPP.  Remove the setting of
	GLOBDIR (we will always put "glob" in SUBDIRS, so automake
	etc. will manage it correctly).  Set an automake conditional
	USE_LOCAL_GLOB to decide whether to compile the glob library.

	* getloadavg.c (main): Include make.h in the "TEST" program to
	avoid warnings.

	* Makefile.am: Remove special rules for loadavg.  Replace them
	with Automake capabilities for building extra programs.

	* signame.c: This file does nothing if the system provide
	strsignal().  If not, it implements strsignal().  If the system
	doesn't define sys_siglist, then we make our own; otherwise we use
	the system version.
	* signame.h: Removed.

	* main.c (main): No need to invoke signame_init().  Update copyright.

	* ABOUT-NLS: Removed.
	* gettext.c: Removed.
	* gettext.h: Get a simplified copy from the gettext package.
	* po/*: Created.
	* i18n/*.po: Moved to po/.
	* i18n/: Removed.

	* config/*: Created.  Contains package configuration helper files.
	* config.guess, config.sub: Moved to config directory.

	* configure.in (AC_CONFIG_FILES): Add po/Makefile.in, config/Makefile.
	Rework to use new-style autoconf features.  Use the "external"
	mode for gettext.  Make the build.sh config file conditional on
	whether build.sh.in exists, to avoid autoconf errors.
	* acinclude.m4: Removed almost all macros as being obsolete.
	Rewrote remaining macros to use AC_DEFINE.
	* acconfig.h: Removed.

	* Makefile.am (EXTRA_DIST): Add config/config.rpath.  Use a
	conditional to handle customs support.  Remove special handling
	for i18n features.

2002-04-20  Paul D. Smith  <psmith@gnu.org>

	* function.c (func_call): Don't mark the argument variables $1,
	etc. as recursive.  They've already been fully expanded so
	there's no need to do it again, and doing so strips escaped $'s.
	Reported by Sebastian Glita <glseba@yahoo.com>.

	* remake.c (notice_finished_file): Walk through double-colon
	entries via the prev field, not the next field!
	Reported by Greg McGary <greg@mcgary.org>.

	* main.c (main): If the user specifies -q and asks for a specific
	target which is a makefile, we got an assert.  In that case it
	turns out we should continue normally instead.

	* i18n/de.po, i18n/fr.po: Installed an updated translation.

	* i18n/he.po: Installed a new translation.

2002-01-07  Paul D. Smith  <psmith@gnu.org>

	* i18n/es.po, i18n/ru.po: Installed an updated translation.

2001-12-04  Paul D. Smith  <psmith@gnu.org>

	* i18n/ja.po: Installed an updated translation.

2001-09-04  Paul D. Smith  <psmith@gnu.org>

	* i18n/da.po: Installed an updated translation.

2001-08-03  Paul D. Smith  <psmith@gnu.org>

	* i18n/fr.po: Installed an updated translation.
	Resolves Debian	bug #106720.

2001-06-13  Paul D. Smith  <psmith@gnu.org>

	* i18n/da.po, configure.in (ALL_LINGUAS): Installed a new
	translation.

2001-06-11  Paul D. Smith  <psmith@gnu.org>

	* i18n/ko.po: Installed a new translation.

2001-05-06  Paul D. Smith  <psmith@gnu.org>

	Modify the EINTR handling.

	* job.c (new_job): Reorganize the jobserver algorithm.  Reorder
	the way in which we manage the file descriptor/signal handler race
	trap to be more efficient.

2001-05-06  Paul Eggert  <eggert@twinsun.com>

	Restart almost all system calls that are interrupted, instead
	of worrying about EINTR.  The lone exception is the read() for
	job tokens.

	* configure.in (HAVE_SA_RESTART): New macro.
	(MAKE_JOBSERVER): Define to 1 only if HAVE_SA_RESTART.
	* main.c (main): Use SA_RESTART instead of the old,
	nonstandard SA_INTERRUPT.

	* configure.in (AC_CHECK_FUNCS): Add bsd_signal.
	* main.c (bsd_signal): New function or macro,
	if the implementation doesn't supply it.
	(The bsd_signal function will be in POSIX 1003.1-200x.)
	(HANDLESIG): Remove.
	(main, FATAL_SIG): Use bsd_signal instead of signal or HANDLESIG.

	* make.h (EINTR_SET): Remove.
	(SA_RESTART): New macro.

	* arscan.c (ar_member_touch): Don't worry about EINTR.
	* function.c (func_shell): Likewise.
	* job.c (reap_children, free_child, new_job): Likewise.
	* main.c (main): Likewise.
	* remake.c (touch_file, name_mtime): Likewise.

