the code that decides whether the cursor is at the beginning of
an otherwise-empty line
- 1/2
- ---
+ 1/2/1997
+ --------
support/bashbug.sh
- fixed a typo, thanks to eggert@twinsun.com
[bash-2.02-alpha1 frozen]
- 1/6/98
- ------
+ 1/6/1998
+ --------
lib/glob/fnmatch.c
- fix define for isgraph so that it does not return success for space
- fix strcompare() so that the call to strcoll is surrounded by
- include <string.h> or <strings.h> as appropriate for strchr()
declaration
- 1/4/99
- ------
+ 1/4/1999
+ --------
print_cmd.c
- don't print bare words when printing conditional expressions in
print_cond_node, use cprintf("%s", cond->op->word)
----
pcomplete.c
- fix to avoid freeing memory twice
+
+ 3/17
+ ----
+[bash-2.04 released]
+
+ 3/20
+ ----
+doc/bash.1
+ - minor typo fix to description of `xpg_echo' option in `echo'
+ description
+
+ 3/23
+ ----
+parse.y
+ - minor fix to parse_string_to_word_list to allow newlines in
+ compound array assignments
+
+ 3/27
+ ----
+lib/readline/rltty.c
+ - fixed a missing-semicolon syntax error in the old 4.3 BSD tty code
+
+support/mksignames.c
+ - if SIGRTMAX is >= 2*NSIG, just #undef SIGRTMIN and SIGRTMAX. This
+ is an issue on AIX 4.3 (only, so far) -- checked by configure, which
+ #defines UNUSABLE_RT_SIGNALS if this is the case
+
+aclocal.m4
+ - new macro, BASH_CHECK_RTSIGS, checks whether or not real-time
+ signals are defined with large values and defines UNUSABLE_RT_SIGNALS
+ if they are
+
+config.h.in
+ - place marker for UNDEF_RT_SIGNALS (initially undefined)
+
+configure.in
+ - call BASH_CHECK_RTSIGS in the `bash miscellaneous' section
+
+ 3/29
+ ----
+
+subst.c
+ - fixed a problem in match_pattern_char, where `string' was checked
+ for a paren following a [?+!@], rather than `pat'
+ - the code that checks whether or not to exit after a failed
+ ${word?msg} substitution needs to check interactive_shell rather
+ than interactive to avoid exiting if such a construct appears in
+ a sourced file
+
+ 3/31
+ ----
+
+bashline.c
+ - add `{' to the list of characters that need to be quoted by the
+ completion code when appearing in a filename
+
+ 4/1
+ ---
+lib/glob/fnmatch.c
+ - fixed an error in brackmatch() while skipping the rest of a bracket
+ expression after a character matched. When `c' was a backslash and
+ *p was a backslash, as in a pattern like [/\\], the old code didn't
+ take both backslashes into account and ended up skipping the `]'
+ after erroneously concluding that it was quoted. Report and fix
+ from akim@epita.fr
+
+ 4/3
+ ---
+lib/readline/rltty.c
+ - slight change to the BSD tty interface code, so that if the first
+ ioctl fails, get_tty_settings returns -1 immediately
+ - better checking of ioctl return values in BSD tty version of
+ get_tty_settings
+
+ 4/5
+ ---
+doc/{bash.1,bashref.texi}
+ - documented use of LINES and COLUMNS variables
+
+ 4/12
+ ----
+lib/readline/rltty.c
+ - change the SETATTR define for the termio tty driver to use
+ TCSETAW (the analog of the termios TCSADRAIN)
+
+lib/glob/fnmatch.c
+ - fix for posix-style bracket expressions ([:xxx:], [=x=]) broken
+ by fix of 4/1 for quotes in bracket expressions
+
+ 4/26
+ ----
+subst.c
+ - fix to list_string and get_word_from_string to make a non-whitespace
+ IFS character preceded by IFS whitespace part of the current field
+ delimiter, not a separate delimiter that would result in a separate
+ null field when splitting. cf POSIX.2, 3.6.5, (3)(b). Bug reported
+ by amc@cs.berkeley.edu
+ - changed the behavior of the (currently undocumented) ${!prefix@}
+ expansion to be analogous to the expansion of $@ in other contexts
+ when double-quoted. ksh93 seems to do this. Bug reported by
+ Ken Pizzini <ken@halcyon.com>
+ - changed read_comsub to ignore NUL bytes in the command substitution
+ output, printing a warning message for now
+
+parse.y
+ - changed shell_getc to ignore NUL bytes in the input, printing a
+ warning message for now
+ - changed read_a_line to print a warning when a NUL byte is encountered,
+ like shell_getc does
+
+ 5/18
+ ----
+subst.c
+ - make sure split_at_delims uses d2 only if it's non-null
+ - make split_at_delims skip leading newlines in the string to split,
+ if newline is a delimiter
+
+jobs.c
+ - prevent hanging processes by marking a child process as not running
+ (even if the shell thinks it is) when waitpid() returns -1/ECHILD --
+ this indicates that waitpid never returned a reasonable status
+ for that child though, which may be a symptom of another problem
+
+builtins/cd.def
+ - make `pwd' print an error message if the write fails when displaying
+ the current directory
+
+lib/readline/histexpand.c
+ - return NULL from history_find_word if history_tokenize_internal
+ returns (char **)NULL rather than attempting to dereference a
+ possibly-NULL pointer
+ - change history_tokenize_internal to set *indp to -1 before any
+ processing if indp is non-null
+
+ 5/19
+ ----
+builtins/trap.def
+ - when in POSIX.2 mode and printing trap dispositions, print the
+ signal name without the leading `SIG', as POSIX.2 requires
+
+ 5/23
+ ----
+lib/readline/readline.c
+ - readline_initialize_everything now assigns $TERM to rl_terminal_name
+ if the application has left it unset, and calls _rl_init_terminal_io
+ with the resultant value
+
+lib/readline/doc/rltech.texinfo
+ - updated the description of rl_terminal_name to note initialization
+
+parse.y
+ - change parse_string_to_word_list to save and restore the value of
+ current_command_line_count before calling back into the parser --
+ it will screw up history if the string extends over one line. Bug
+ reported by Jan.Djarv@mbox200.swipnet.se
+
+ 6/7
+ ---
+unwind_prot.h
+ - change unwind_protect_string to pass the address of the variable to
+ unwind_protect_var on machines where ints and char *s are not the
+ same size
+
+unwind_prot.c
+ - make sure we save the value of a variable in unwind_protect_var on
+ machines where sizeof(char *) != sizeof(int), not the contents the
+ value points to. Fix from Andreas Schwab <schwab@suse.de>
+
+ 6/8
+ ---
+configure.in
+ - auto-configure MacOS X without bash malloc (like Rhapsody)
+
+ 6/12
+ ----
+configure.in
+ - replace `cygwin32' with `cygwin'
+
+builtins/evalfile.c
+ - in _evalfile, change all instances of \r to \n in the file to be
+ sourced and evaluated if __CYGWIN__ is defined
+
+input.c
+ - change \r to \n in the buffer argument to make_buffered_stream and
+ after calling zread() from b_fill_buffer if __CYGWIN__ is defined
+
+ 6/29
+ ----
+lib/sh/Makefile.in
+ - add rest of dependency rules for makepath.o
+
+ 7/3
+ ---
+jobs.c
+ - changed `int' to `WAIT' in a couple of places for non-POSIX
+ environments
+
+ 7/5
+ ---
+locale.c
+ - try to avoid an unnecessary memory allocation in localetrans() in
+ the case that a translation for the string exists
+
+lib/readline/doc/rluserman.texinfo
+ - change the `direntry' information to make it use rluserman instead
+ of readline
+
+ 7/6
+ ---
+execute_cmd.c
+ - changes to execute_in_subshell so it honors a `!' prefixing a command
+ inside a user subshell (...)
+
+support/config.{guess,sub}
+ - add cases for Apple Darwin (Mac OS X)
+
+configure.in
+ - Apple Darwin (MacOS X) uses its own malloc
+
+lib/readline/readline.h
+ - removed bogus extern declaration of savestring
+
+ 7/7
+ ---
+builtins/common.c
+ - in get_working_directory, don't count on getcwd returning anything
+ useful in the buffer passed as the first argument on failure; just
+ use bash_getcwd_errstr in the error message
+
+examples/loadables/cut.c
+ - loadable version of cut(1), from the FreeBSD source tree
+
+ 7/18
+ ----
+lib/readline/vi_mode.c
+ - changed two instances of (*rl_getc_function) (rl_instream) to calls
+ to rl_read_key(), which works better when using the readline
+ callback interface
+
+builtins/evalfile.c
+ - added FEVAL_CHECKBINARY flag for _evalfile; controls whether a check
+ for a binary file is performed
+ - FEVAL_CHECKBINARY is not set by source_file or maybe_execute_file,
+ which means that the startup file execution and the `.' builtin will
+ not check for binary files
+ - FEVAL_CHECKBINARY is not set by fc_execute_file, so that files
+ composed of history commands run by the `fc' builtin will not check
+ for binary files
+
+ 7/21
+ ----
+shell.c
+ - added `--init-file' as a synonym for `--rcfile'
+
+ 7/26
+ ----
+support/shobj-conf
+ - added commented-out stanzas for hpux 10 and hpux 11 when the HP
+ unbundled ANSI C compiler is being used
+
+ 7/27
+ ----
+shell.c
+ - don't call end_job_control() from exit_shell if subshell_environment
+ is non-zero, even if the shell is interactive
+
+ 8/1
+ ---
+bashhist.c
+ - change maybe_add_history to remember if we saved the first line of
+ a command and don't save the second and subsequent lines of a
+ multi-line command if we didn't save the first line
+
+configure.in, config.h.in
+ - add autoconf check for getservbyname()
+
+lib/sh/netopen.c
+ - add support for named services in /dev/(tcp|udp)/host/service
+ translation
+ - make sure _netopen sets errno to something other than 0 if something
+ goes wrong, so the redirection functions behave
+
+ 8/3
+ ---
+parse.y
+ - changes to shell_getc so that history saving of blank lines in
+ multi-line commands does a better job of syntactic correctness
+
+bashline.c
+ - bind ^O to operate-and-get-next in emacs_standard_keymap explicitly,
+ rather than letting rl_add_defun do it in whatever keymap is current
+
+ 8/4
+ ---
+builtins/ulimit.def
+ - removed some of the special handling of RLIM_INFINITY -- the old
+ code would let people use RLIM_INFINITY as a shorthand for setting
+ the soft limit to the hard limit. Now, the code just passes
+ RLIM_INFINITY through if the hard limit is being set, so any kernel
+ errors will get through
+
+ 8/10
+ ----
+parse.y
+ - change the grammar rule for group_command so that it's composed of
+ a compound_list instead of a list, which requires a terminator
+ This means that you can now do (legal) things like
+
+ { { echo a b c ; } }
+
+ since the second close brace will be recognized as such because a
+ reserved word is legal in that context
+
+ 8/23
+ ----
+subst.c
+ - fix parameter_brace_substring to free all memory allocated by
+ get_var_and_type in the VT_ARRAYMEMBER case (returned parameter
+ `val')
+
+ 8/30
+ ----
+variables.c
+ - now that we have dynamic array variables, remove the check for
+ array_p in bind_variable and let array variables with assignment
+ functions call the assignment function
+
+subst.c
+ - change cond_expand_word to perform tilde expansion (like it should
+ have done all along)
+
+lib/readline/complete.c
+ - don't allow backslash to quote anything inside a single-quoted
+ string in find_completion_word
+
+doc/{bash.1,bashref.texi}
+ - note that `set -a' will cause functions to be exported
+
+lib/readline/parens.c
+ - changed the blink timeout so it's settable by the application. A
+ new function, int rl_set_paren_blink_timeout (int u), (still
+ undocumented) will set the timeout to U usec
+
+variables.c
+ - at startup of an interactive login shell, if the current directory
+ is $HOME, and $HOME and $PWD are not the same, set the current
+ directory (and $PWD) to $HOME
+ - break code that initializes $PWD and $OLDPWD into set_pwd function
+
+variables.h
+ - new extern declaration for set_pwd()
+
+ 9/6
+ ---
+lib/readline/{readline,callback}.c
+ - allocate memory for rl_prompt instead of simply using what's passed
+ as the `prompt' argument to readline() or
+ rl_callback_handler_install(). This will allow constant strings to
+ be passed to readline(), since the prompt processing code wants to
+ write to rl_prompt
+
+lib/sh/pathcanon.c
+ - new file, does pathname canonicalization
+
+externs.h
+ - extern declarations for stuff in lib/sh/pathcanon.c
+
+general.c
+ - canonicalize_pathname now simply calls sh_canonpath()
+
+ 9/7
+ ---
+pcomplete.h
+ - new member in a `struct compspec': options
+ - define some flag values for options field
+
+pcomplib.c
+ - make sure options field is allocated and copied by compspec utility
+ functions
+
+builtins/complete.def
+ - add struct for completion options. These control the `meta-behavior'
+ of the compspec. Initially, there are three:
+
+ default - perform bash default completion if programmable
+ completion produces no matches
+ dirnames - perform directory name completion if programmable
+ completion produces no matches
+ filenames - tell readline that the compspec produces filenames,
+ so it can do things like append slashes to
+ directory names and suppress trailing spaces
+
+ - add `-o option' to complete and compgen, setting new options field
+ in a compspec
+ - add code to display any -o options when a compspect is printed
+ with the `-p' option to complete
+
+pcomplete.c
+ - programmable_completions now passes back to the caller in the last
+ argument any options associated with the active compspec (from the
+ options field)
+ - change gen_compspec_completions to re-call gen_action_completions
+ with a CA_DIRECTORY action if the rest of the compspec actions
+ generate no matches and the COPT_DIRNAMES option is present
+
+bashline.c
+ - if a programmable completion was specified to return filenames with
+ the `-o filenames' option, set rl_filename_completion_desired to 1
+ so readline does its special filename-specific things
+ - if a compspec was defined to fall back to the readline default with
+ the `-o default' option, suppress setting the
+ rl_attempted_completion_over variable in attempt_shell_completion
+ so readline will go on and do filename completion
+
+include/ansi_stdlib.h
+ - add extern declarations for atof and strtod
+
+builtins/printf.def
+ - change getdouble() to use strtod (since we provide a version in
+ libsh if the C library doesn't have one), and change the calling
+ convention to match the other getXXX functions (take value pointer
+ as argument, return success or failure). This makes printf handle
+ invalid floating point numbers better
+
+doc/bash.1, lib/readline/doc/rluser.texinfo
+ - documented new `-o' option to complete and compgen; described the
+ arguments and their effects
+
+lib/readline/{macro,util}.c
+ - renamed _rl_executing_macro to rl_executing_macro; made it public
+
+lib/readline/readline.h
+ - new extern declaration for rl_executing_macro
+
+lib/readline/rltech.texinfo
+ - documented rl_executing_macro
+
+ 8/8
+ ---
+lib/readline/readline.c
+ - new state variable, rl_readline_state
+
+lib/readline/readline.h
+ - extern declaration for rl_readline_state
+ - define some flag bits for rl_readline_state and macros to test and
+ set them
+
+lib/readline/{readline,complete,isearch,macro,rltty,search,signals,undo,vi_mode}.c
+ - added calls to the macros that set and unset various readline states
+
+lib/readline/readline.c
+ - new private function, _rl_free_saved_history_line, to free up the
+ history entry saved in saved_line_for_history
+
+lib/readline/search.c
+ - free any saved history entry in rl_history_search_reinit(), because
+ rl_get_previous_history makes a call to maybe_save_line without
+ freeing up the saved line when it returns (so rl_get_next_history
+ can use it)
+ - fix up noninc_search_from_pos to deal with invalid history indices
+ better
+
+ 9/12
+ ----
+support/shobj-conf
+ - HPUX 11.0 needs the `-fpic' flag passed to the linker as well as
+ the compiler
+
+execute_cmd.c
+ - on MS-DOS systems, let the #! checking stuff in execute_shell_script
+ treat \r as EOL (identical to \n)
+
+input.c
+ - on DJGPP systems, allow buffered_getchar to ignore \r. Change
+ suggested by snowball3@bigfoot.com
+
+ 9/13
+ ----
+lib/sh/tmpfile.c
+ - new file, with functions for creating and (hopefully safely) opening
+ temporary files for the shell
+
+externs.h
+ - extern declarations for functions in tmpfile.c
+
+variables.c
+ - new function, get_random_number, returns random values from the
+ shell's (lame) RNG to the rest of the shell
+
+variables.h
+ - new extern declaration for get_random_number()
+
+subst.c
+ - changed make_named_pipe to call sh_mktmpname instead of mktemp
+
+redir.c
+ - changed here_document_to_fd to call sh_mktmpfd instead of using
+ inline code
+
+builtins/fc.def
+ - changed fc_builtin to call sh_mktmpfp instead of using inline code
+
+builtins/common.h
+ - new extern declaration for fc_execute_file()
+
+bashline.c
+ - if RL_ISSTATE(RL_STATE_INITIALIZED) is 0 in initialize_readline,
+ call rl_initialize to bind the defaults before setting up our
+ custom shell functions and key bindings
+
+lib/readline/doc/{rluser,hsuser}.texinfo
+ - some markup changes proposed by dima@chg.ru
+
+lib/readline/rlstdc.h, lib/tilde/tilde.h
+ - to get ready for some `const' changes to the source, define const
+ as either `__const' (gcc -traditional) or nothing if __STDC__ and
+ __cplusplus are not defined
+
+lib/readline/readline.c
+ - upped the default readline version to `4.2-alpha'
+
+lib/readline/readline.[ch]
+ - readline now takes a `const char *' as an argument
+
+lib/readline/history.h
+ - read_history(), read_history_range(), write_history(),
+ append_history(), history_truncate_file() now take `const char *'
+ filename arguments
+
+lib/readline/histfile.c
+ - history_filename(), read_history(), read_history_range(),
+ write_history(), append_history(), history_truncate_file(),
+ history_do_write() now take `const char *' filename arguments
+
+lib/readline/readline.h
+ - rl_read_init_file() now takes a `const char *' filename argument
+
+lib/readline/bind.c
+ - current_readline_init_file is now a `const char *'
+ - _rl_read_init_file(), rl_read_init_file() now take `const char *'
+ filename arguments
+
+lib/tilde/tilde.c
+ - tilde_expand(), tilde_expand_word() now take a `const char *'
+ as their first argument
+
+lib/readline/chardefs.h
+ - fixed CTRL_CHAR macro to be right on systems where chars are unsigned
+ by default
+
+ 9/14
+ ----
+lib/readline/doc/{rl,hs}tech.texinfo
+ - documented new `const char *' function arguments where appropriate
+
+lib/readline/vi_mode.c
+ - clamp the numeric argument in vi mode at 999999, just like in emacs
+ mode
+
+lib/readline/doc/rltech.texinfo
+ - documented rl_readline_state and its possible bit values
+
+builtins/source.def
+ - in maybe_pop_dollar_vars, don't overwrite a shell function's
+ argument list even if the sourced script changes the positional
+ parameters
+
+lib/readline/{util.c,rlprivate.h}
+ - made _rl_strindex() and _rl_savestring() take `const char *'
+ arguments
+
+ 9/15
+ ----
+subst.c
+ - renamed `doing_completion' to `no_longjmp_on_fatal_error', a more
+ accurate description of its current behavior
+
+pcomplete.[ch], externs.h, lib/sh/stringlist.c
+ - moved the stringlist code from pcomplete.c to lib/sh/stringlist.c
+ - moved the type declaration and extern function declarations from
+ pcomplete.h to externs.h
+
+stringlib.c, externs.h, lib/sh/stringvec.c
+ - moved the char array code from stringlib.c to lib/sh/stringvec.c
+
+ 9/18
+ ----
+subst.c
+ - new static variable, expand_no_split_dollar_star, to ensure that
+ $* isn't `split and rejoined' when expanded via a call to
+ expand_string_unsplit or expand_word_no_split
+
+ 9/19
+ ----
+subst.[ch], execute_cmd.c
+ - renamed expand_word_no_split to expand_word_unsplit
+
+copy_cmd.c
+ - in copy_word, just assign w->flags to new_word->flags rather than
+ bcopying the entire structure
+
+configure.in, config.h.in
+ - add autoconf check for pathconf(3)
+
+lib/sh/tmpfile.c
+ - test for systems with short filenames using pathconf and
+ _PC_NAME_MAX, so generated temp file names don't exceed that length
+ (only do this if the file name length is less than 32, since the
+ current code doesn't generate filenames longer than about 17 chars)
+
+ 9/20
+ ----
+command.h
+ - added `W_NOSPLIT2' word flag to prepare for supporting weird $@
+ expansion on the rhs of assignment statements (inspecting the ksh93
+ behavior, it seems like nothing except $@ is split, and $@ is split
+ on $IFS and rejoined as a single string)
+
+ 9/21
+ ----
+subst.c
+ - since verify_substring_values looks for arithmetic expressions
+ bounded by `:', do some ad-hoc parsing of the substring passed as
+ the argument to avoid cutting the expression off at the wrong `:'
+ (since the inclusion of conditional expressions, expressions may
+ contain a `:'). Do some paren balancing, too.