	* arscan.c (ar_member_touch): Fix bug uncovered by EINTR removal;
	if fstat failed with errno!=EINTR, the error was ignored.

	* job.c (set_child_handler_action_flags): New function.
	(new_job): Use it to temporarily clear the SIGCHLD action flags
	while reading the token.

2001-05-02  Paul D. Smith  <psmith@gnu.org>

	* job.c (start_job_command): Don't add define/endef per-line flags
	to the top-level flags setting.

2001-04-03  Paul D. Smith  <psmith@gnu.org>

	* arscan.c (VMS_get_member_info,ar_scan) [VMS]: VMS sets the low
	bit on error, so check for odd return values, not non-0 return
	values.
	(VMS_get_member_info): Calculate the timezone differences correctly.
	Reported by John Fowler <jfowler@nyx.net>.


2001-03-14  Paul D. Smith  <psmith@gnu.org>

	* variable.c (lookup_variable) [VMS]: Null-terminate the variable
	value before invoking define_variable().
	Reported by John Fowler <jfowler@nyx.net>.

2001-02-07  Paul D. Smith  <psmith@gnu.org>

	* read.c (record_target_var): If we reset the variable due to a
	command-line variable setting overriding it, turn off the "append"
	flag.

2001-01-17  Paul D. Smith  <psmith@gnu.org>

	* variable.c (lookup_variable) [VMS]: When getting values from the
	environment, allocate enough space for the _value_ plus escapes,
	not enough space for the name plus escapes :-/.
	Reported by John Fowler <jfowler@nyx.net>.

	* remake.c (f_mtime): Removed the "***" prefix from the mod time
	warnings that make generates, so it doesn't look like an error.
	Reported by Karl Berry <karl@gnu.org>.


	Fix for PR/2020:  Rework appended target-specific variables.  I'm
	fairly confident this algorithm is finally correct.

	* expand.c (allocated_variable_append): Rewrite.  Instead of
	expanding each appended variable then adding all the expanded
	strings together, we append all the unexpanded values going up
	through the variable set contexts, then expand the final result.
	This behaves just like non-target-specific appended variable
	values, while the old way didn't in various corner cases.
	(variable_append): New function: recursively append the unexpanded
	value of a variable, walking from the outermost variable scope to
	the innermost.
	* variable.c (lookup_variable): Remove the code that looked up the
	variable set list if the found variable was "append".  We don't
	need this anymore.
	(lookup_variable_in_set): Make this non-static so we can use it
	elsewhere.
	(try_variable_definition): Use lookup_variable_in_set() rather
	than faking out current_variable_set_list by hand (cleanup).
	* variable.h: Add a prototype for the now non-static
	lookup_variable_in_set().

2000-11-17  Paul D. Smith  <psmith@gnu.org>

	* remake.c (f_mtime) [WINDOWS32]: On various advice, I changed the
	WINDOWS32 port to assume timestamps can be up to 3 seconds away
	before throwing a fit.

2000-11-17  Paul D. Smith  <psmith@gnu.org>

	* read.c (readline): CRLF calculations had a hole, if you hit the
	buffer grow scenario just right.  Reworked the algorithm to avoid
	the need for len or lastlen at all.  Problem description with
	sample code chages provided by Chris Faylor <cgf@redhat.com>.

2000-10-24  Paul D. Smith  <psmith@gnu.org>

	* gettext.c (SWAP): Declare this with the prototype, otherwise
	some systems don't work (non-32-bit?  Reported for Cray T3E).
	Reported by Thorstein Thorsteinsson <thor@signe.teokem.lu.se>.

2000-10-05  Paul D. Smith  <psmith@gnu.org>

	* acinclude.m4 (AM_LC_MESSAGES): Remove undefined macro
	AM_LC_MESSAGES; it doesn't seem to do anything anyway??

	* i18n/gl.po, configure.in (ALL_LINGUAS): New Galician translation.

2000-09-22  Paul D. Smith  <psmith@gnu.org>

	* gettext.c: Don't #define _GETTEXT_H here; we only include some
	parts of the real gettext.h here, and we expect to really include
	the real gettext.h later.  If we keep this #define, it's ignored.

2000-09-21  Paul D. Smith  <psmith@gnu.org>

	* main.c (log_working_directory): Rework the text to use complete
	sentences, to make life simpler for the translators.