+
+ 9/22
+ ----
+command.h
+ - new #define, CLOBBERING_REDIRECT, evaluates to true for any redirect
+ that will clobber a file and should be subject to `noclobber'
+
+redir.c
+ - changed redir_open to use CLOBBERING_REDIRECT instead of
+ OUTPUT_REDIRECT. This means that <> no longer tests the value of
+ noclobber, as POSIX.2 specifies
+
+parse.y
+ - made a change to cond_term so that expressions like [[ (x) ]] are
+ parsed correctly
+
+ 9/25
+ ----
+parse.y
+ - small change to grammar for arith_for_command and special_case_tokens
+ to allow `do...done' or {...} after ((...)) without an intervening
+ list_terminator (for ksh93 compatibility)
+
+builtins/printf.def
+ - `printf' now treats \E and \e the same when performing backslash
+ escape expansion for the %b format character for compatibility with
+ $'...' quoting (for ksh93 compatibility)
+
+ 9/26
+ ----
+builtins/source.def
+ - new variable, source_searches_cwd, initially 1. If non-zero, `.'
+ searches $PWD for the filename argument if it's not found in $PATH
+
+general.c
+ - changed posix_initialize to turn off source_searches_cwd if posix
+ mode is being enabled; turn it on if posix mode is being disabled
+ - changed things so interactive_comments is not turned off if posix
+ mode is disabled, since that's rarely what is desired
+
+variables.c
+ - changed uidset() to set UID and EUID only if they are not already
+ set when the shell starts up
+ - changed things to HOSTNAME is set only if it's not set when the
+ shell starts up
+
+execute_cmd.c
+ - changed execute_builtin_or_function so that if a command like
+ `command exec 4<file' is executed, the effect is as if the `command'
+ had been omitted, as POSIX.2 requires
+
+subst.c
+ - changed array_length_reference to support ${foo[X]} when foo is not
+ an array variable and X is 0, @, or * (everything else returns 0)
+ for ksh compatibility
+ - changed array_value_internal to support ${foo[@]} and ${foo[*]}
+ when foo is not an array variable (as ksh does)
+
+ 9/27
+ ----
+general.h
+ - added defines for ABSPATH (absolute pathname), RELPATH (relative
+ pathname, and ROOTEDPATH (pathname rooted at /), with different
+ definitions for Unix and Cygwin
+ - added defines for DIRSEP (the directory separator in pathnames,
+ and ISDIRSEP (whether a particular character is the directory
+ separator)
+ - added PATHSEP define: takes a character and decides whether it
+ delimits a filename component of a pathname
+ - changed isletter(c) #define to just call isalpha(c)
+
+lib/sh/pathcanon.c
+ - changed to use new pathname #defines from general.h above
+
+general.c
+ - changed make_absolute, absolute_pathname to understand
+ Cygwin's idea of an absolute pathname: [[:upper:][:lower:]]:/
+ - rewrote absolute_pathname to use PATHSEP and ABSPATH
+ - changed full_pathname to use ABSPATH
+
+variables.c
+ - changed get_bash_name to use ABSPATH and RELPATH, and to handle the
+ case when $PWD is unset when it is called
+
+subst.c
+ - if compiling on cygwin, translate \r\n to \n when reading the output
+ of a command substitution in read_comsub()
+
+ 9/28
+ ----
+input.c, shell.c, builtins/read.def, lib/tilde/tilde.c
+ - changed __CYGWIN32__ to __CYGWIN__
+
+lib/readline/{bind,funmap,kill}.c, lib/readline/readline.h
+ - changed __CYGWIN32__ to __CYGWIN__
+
+input.c
+ - a couple of minor changes to clean up the code
+
+parse.y
+ - changed shell_getc and shell_ungetc to implement one character of
+ lookahead/lookbehind when pushing a character back at the start
+ of an input line (the old code would have discarded it because
+ shell_input_line_index was 0)
+
+lib/readline{histsearch.c,history.h}
+ - history_search_internal, history_search, history_search_prefix,
+ history_search_pos now take const char * string arguments
+
+lib/readline/{histexpand.c,history.h}
+ - get_history_event, history_arg_extract, history_tokenize_internal,
+ history_tokenize now take const char * string arguments
+
+lib/readline/{history.c,history.h}
+ - add_history, replace_history_entry now take const char * string
+ arguments
+
+lib/readline/{bind.c,readline.h}
+ - rl_add_defun, rl_unbind_command_in_map, rl_set_key, rl_macro_bind,
+ rl_generic_bind, rl_translate_keyseq, rl_named_function,
+ rl_function_of_keyseq, rl_variable_bind, rl_get_keymap_by_name
+ now take const char * string arguments where appropriate
+ - rl_function_dumper, rl_list_funmap_names needed changes to deal
+ with rl_funmap_names returning const char **
+ - rl_readline_name is now a const char * variable
+
+lib/readline/keymaps.h
+ - duplicate definition of rl_get_keymap_by_name needs const changes
+
+lib/readline/{bind.c,rlprivate.h}
+ - _rl_bind_if_unbound takes a const char * string argument
+
+lib/readline/{funmap.c,readline.h}
+ - rl_add_funmap_entry now takes a const char * string argument
+ - rl_funmap_names now returns a const char **
+ - a FUNMAP now has a const char *name member
+
+lib/readline/{terminal.c,readline.h,rlprivate.h}
+ - _rl_init_terminal_io, rl_get_termcap, rl_reset_terminal,
+ _rl_output_some_chars now take const char * string arguments
+
+lib/readline/{readline.[ch],rlprivate.h}
+ - rl_library_version and rl_terminal_name are now const char *
+ variables
+ - rl_insert_text, _rl_replace_text now take const char * string
+ arguments
+
+lib/readline/{callback.c,readline.h}
+ - rl_callback_handler_install takes a const char * string argument
+
+lib/readline/doc/{rltech,hstech}.texinfo
+ - changed documentation for newly-consted functions and variables
+
+lib/readline/{complete.c,readline.h}
+ - rl_strpbrk now takes const char * string arguments
+ - username_completion_function and filename_completion_function
+ now take const char * string arguments
+ - rl_basic_word_break_characters, rl_basic_quote_characters,
+ rl_completer_word_break_characters, rl_completer_quote_characters,
+ rl_filename_quote_characters, rl_special_prefixes are all
+ const char * variables
+
+lib/readline/{display.c,rlprivate.h}
+ - _rl_move_cursor_relative takes a const char * string argument
+
+redir.c
+ - change from tbisp@uswest.net to fix a problem with cygwin and
+ here_document_to_fd, since cygwin doesn't allow a file to be
+ unlinked if there is a descriptor open on it
+
+display.c
+ - fixed a slight bug in the logic deciding whether or not to redraw
+ a prompt containing invisible characters in update_line. Fixes
+ bug reported by bdluevel@heitec.net
+
+ 9/29
+ ----
+shell.c
+ - fixed a problem with a too-aggressive setting of restricted shell
+ mode based on shell_name. Old code set `restricted' even after a
+ longjmp due to shell_execve trying to execute an executable script
+ without a leading `#!', though that case should result in an
+ unrestricted shell
+
+ 10/2
+ ----
+copy_cmd.c
+ - make sure copy_function_def copies the `flags' and `line' members
+ of the function_def
+
+parse.y
+ - make the function_def production take a new function_body parse
+ unit, which is currently just a group_command (in prep for future
+ changes)
+
+builtins/printf.def
+ - changed getint, getlong, getulong, getdouble to print warnings on
+ range errors and return the clamped values that strtol/strtod
+ return, as POSIX.2 specifies
+
+ 10/3
+ ----
+parse.y
+ - made a function_def a `command' instead of a `shell_command'
+ This makes the grammar more closely resemble that supplied in
+ the POSIX.2 standard
+ - made a `function_body' be like the function_body production in
+ the POSIX.2 standard. This relaxes the restriction that the
+ body be composed of a group_command, but introduces a YACC
+ shift/reduce conflict
+
+print_cmd.c
+ - changes to named_function_string (for normal function printing)
+ and print_function_def (for printing function definitions embedded
+ in other commands) to print redirections that should be attached
+ to the function as a whole after the closing brace, because of
+ the changes to the parser
+
+tests/func2.sub
+ - tests for new function declarations whose bodies are not group
+ commands -- the command printing code still prints them with the
+ braces
+
+ 10/4
+ ----
+lib/sh/pathphys.c
+ - new function, sh_physpath, resolves symlinks in pathname passed
+ as argument
+
+builtins/cd.def
+ - changed pwd_builtin to use sh_physpath if -P supplied or in
+ physical mode
+ - changed bindpwd to use sh_physpath if the no_symlinks argument is
+ non-zero
+ - changed change_to_directory to use sh_physpath and set
+ the_current_working_directory appropriately if the `nolinks'
+ argument is non-zero
+
+ 10/5
+ ----
+variables.c, bashline.c, builtins/cd.def
+ - replaced calls to canonicalize_pathname with sh_canonpath
+
+builtins/cd.def
+ - rewrote change_to_directory to make it much shorter and simpler
+
+command.h
+ - added SUBSHELL_PIPE as possible flag value for subshell_environment
+
+execute_cmd.c
+ - execute_simple_command sets SUBSHELL_PIPE in subshell_environment
+ if the command has a pipe in or out
+
+subst.c
+ - don't reset pipeline_pgrp to shell_pgrp in command_substitute if
+ we're already in a pipline
+ ((subshell_environment & SUBSHELL_PIPE) != 0)
+
+ 10/6
+ ----
+subst.c
+ - change command_substitute to call setmode(fd, O_TEXT) on the pipe
+ file descriptor if __CYGWIN__ is defined; remove the other code
+ that converts CR/LF to LF on cygwin
+
+ 10/9
+ ----
+stringlib.c
+ - change strcreplace() to make sure that RESIZE_MALLOCED_BUFFER is
+ called everywhere characters are added
+
+ 10/11
+ -----
+lib/sh/spell.c
+ - new file, with the spelling correction code from builtins/cd.def
+
+builtins/cd.def
+ - a little surgery on cd_builtin, to make the structure clearer
+
+ 10/12
+ -----
+builtins/read.def
+ - include fcntl.h and io.h on __CYGWIN__
+ - set mode of fd 0 to O_TEXT if __CYGWIN__ and O_TEXT are both
+ defined
+
+shell.c
+ - in open_shell_script, set mode of fd opened to the shell script
+ to O_TEXT if __CYGWIN__ and O_TEXT are both defined
+
+builtins/evalfile.c
+ - changed the __CYGWIN__ code to call setmode(fd, O_TEXT) instead
+ of translating \r to \n after the read() returns
+
+ 10/13
+ -----
+lib/sh/strtrans.c, stringlib.c
+ - moved ansicstr() from stringlib.c to new file lib/sh/strtrans.c
+ - new function ansic_quote, to quote a string with $'...'
+
+lib/sh/strindex.c, stringlib.c
+ - moved strindex() from stringlib.c to new file lib/sh/strindex.c
+
+lib/sh/Makefile.in
+ - compile `shquote.c' in as part of the shell library
+ - compile `strtrans.c' in as part of the shell library
+ - compile `strindex.c' in as part of the shell library
+
+Makefile.in
+ - link with libsh before any of the readline libraries
+
+builtins/common.c
+ - removed the functions also appearing in lib/sh/shquote.c
+
+externs.h, builtins/common.h
+ - moved extern declarations for functions in lib/sh/shquote.c to
+ externs.h from builtins/common.h
+
+variables.c
+ - print_var_value (used in the output of `set') now calls
+ ansic_quote to quote the value, so any non-printable characters
+ are displayed
+
+ 10/14
+ -----
+syntax.h
+ - new file, with definitions from shell.h having to do with shell
+ syntax, metacharacters, and quoting
+
+parse.y
+ - moved defines for shell metachars and other char class declarations
+ to syntax.h
+
+shell.h
+ - now includes syntax.h
+
+Makefile.in,builtins/Makefile.in,lib/sh/Makefile.in
+ - added syntax.h to appropriate dependency lists
+
+ 10/16
+ -----
+mksyntax.c
+ - new program, to build a syntax table for the shell at compile time
+
+Makefile.in
+ - new generated file, syntax.c, generated by mksyntax
+
+general.h
+ - `digit(x)' is now just defined as `isdigit(x)'
+
+syntax.h
+ - now that we have sh_syntaxtab[], change the appropriate defines
+ to use it
+
+parse.y, subst.c
+ - change checks of member(c, slashify_in_quotes) to test
+ sh_syntaxtab[c] & CBSDQUOTE
+
+builtins/set.def
+ - corrected the help text for `unset', since PATH and IFS may be
+ unset
+
+subst.c
+ - changed explicit tests for shell special parameters using member()
+ or strchr() to test (sh_syntaxtab[c] & CSPECVAR) instead
+ - changed some boilerplate calls to member() that test for validity
+ of various characters in the expansion process to macros for
+ efficiency
+
+lib/sh/shquote.c
+ - changed some of the functions to use checks against sh_syntaxtab[]
+ flag values
+
+ 10/19
+ -----
+findcmd.c
+ - find_user_command_in_path no longer finds directories if the caller
+ specified FS_NODIRS in the flags argument. This fixes the bug of
+ hashing directory names, at the loss of some information about
+ attempting to execute directories
+
+builtins/hash.def
+ - hash -p dirname name no longer puts directories into the hash table
+
+ 10/26
+ -----
+builtins/read.def
+ - make sure that the last unwind-protect added is the one that calls
+ xfree(input_string), so that this is the top element on the
+ unwind-protect stack
+ - if xrealloc() is called to increase the size of input_string, make
+ sure to remove the top element from the unwind-protect stack (the
+ old xfree) and add a new one, in case realloc() freed the original
+ input_string and allocated new space. This fixes the bug of bash
+ crashing because it's trying to free already-freed memory if the
+ `read' is interrupted (reported by klausf@Pool.Informatik.RWTH-Aachen.DE)
+
+ 10/27
+ -----
+subst.c
+ - fixed a problem in expand_word_list_internal so that
+ last_command_exit_value is set to 1 if a variable assignment fails
+ and a jump_to_top_level (DISCARD) is taken. This fixes the problem
+ of assignment statements to read-only variables appearing to succeed
+ (though it always worked right in non-interactive posix-mode shells)
+
+ 10/30
+ -----
+lib/readline/readline.h
+ - rl_completion_entry_function is now of type CPFunction *, since
+ that's the actual type that's used, and casting function pointers
+ of different types (which happened all the time as a result), is
+ problematic on some architectures
+
+lib/readline/complete.c
+ - changed entry function argument to gen_completion_matches to be of
+ type CPFunction *, since that's how it's used
+ - changed variables that are completion entry function pointers to be
+ of type CPFunction *, removed problematic typecasts
+
+lib/readline/doc/rltech.texinfo
+ - documented type change for rl_completion_entry_func
+
+lib/readline/rltypedefs.h
+ - new file, with all the readline function typedefs
+
+lib/readline/rltypedefs.h
+ - new typedef, rl_command_func_t, which is a function returning
+ an int
+ - a KEYMAP_ENTRY now contains a pointer to an rl_command_func_t
+ - new typedef, rl_compentry_func_t, which is a function returning
+ a char *
+ - new typdef, rl_completion_func_t, which is a function returning
+ a char **
+ - new typedef, rl_quote_func_t, which is a function returning a
+ char *
+ - new typedef, rl_dequote_func_t, which is a function returning a
+ char *
+ - new typedef, rl_compignore_func_t, which is a function returning a
+ char *
+ - new typedef, rl_compdisp_func_t, which is a void function taking
+ a char **, int, int argument list
+ - new typedef, rl_hook_func_t, which is a function taking no args
+ and returning an int (used for input and other hooks)
+ - new typedef, rl_modterm_func_t, which is a void function tkaing
+ an int, used to modify terminal (rl_prep_term_function)
+ - new typedef, rl_getc_func_t, which is an int-returning function
+ that takes a FILE *
+ - new typedef, rl_linebuf_func_t, which takes a string and an int
+ index and returns an int
+ - new generic typedefs, rl_intfunc_t and rl_voidfunc_t, which are
+ functions which take and return int and void, respectively
+ - new generic typedef, rl_vintfunc_t, which takes an int and returns
+ void
+ - new generic typedefs, rl_icpfunc_t and rl_vcpfunc_t, which take a
+ char * and return int and void, respectively
+ - new generic typedefs, rl_icppfunc_t and rl_vcppfunc_t, which take
+ a char ** and return int and void, respectively
+
+lib/readline/rlprivate.h
+ - rl_last_func is now of type rl_command_func_t *
+
+lib/readline/readline.c
+ - declare rl_last_func as rl_command_func_t *
+
+lib/readline/{bind,readline,vi_keymap,emacs_keymap,funmap,keymaps,isearch}.c
+lib/readline/readline.h
+ - change appropriate cases of (Function *) to (rl_command_func_t *)
+
+lib/readline/{readline.h,bind.c}
+ - rl_named_function and rl_function_of_keyseq now return a
+ (rl_command_func_t *)
+
+lib/readline/bind.c
+ - changed type of functions in parser function table from Function *
+ to _rl_parser_func_t *
+
+lib/readline/funmap.c
+ - removed `vi-bracktype' from vi function map -- it's not a
+ bindable function
+
+lib/readline/{readline.h,complete.c}
+ - changed appropriate uses of CPFunction * to rl_compentry_func_t *
+ or rl_quote_func_t *
+
+lib/readline/{rltty.c,readline.h}
+ - rl_prep_term_function is now of type rl_vintfunc_t *
+ - rl_deprep_term_function is now of type rl_voidfunc_t *
+
+lib/readline/{keymaps,history.h}
+ - include rltypedefs.h or <readline/rltypedefs.h> as appropriate
+
+lib/readline/{display.c,readline.h}
+ - rl_redisplay_func is now of type rl_voidfunc_t *
+
+lib/readline/{callbacks.c,readline.h}
+ - the `linefunc' argument to rl_callback_handler_install and the
+ rl_linefunc variable are now of type rl_vcpfunc_t *
+
+lib/readline/{complete.c,readline.h}
+ - rl_completion_display_matches_hook is now of type
+ rl_compdisplay_func_t *
+ - rl_directory_completion_hook is now of type rl_icppfunc_t *
+
+lib/readline/{history.h,histexpand.c}
+ - history_inhibit_expansion_function is now of type
+ rl_linebuf_func_t *
+
+lib/readline/histexpand.c
+ - new private type, _hist_search_func_t, used to indirect through
+ the various history search functions
+
+lib/tilde/tilde.h
+ - new typedef, tilde_hook_func_t, which takes a char * and returns
+ a char *
+
+lib/tilde/tilde.[ch]
+ - tilde_expansion_preexpansion_hook and tilde_expansion_failure_hook
+ are now of type tilde_hook_func_t *
+
+[THERE ARE NO LONGER ANY REFERENCES TO Function *, CPFunction *, or
+CPPFunction * IN THE READLINE SOURCE, WHICH INCLUDES THE TILDE LIBRARY]
+
+{bashline,bracecomp}.c, bashline.h
+ - changes for changed type of rl_completion_entry_func and other
+ changed types (rl_compentry_func_t, rl_completion_func_t,
+ const changes for prototypes, etc.)