2000-08-29  Paul D. Smith  <psmith@gnu.org>

	* file.c (remove_intermediates): Print a debug message before we
	remove intermediate files, so the user (if she uses -d) knows
	what's going on.

2000-08-21  Paul D. Smith  <psmith@gnu.org>

	* variable.c (try_variable_definition): Change how we handle
	target-specific append variable defns: instead of just setting the
	value, expand it as an append _but_ only within the current
	target's context.  Otherwise we lose all but the last value if the
	variable is appended more than once within the current target
	context.  Fixes PR/1831.

2000-08-16  Paul D. Smith  <psmith@gnu.org>

	* function.c (func_shell): Nul-terminate the buffer before
	printing an exec error message (just in case it's not!).
	Fixes PR/1860, reported by Joey Hess <joey@valinux.com>.

2000-07-25  Paul D. Smith  <psmith@gnu.org>

	* job.c (construct_command_argv_internal): Add "~" to the list of
	sh_chars[] which disallow optimizing out the shell call.

2000-07-23  Paul Eggert  <eggert@twinsun.com>

	* NEWS, make.texinfo: Document .LOW_RESOLUTION_TIME, which
	supersedes --disable-nsec-timestamps.
	* make.texinfo: Consistently use "time stamp" instead of "timestamp".
	* README: Remove --disable-nsec-timestamps.

	* filedef.h (struct file.low_resolution_time): New member.
	* file.c (snap_deps): Add support for .LOW_RESOLUTION_TIME.
	* remake.c (update_file_1):
	Avoid spurious rebuilds due to low resolution time stamps,
	generalizing the earlier code that applied only to archive members.
	(f_mtime): Archive members always have low resolution time stamps.

	* configure.in: Remove --disable-nsec-timestamps, as this has
	been superseded by .LOW_RESOLUTION_TIME.

2000-07-23  Paul Eggert  <eggert@twinsun.com>

	* configure.in (enable_nsec_timestamps): Renamed from
	make_cv_nsec_timestamps, since enable/disable options
	shouldn't be cached.

2000-07-23  Bruno Haible  <haible@clisp.cons.org>
       and  Paul Eggert  <eggert@twinsun.com>

	* file.c (file_timestamp_now):
	Use preprocessor-time check for FILE_TIMESTAMP_HI_RES
	so that clock_gettime is not linked unless needed.

	* filedef.h (FILE_TIMESTAMP_HI_RES):
	Remove definition; "configure" now does this.

	* configure.in (jm_AC_TYPE_UINTMAX_T): Move up,
	to before high resolution file timestamp check,
	since that check now uses uintmax_t.
	(FILE_TIMESTAMP_HI_RES): Define to nonzero if the code should use
	high resolution file timestamps.
	(HAVE_CLOCK_GETTIME): Do not define if !FILE_TIMESTAMP_HI_RES,
	so that we don't link in clock_gettime unnecessarily.

2000-07-17  Paul D. Smith  <psmith@gnu.org>

	* i18n/ja.po: New version of the translation file.

2000-07-07  Paul D. Smith  <psmith@gnu.org>

	* remake.c (f_mtime): If NO_FLOAT is defined, don't bother with
	the offset calculation.
	(name_mtime): Replace EINTR test with EINTR_SET macro.

2000-07-07  Paul Eggert  <eggert@twinsun.com>

	Fix for PR/1811:

	* remake.c (update_file_1):
	Avoid spurious rebuilds of archive members due to their
	timestamp resolution being only one second.
	(f_mtime): Avoid spurious warnings of timestamps in the future due to
	the clock's resolution being lower than file timestamps'.
	When warning about future timestamps, report only the discrepancy,
	not the absolute value of the timestamp and the current time.

	* file.c (file_timestamp_now): New arg RESOLUTION.
	* filedef.h (file_timestamp_now): Likewise.
	(FILE_TIMESTAMP_NS): Now returns int.  All uses changed.

2000-07-05  Paul D. Smith  <psmith@gnu.org>

	* variable.c (lookup_variable) [VMS]: Remove vestigial references
	to listp.  Fixes PR/1793.

2000-06-26  Paul Eggert  <eggert@twinsun.com>

	* Makefile.am (MAINTAINERCLEANFILES): New macro, with stamp-pot in it.

	* dir.c (vms_hash): Ensure ctype macro args are nonnegative.

	* remake.c (f_mtime): Remove unused var memtime.

2000-06-25  Martin Buchholz  <martin@xemacs.org>

	* make.texinfo, NEWS, TODO.private: Minor spelling corrections.
	Ran spell-check	on make.texinfo.


See ChangeLog.2 for earlier changes.