+
+pathexp.[ch]
+ - shell_glob_filename now takes a `const char *' argument
+ - quote_string_for_globbing now takes a `const char *' argument
+
+ 10/31
+ -----
+lib/sh/zwrite.c
+ - fix zwrite() to update the buffer pointer in the case of a
+ partial write, in addition to decreasing the count of bytes to
+ be written
+
+support/config.{guess,sub}
+ - added entries for MIPS-based Compaq machines running NonStopUX
+
+ 11/1
+ ----
+lib/readline/undo.c
+ - renamed free_undo_list to be rl_free_undo_list, made a dummy
+ replacement function free_undo_list that just calls
+ rl_free_undo_list
+
+lib/readline/readline.h
+ - extern declaration for rl_free_undo_list
+
+lib/readline/{readline,signals}.c
+ - changed calls to free_undo_list to rl_free_undo_list
+
+lib/readline/doc/rltech.texinfo
+ - changed all function types and arguments for new typedefs
+
+ 11/2
+ ----
+lib/readline/{rlprivate.h,terminal.c,display.c}
+ - term_clreol, term_clrpag, term_cr, term_backspace, term_goto,
+ term_pc, term_im, term_ei, term_ic, term_ip, term_IC, term_dc,
+ term_DC, term_forward_char, term_up, term_mm, term_mo, term_ku,
+ term_kd, term_kr, term_kl, term_ks, term_ke, term_kh, term_kH,
+ terminal_can_insert variables all now have a _rl_ prefix
+
+lib/readline/{rlprivate.h,terminal.c,display.c, readline.c}
+ - screenheight, screenwidth, screenchars now have an _rl_ prefix
+
+lib/readline/{terminal,complete,isearch,kill,macro,readline,search,undo,util,vi_mode}.c
+ - changed calls to ding to calls to rl_ding
+
+lib/readine/{isearch,readline,search}.c
+ - saved_line_for_history now has an _rl_ prefix
+
+lib/readline/readline.[ch]
+ - renamed maybe_{replace,save,unsave} line to have an rl_ prefix
+
+lib/readline/{readline,search,isearch}.c
+ - changed calls to maybe_{replace,save,unsave} line to call the
+ rl_-prefixed equivalents
+
+lib/readline/{rlprivate.h,readline.c,util.c}
+ - renamed alphabetic to rl_alphabetic; changed calls
+
+lib/readline/{bind.c,funmap.c,rlprivate.h}
+ - possible_control_prefixes -> _rl_possible_control_prefixes
+ - possible_meta_prefixes -> _rl_possible_meta_prefixes
+
+lib/readline/compat.c
+ - new file for entry points for backwards compatibility functions
+ - moved free_undo_list here from undo.c
+ - maybe_{replace,save,unsave}_line function that call their rl_
+ counterparts
+ - moved ding() here; it just calls rl_ding()
+ - moved alphabetic here; it just calls _rl_alphabetic()
+
+lib/readline/Makefile.in
+ - make sure compat.o is built and linked appropriately
+
+bashline.c,pcomplete.c
+ - changed calls to ding() to call rl_ding()
+
+support/shobj-conf
+ - changed linux stanza SHLIB_XLDFLAGS to make the `soname' in a
+ shared library have only the major version number, rather than
+ major and minor numbers
+ - made analogous changes for bsd/os-4
+ - changed freebsd 3.x stanza to match `freebsdelf[3-9]*' for
+ FreeBSD elf systems and `freebsdaout[3-9]*' for FreeBSD
+ a.out systems (are there any?)
+
+lib/readline/doc/{rluser,hsuser}.texinfo
+ - changed uses of @samp to @option where appropriate
+
+doc/bashref.texi
+ - changed uses of @samp to @option where appropriate
+
+ 11/3
+ ----
+lib/readline/{readline.h,complete.c,compat.c}
+ - completion_matches, username_completion_function,
+ filename_completion_function all have an rl_ prefix
+
+{bashline,bracecomp,pcomplete}.c
+ - changed calls to completion_matches, username_completion_function,
+ and filename_completion_function
+
+lib/sh/shquote.c
+ - single_quote, double_quote, un_double_quote, backslash_quote,
+ backslash_quote_for_double_quotes, contains_shell_metas all now
+ have an sh_ prefix
+
+externs.h
+ - changed extern declarations for functions from shquote.c
+
+{array,bashline,pcomplete,print_cmd,variables}.c, parse.y
+builtins/{alias,complete,trap,type}.def
+ - changed callers of single_quote
+
+{array,bashline}.c, builtins/setattr.def
+ - changed callers of double_quote
+
+{bashline,bracecomp}.c, parse.y, builtins/printf.def
+ - changed callers of backslash_quote
+
+parse.y
+ - changed callers of backslash_quote_for_double_quotes
+
+{print_cmd,variables}.c
+ - changed callers of contains_shell_metas
+
+lib/readline/{rlshell.h,shell.c,histexpand.c}
+ - single_quote -> sh_single_quote
+
+lib/readline/{rlshell.h,shell.c,input.c}
+general.[ch],shell.c,subst.c
+ - unset_nodelay_mode -> sh_unset_nodelay_mode
+
+lib/readline/{rlshell.h,shell.c}, lib/tilde/tilde.c, variables.c
+ - get_home_dir -> sh_get_home_dir
+
+lib/readline/rlshell.h,lib/readline/bind,histfile,nls,readline,shell,terminal}.c, lib/tilde/tilde.c, variables.c
+ - get_env_value -> sh_get_env_value
+
+lib/readline/rlshell.h,lib/readline/{shell,terminal}.c
+variables.h,{jobs,nojobs,variables}.c
+ - set_lines_and_columns -> sh_set_lines_and_columns
+
+ 11/6
+ ----
+configure.in
+ - bash can only link with an already-installed readline library whose
+ version is 4.2 or greater
+
+variables.h
+ - new variable attribiute, noassign, with its accompanying test macro,
+ noassign_p. Completely internal, not user-settable
+
+execute_cmd.c
+ - treat noassign variables like readonly variables, except an attempt
+ to assign to them does not cause a non-interactive shell in POSIX
+ mode to abort
+
+{subst,variables}.c,builtins/declare.def
+ - attempts to assign to `noassign' variables return an error status,
+ but no error messages -- almost identical to readonly variables
+
+variables.c
+ - GROUPS and FUNCNAME are `noassign' variables, meaning assignment to
+ them is disallowed, but that they may be unset (in which case they
+ lose their special properties)
+
+doc/{bash.1,bashref.texi}
+ - documented that assignments to GROUPS and FUNCNAME are disallowed
+ and return an error status
+
+ 11/8
+ ----
+lib/readline/{funmap.c,bind.c,rlprivate.h}
+ - _rl_possible_{control,meta}_prefixes are now arrays of `const char *'
+ and are defined in bind.c, since that's where they're used
+
+lib/readline/rltty.c
+ - changed prepare_terminal_settings to take an `oldtio' argument
+ instead of `otio', which shadows the global declaration
+
+lib/readline/bind.c
+ - the `name' member of a parser_directive element is now a
+ `const char *'
+ - the `name' member of a boolean_varlist element is now a
+ `const char *'
+ - the `name' member of a string_varlist element is now a
+ `const char *'
+ - the `name' member of an assoc_list element is now a
+ `const char *'
+ - the `name' member of an keymap_names element is now a
+ `const char *'
+
+lib/readline/display.c
+ - changed scope-local char * variable `temp' to `vtemp' because
+ there's already a function local variable named `temp' of a
+ different type (int) in rl_redisplay
+
+lib/readline/util.c
+ - pathname_alphabetic_chars is now `const char *'
+
+lib/readline/terminal.c
+ - the `tc_var' member of a _tc_string element is now a
+ `const char *'
+
+lib/readline/histexpand.c
+ - changed history_expand_internal to remove the duplicate local
+ declaration of `t' in the `s'/`&' case
+
+lib/readline/histfile.c
+ - in history_filename, `home' is now a `const char *'
+
+lib/tilde/tilde.c
+ - default_prefixes and default_suffixes are now arrays of
+ `const char *'
+
+lib/readline/vi_mode.c
+ - vi_motion and vi_texmod are now `const char *'
+
+lib/readline/complete.c
+ - simplified the conditional declaration of getpwent()
+
+variables.[ch]
+ - get_string_value now takes a `const char *' argument
+
+ 11/10
+ -----
+display.c
+ - changes to expand_prompt and rl_redisplay to keep track of the number
+ of invisible characters on the first line of the prompt, in case it
+ takes up more than one physical line, and not all invisible chars are
+ on the first line
+ - the code that determines the index into the line at which to wrap the
+ first line if the visible prompt length exceeds the screen width was
+ changed to use the number of invisible chars in the first physical
+ prompt line rather than the total number of invisible characters
+ - changed redraw_prompt to save and restore prompt_invis_chars_first_line
+
+subst.c
+ - changed expand_string_internal to savestring() the string argument
+ to avoid possible doubly-freed memory on substitution errors
+ - fixed expand_word_list_internal so that commands with assignment
+ statements preceding words that expand to nothing return a failure
+ status if one of the assignments fails; for example
+
+ readonly TVAR
+ TVAR=7 $ECHO $ABCDE
+ - made sure that all cases of jump_to_top_level set
+ last_command_exit_value to something non-zero before jumping, since
+ they are by definition errors
+
+builtins/evalstring.c
+ - changed parse_and_execute to set last_result to EXECUTION_FAILURE
+ along with last_command_exit_value in the `DISPOSE' case of a
+ jump_to_top_level, since we can return this value in most
+ (interactive) circumstances
+
+ 11/14
+ -----
+general.c
+ - fix legal_number so that it skips any trailing whitespace in its
+ string argument after strtol returns, since strtol does not
+
+lib/readline/{complete.c,readline.h}
+ - new function pointer, rl_directory_rewrite_hook, identical in effect
+ to rl_directory_completion_hook except that it does not modify what
+ is displayed when the possible completions are listed or inserted
+
+bashline.c
+ - changed tcsh_magic_space to try and insert a space closer to where
+ the point was originally, rather than always inserting it at the
+ end of the line
+
+ 11/16
+ -----
+jobs.c
+ - fixed a bug that happened on cygwin systems that caused the jobs
+ table to fill up if a file sourced from an interactive shell ran
+ lots of jobs -- notify_and_cleanup() never called notify_of_job_status
+
+subst.c
+ - fixed pat_match to avoid infinite recursion on zero-length matches
+ from match_pattern
+
+ 11/17/2000
+ ----------
+[bash-2.05-alpha1 released]
+
+ 11/18
+ -----
+support/shobj-conf
+ - added GNU Hurd support, identical to linux
+
+variables.c
+ - cygwin systems don't use the exportstr stuff at all, so call
+ INVALIDATE_EXPORTSTR before testing var->exportstr, since just
+ blindly using the value_cell will break exported functions that
+ already have an exportstr set
+ - when recomputing the exported value because valid_exportstr fails,
+ make sure to call named_function_string if the variable is a shell
+ function
+
+ 11/20
+ -----
+variables.c
+ - removed call to valid_exportstr; the exportstr code has been
+ adequately debugged
+
+lib/glob/fnmatch.c
+ - reverted to the POSIX.2 behavior of using the current locale and
+ strcoll() to evaluate range expressions in bracket expressions
+
+doc/{bash.1,bashref.texi}
+ - added note to Pattern Matching section emphasizing that the current
+ locale plays a part in evaluating range expressions within [...]
+
+aclocal.m4
+ - added an #include <unistd.h> to the test for posix sigaction
+ restarting interrupted system calls
+
+support/config.guess
+ - changes from Tim Mooney to support `letter version' releases of
+ Tru 64 Unix on the Alpha (e.g., 5.1a)
+
+ 11/21
+ -----
+lib/sh/pathphys.c
+ - make sure MAXSYMLINKS is defined
+
+aclocal.m4
+ - changed BASH_CHECK_PWFUNCS to just use AC_EGREP_CPP on a file
+ that includes pwd.h and other header files, checking for getpwuid.
+ It #defines HAVE_GETPW_DECLS if the egrep succeeds
+
+lib/readline/complete.c
+ - provide extern declaration of getpwent if _POSIX_SOURCE is defined,
+ since most systems don't declare it in <pwd.h> unless _POSIX_SOURCE
+ is undefined
+
+ 11/22
+ -----
+input.c
+ - fixed a small typo in the cygwin-specific code
+
+doc/{bashref.texi,bash.1}
+ - added some more explanatory text about the effect of the current
+ locale's collating order on range expressions within bracket
+ expressions
+
+doc/bashref.texi, lib/readline/doc/{rltech,rluser,hstech}.texinfo
+ - changed to use @env instead of @code for the names of environment
+ and shell variables in text (not in tables yet)
+
+ 11/27
+ -----
+configure.in
+ - configure without the bash malloc on Interix
+
+lib/readline/doc/rltech.texinfo
+ - changed completion_matches() to rl_completion_matches()
+ - changed filename_completion_function() to
+ rl_filename_completion_function()
+ - changed username_completion_function() to
+ rl_username_completion_function()
+
+lib/readline/rlprivate.h
+ - changed declaration of _rl_alphabetic to rl_alphabetic, since that's
+ what's used in the C files
+
+support/bashbug.sh
+ - Debian has /usr/bin/editor as a link to the system default editor;
+ make that one of the alternatives if $EDITOR is not set
+ - if the OS provides mktemp(1) or tempfile(1), use those to create
+ the temp file for the bug report
+
+lib/readline/{readline.[ch], callback.c}
+ - new function, rl_set_prompt, called with prompt string as argument,
+ sets and, if necessary, expands rl_prompt
+
+lib/readline/{complete.c,compat.c,readline.h}
+ - rl_completion_matches now takes a `const char *' as its first
+ argument
+ - compute_lcd_of_matches now takes a `const char *' as its third
+ argument
+
+lib/readline/rltypedefs.h
+ - a rl_completion_func_t now takes a `const char *' as its first
+ argument
+
+lib/readline/Makefile.in
+ - added dependencies on rltypedefs.h
+
+bashline.c
+ - attempt_shell_completion now takes a `const char *' as its first
+ argument
+ - removed bogus cast when assigning to rl_attempted_completion_function
+ and rl_completion_entry_function
+ - bash_directory_completion_matches takes a `const char *' argument
+
+pcomplete.c
+ - gen_matches_from_itemlist, pcomp_filename_completion_function,
+ gen_action_completions, gen_globpat_matches, gen_wordlist_matches,
+ gen_compspec_completions now take `const char *' arguments as
+ appropriate
+ - programmable_completions now takes `const char *' first and
+ second arguments
+
+pcomplib.c
+ - find_compspec now takes a `const char *' argument
+
+ 11/29
+ -----
+shell.c
+ - in open_shell_script, if we have /dev/fd, don't perform the check
+ for a binary file if isatty(fd) (where fd is opened on the script
+ name) returns non-zero
+ - in open_shell_script, don't call init_noninteractive again, since it
+ can undo changes made by the startup files ($BASH_ENV)
+
+input.c
+ - in fd_to_buffered_stream, make sure buffer size (and hence size
+ arg passed to read(2)) is greater than zero. This makes
+ bash /dev/stdin work right.
+ - return -1 from sync_buffered_stream immediately if buffers is NULL
+ (can happen when running `bash -i /dev/stdin' and executing a
+ command that requires a fork)
+
+ 11/30
+ -----
+jobs.c
+ - renamed cont_signal_handler to sigcont_sighandler for consistency
+ - renamed stop_signal_handler to sigstop_sighandler for consistency
+ - if there is no jobs list or no outstanding jobs, return immediately
+ from notify_of_job_status, before fussing with the signal state
+ - change map_over_jobs to not fuss with the signal state unless there
+ are jobs
+
+lib/malloc/malloc.c
+ - encapsulated the code to block and unblock signals during morecore()
+ execution into two functions: block_signals and unblock_signals
+ - only block signals in morecore() if interrupt_immediately is non-zero
+ or SIGINT or SIGCHLD is trapped
+
+ 12/1
+ ----
+lib/readline/readline.c
+ - fix to rl_forward to make sure that lend doesn't get set < 0, since
+ that will result in rl_point being < 0
+
+ 12/5
+ ----
+parse.y
+ - fix to parse_matched_pair to allow backslash to escape a ' inside a
+ $'...' construct inside another grouping construct like ${...}
+ - translate $'...' and $"..." in parse_matched_pair, because nothing
+ else will do it if such a construct appears inside a grouping
+ construct
+
+ 12/6
+ ----
+configure.in
+ - make sure all bracket tests for pattern matching in case statements
+ are appropriately quoted, so autoconf doesn't swallow the brackets
+ - add -rdynamic to LOCAL_LDFLAGS for freebsd-4.x
+
+examples/loadables/realpath.c
+ - new loadable builtin, canonicalizes pathnames and expands symlinks
+
+builtins/ulimit.def
+ - added undocumented -w option for setting RLIMIT_SWAP, if it's defined
+
+ 12/7
+ ----
+bashline.c
+ - reordered code in initialize_readline so all of the rl_add_defun
+ calls are made before the call to rl_initialize, and all of the
+ rl_{bind,unbind}_key_in_map calls are made after initializing
+ readline. This allows the bash-specific function names to be used
+ in ~/.inputrc
+
+ 12/12
+ -----
+lib/readline/callback.c
+ - in rl_callback_read_char, loop after returning from
+ readline_internal_char and test rl_pending_input in case some
+ function (like the incremental search functions and digit argument
+ functions) pushed input back with rl_execute_next
+
+lib/readline/readline.h
+ - new rl_state flag: RL_STATE_INPUTPENDING, indicating that
+ rl_execute_next has pushed input back for rl_get_key to read
+
+lib/readline/input.c
+ - new function to clear rl_pending_input: rl_clear_pending_input()
+
+lib/readline/{input,readline,signals}.c
+ - when rl_pending_input is set, set state RL_STATE_INPUTPENDING;
+ unset that state when rl_read_key returns the pending input
+ - call rl_clear_pending_input at appropriate points
+
+ 12/15
+ -----
+lib/readline/histfile.c
+ - don't leak file descriptors while attempting to truncate a non-regular
+ file
+
+input.c
+ - make sure check_bash_input sets the close-on-exec flag for the file
+ descriptor used to save the bash input fd
+
+parse.y
+ - if default_buffered_input gets changed while it's on the saved
+ stream stack, make sure it gets set to close-on-exec when it's
+ restored by pop_stream()
+
+ 12/18
+ -----
+builtins/ulimit.def
+ - change set_limit to account for Cygwin's setdtablesize() being
+ declared as a void function
+
+support/config.{guess,sub}
+ - change cygwin32 to cygwin
+
+ 12/20
+ -----
+nojobs.c
+ - make wait_sigint_handler handle the case of interrupt_immediately
+ being non-zero when it's called, and jump out appropriately
+
+jobs.c
+ - new function to run sigchld traps: run_sigchld_trap(); called from
+ waitchld
+ - broke the code that computes the printable job status messages for
+ pretty_print_job out into a separate function
+ - rearranged some code eliminating a useless `else' branch in wait_for
+ - shell_tty_info is now a TTYSTRUCT, using defines from shtty.h
+ - broke the code that sets a job's status after waitpid says its state
+ has changed and cleans up after a received SIGINT out into a separate
+ function: set_job_status_and_cleanup
+ - last_stopped_job() -> job_last_stopped() to avoid collision with a
+ variable name
+ - last_running_job() -> job_last_running() for consistency
+
+ 12/21
+ -----
+variables.c
+ - new function, print_func_list (SHELL_VAR **list); prints shell
+ functions from list in a form that may be re-used as input to
+ redefine the functions
+
+variables.h
+ - new extern declaration for print_func_list
+
+builtins/set.def
+ - set without options now calls print_func_list to have functions
+ printed in a way that can be reused as input (this affects
+ `declare' and `declare -p' also)
+
+ 12/27
+ -----
+general.h
+ - add an ISOCTAL define
+
+builtins/common.h
+ - use ISOCTAL in read_octal function
+
+parse.y
+ - in decode_prompt_string, don't just skip STRING forward by 3 if an
+ octal sequence is translated; handle sequences of fewer than three
+ characters by checking each character
+
+ 12/28
+ -----
+doc/{bash.1,bashref.texi}
+ - make sure `vi-move' is a documented keymap argument to bind's
+ `-m' option
+
+ 12/29
+ -----
+variables.c
+ - change print_assignment so that functions are printed in a way that
+ may be reused as input (without the `=')
+
+ 1/2/2001
+ --------
+test.c
+ - fix test_command to print an error message if `[' is supplied a
+ single argument that is not `]'
+
+support/shobj-conf
+ - change test of $SHOBJ_CC from `gcc*' to `*gcc*' to handle full
+ pathnames
+
+ 1/11
+ ----
+execute_cmd.c
+ - in initialize_subshell, zero out builtin_env, since those aren't
+ supposed to be exported
+
+variables.c
+ - new function: shell_var_from_env_string, creates a SHELL_VAR from
+ a `name=value' style environment string and returns it
+ - new function: bind_name_in_env_array(name, value, array), makes
+ NAME have VALUE in ARRAY, a temporary environment
+ - new function: bind_tempenv_variable (name, value), makes NAME have
+ VALUE in one of the temporary environments -- temporary_env,
+ builtin_env, or function_env
+ - changed bind_variable to call bind_tempenv_variable if one of the
+ temporary environments is non-null before modifying the
+ shell_variables table. This fixes a problem with sourced scripts
+ modifying variables in a temporary environment, but not seeing the
+ updated values
+
+ 1/12
+ ----
+variables.c
+ - changed things so OSTYPE, HOSTTYPE, and MACHTYPE are set only if
+ they don't have a value at shell startup
+
+ 1/17
+ ----
+builtins/fc.def
+ - create temp file with MT_USETMPDIR
+
+variables.c
+ - allow function local readonly variables to overwrite (or shadow)
+ other (calling) function readonly variables, but not global
+ readonly ones
+
+builtins/wait.def
+ - replace unwind_protect of interrupt_immediately, since that's a
+ variable that you cannot unwind-protect (the unwind-protect code
+ uses it)
+
+ 1/22
+ ----
+Makefile.in
+ - changed RELSTATUS to `beta1'
+
+lib/readline/terminal.c
+ - added rl_set_screen_size function, wrapper around
+ _rl_set_screen_size so bash doesn't call _rl_ functions
+
+lib/readline/{rlprivate,readline}.h
+ - moved rl_explicit_arg to readline.h
+ - moved rl_numeric_arg to readline.h
+ - moved rl_editing_mode to readline.h
+ - moved rl_last_func to readline.h
+ - added rl_set_screen_size to readline.h
+
+lib/readline/doc/rltech.texinfo
+ - documented rl_explicit_arg, rl_editing_mode, rl_numeric_arg,
+ rl_last_func
+ - documented rl_set_screen_size
+
+{jobs,nojobs}.c
+ - _rl_set_screen_size -> rl_set_screen_size
+
+{bashline,pcomplete}.c, builtins/set.def
+ - removed extern declarations of variables defined in readline.h
+
+ 1/23
+ ----
+support/rlvers.sh
+ - use $TMPDIR if it's defined
+ - use $RL_INCDIR (set by -I option, /usr/local/include by default)
+ as the directory containing the installed readline header files
+ - cd to the temp directory before attempting compilation
+
+configure.in
+ - pass `-I ${includedir}' to rlvers.sh
+
+ 1/29
+ ----
+subst.c
+ - fixed parameter_brace_expand to only throw away the state of
+ quoted dollar-at if we're going to be using the rhs of the
+ brace expansion, since whatever parameter_brace_expand_rhs sets
+ should be what goes (corrects bad fix of 2/16/2000)
+ - fixed param_expand to only free temp after calling
+ parameter_brace_expand if it's a quoted null and we *did not*
+ have a quoted $@, since in that case we want to keep it and
+ do the $@ special handling (fixes real problem from 2/16/2000)
+
+ 1/30
+ ----
+variables.c
+ - remove export attribute from SSH2_CLIENT, like SSH_CLIENT
+
+ 2/1
+ ---
+unwind_prot.c
+ - added new function, clear_unwind_protect_list, to zero out the
+ unwind_protect list, optionally freeing the elements
+
+unwind_prot.h
+ - new extern declaration for clear_unwind_protect_list
+
+execute_cmd.c
+ - call clear_unwind_protect_list in initialize_subshell. This fixes
+ the problem of core dumps when calling a shell script without a
+ leading `#!' from a shell function or other context with an
+ unwind-protect list
+ - set variable_context and return_catch_flag to 0 in
+ initialize_subshell, since a new subshell is no longer in a shell
+ function
+
+ 2/2
+ ---
+doc/readline.3
+ - updated the text; clarified several points
+ - changed the usage synopsis to include <readline/readline.h> and
+ <readline/history.h>
+
+lib/readline/doc/hstech.texinfo
+ - made sure all function prototypes are valid ANSI C (mostly
+ changing () to (void))
+
+lib/readline/doc/rluser.texinfo
+ - a few clarifications
+
+lib/readline/doc/rltech.texinfo
+ - a few clarifications
+ - added two new subsections for character input and terminal
+ management
+ - changed all function prototypes to be valid ANSI C (mostly
+ replacing () with (void))
+ - documented some variables in readline.h but previously not in
+ the reference manual (rl_num_chars_to_read, rl_dispatching,
+ rl_gnu_readline_p, rl_completion_type)
+ - documented some functions in readline.h but previously not in
+ the reference manual (rl_clear_pending_input, rl_prep_terminal,
+ rl_deprep_terminal, rl_tty_set_default_bindings, rl_alphabetic,
+ rl_crlf)
+
+lib/readline/readline.h
+ - added extern declaration for rl_clear_pending_input()
+ - added extern declaration for rl_alphabetic
+ - changed rltty_set_default_bindings to rl_tty_set_default_bindings
+ - changed crlf() -> rl_crlf()
+
+lib/readline/rltty.c
+ - new function, rl_tty_set_default_bindings, just calls
+ rltty_set_default_bindings (needed `rl_' prefix)
+
+lib/readline/readline.c
+ - readline_default_bindings now calls rl_tty_set_default_bindings
+
+lib/readline/terminal.c
+ - renamed crlf() to rl_crlf(); crlf() is still defined for backwards
+ compatibility
+
+bashline.c, lib/readline/{complete,display}.c
+ - changed all calls to crlf() to rl_crlf()
+
+test.c
+ - fix filecomp() so that `test file1 -nt file2' doesn't fail if file1
+ exists but file2 does not
+
+ 2/5
+ ---
+lib/readline/{terminal,compat}.c
+ - move crlf() backwards-compatibility definition to compat.c
+
+execute_cmd.c
+ - if we're running a command `( command ) &' from a non-interactive
+ shell, don't unconditionally disable expand_aliases. Only do it
+ when running such a command from an interactive shell, since then
+ we're changing to a non-interactive shell
+
+ 2/6
+ ---
+[bash-2.05-beta1 released]
+
+lib/readline/compat.c
+ - add extern declaration for rl_crlf()
+
+lib/readline/undo.c
+ - initialize start and end in rl_do_undo to avoid bogus gcc compiler
+ warning with -Wall
+
+lib/readline/rlprivate.h
+ - fix typo in extern declaration of _rl_current_display_line()
+ - add extern declaration of _rl_free_saved_history_line()
+
+lib/readline/terminal.c
+ - return immediately from _rl_set_screen_size if one of the args is 0
+ - new function, rl_get_screen_size (int *rows, int *cols) to return
+ readline's idea of the screen dimensions
+
+lib/readline/doc/rltech.texinfo
+ - documented rl_get_screen_size
+
+ 2/7
+ ---
+pathexp.c
+ - a couple of fixes for the POSIX_GLOB_LIBRARY version of the globbing
+ code
+
+expr.c
+ - fixed omission of ^= from the special assignment operators
+
+shell.c
+ - change logic for setting `posixly_correct' at shell startup: if
+ $POSIXLY_CORRECT or $POSIX_PEDANTIC is found in the environment,
+ set it to 1, otherwise retain the previous value (which is
+ initialized to 0)
+
+ 2/10
+ ----
+builtins/evalstring.c
+ - since subshell_environment is now a flags word, test for the
+ SUBSHELL_COMSUB bit being set when decided whether or not to
+ evaluate $(<file), not just for subshell_environment ==
+ SUBSHELL_COMSUB
+
+ 2/12
+ ----
+lib/readline/doc/rltech.texinfo
+ - fixed a couple of problems with the documentation for
+ rl_prep_terminal and rl_deprep_terminal
+ - added description of rl_set_prompt
+
+support/config.sub
+ - added support for Linux running on the IBM S390
+
+ 2/13
+ ----
+input.c
+ - new function, fd_is_bash_input (fd), encapsulates the test for
+ whether or not FD is the file descriptor from which bash is reading
+ its input
+ - change check_bash_input to call fd_is_bash_input
+ - new function, save_bash_input (fd, new_fd), handles saving the
+ buffered stream associated with file descriptor FD to one
+ associated with NEW_FD. If NEW_FD == -1, this function allocates a
+ file descriptor itself with fcntl
+ - new function, set_bash_input_fd(fd), sets the fd bash is using to
+ read input to FD
+ - changed duplicate_buffered_stream to note that the buffer it's
+ duplicating to was the bash input buffer (B_WASBASHINPUT)
+
+input.h
+ - extern declarations for fd_is_bash_input and save_bash_input
+ - new flag for b->flag: B_WASBASHINPUT
+
+redir.c
+ - in add_undo_redirect, if FD is 0, make the undo redirect be of type
+ r_duplicating_input instead of r_duplicating_output, since that fd
+ requires special handling in the presence of buffered streams (fixes
+ problem with scripts reading input through a pipe restoring bash
+ input buffered stream correctly after a `read' with an input
+ redirection
+
+ 2/14
+ ----
+lib/readline{complete,util}.c
+ - moved rl_strpbrk to util.c with rest of string utility functions,
+ renamed to _rl_strpbrk
+
+lib/readline/rlprivate.h
+ - extern declaration for _rl_strpbrk
+
+lib/readline/input.c
+ - make the timeout in rl_gather_tyi() be a variable
+ (_keyboard_input_timeout) rather than a constant
+ - new function, rl_set_keyboard_input_timeout (int u) to set
+ _keyboard_input_timeout. The argument is in microseconds; returns
+ the old value of _keyboard_input_timeout
+ - _rl_input_available uses _keyboard_input_timeout also
+
+lib/readline/readline.h
+ - extern declaration for rl_set_keyboard_input_timeout
+
+lib/readline/doc/rltech.texinfo
+ - documented rl_variable_bind(), rl_macro_bind(), rl_macro_dumper(),
+ rl_variable_dumper(), rl_push_macro_input(), rl_show_char()
+
+ 2/20
+ ----
+findcmd.c
+ - new function, executable_or_directory(name), returns 1 if NAME is
+ an executable file or a directory
+
+findcmd.h
+ - extern declaration for executable_or_directory()
+
+bashline.c
+ - call executable_or_directory() from command_word_completion_function,
+ which requires a single stat(2), instead of executable_file() and
+ is_directory(), each of which requires a stat(2)
+
+ 2/21
+ ----
+execute_cmd.c
+ - changed execute_builtin to handle `return' with a preceding variable
+ assignment list specially when in POSIX mode. POSIX says that the
+ variable assignments should persist after the builtin returns, since
+ `return' is a special builtin. Since `return' does a longjmp and
+ doesn't go through the normal cleanup path, we need to do special
+ things
+
+variables.c
+ - new function, merge_function_env(), merges the function_env into
+ the shell_variables table
+
+variables.h
+ - extern declaration for merge_function_env()
+
+execute_cmd.c
+ - changed execute_function to merge any variable assignments preceding
+ the function call into shell_variables if the shell is running in
+ POSIX mode
+
+subst.c
+ - changed parameter_brace_expand_error to do tilde expansion on the
+ rhs of the expansion, if it's not null
+ - changed command_substitute so that the subshell begun to execute
+ the command substitution inherits the setting of the `-e' option
+ when in POSIX mode
+
+builtins/break.def
+ - changed check_loop_level to not print a complaining message when
+ the shell is running in POSIX mode, even if BREAK_COMPLAINS is
+ defined. This affects `break' and `continue'
+ - changed `break' and `continue' to return success even if the shell
+ is not executing a loop, as POSIX specifies
+
+builtins/set.def
+ - changed reset_shell_options() to not clear posixly_correct, so
+ subshells spawned to execute shell scripts without a leading `#!'
+ inherit posix mode from their parent shell
+
+variables.c
+ - changed make_new_variable to create the shell_variables hash table
+ if it hasn't been done yet. This is needed because bind_variable
+ may be called before initialize_shell_variables (e.g., when
+ `bash -o posix' calls bind_variable("POSIXLY_CORRECT", "y") from
+ parse_shell_options -> set_minus_o_option -> set_posix_mode
+ This makes `bash --posix' and `bash -o posix' equivalent, as they
+ should always have been
+ - changed maybe_make_export_env to merge the contents of `builtin_env'
+ into export_env after function_env and before temporary_env. This
+ fixes the problems with variable assignments preceding `eval' or
+ `source' not making it into the environment passed to executed disk
+ commands (var=20 eval printenv var)
+
+doc/bashref.texi
+ - noted that variable assignments preceding shell function calls
+ persist in the shell environment after the function returns when
+ the shell is running in POSIX mode
+ - noted that when the shell is running in POSIX mode, subshells
+ spawned to execute command substitutions inherit the value of
+ the `-e' option from the parent shell
+
+ 2/26
+ ----
+doc/bashref.texi
+ - augmented description of bash non-posix behavior in posix mode
+ with information from running the vsc-lite test suite
+
+ 2/28
+ ----
+general.c
+ - since POSIX.2 says that aliases should be expanded even in
+ non-interactive shells, enable alias expansion in
+ posix_initialize() when posix mode is enabled, and set it to the
+ value of interactive_shell if posix mode is disabled
+
+shell.c
+ - before running the startup scripts, set expand_aliases to
+ posixly_correct, since we don't want alias expansion in startup
+ files unless we've been given the --posix invocation option
+
+doc/bashref.texi
+ - added item in BASH Posix Mode section noting that alias expansion
+ is always enabled when in posix mode
+
+Makefile.in
+ - changed release level to `beta2'
+
+ 3/1
+ ---
+jobs.c
+ - expanded retcode_name_buffer to 64
+ - changed printable_job_status to conform to posix.2 for text of
+ stopped job status: `Stopped(SIGTSTP)' or `Stopped(SIGSTOP)'
+
+doc/bashref.texi
+ - Noted above change in posix mode section
+
+ 3/5
+ ---
+lib/readline/bind.c
+ - changed rl_get_keymap_by_name to work case-insensitively
+
+ 3/6
+ ---
+doc/readline.{0,3,ps}
+ - removed from distribution; moved to readline distribution
+
+lib/readline/doc/history.3
+ - man page for history library; neither built nor installed
+ (part of the standalone readline distribution)
+
+lib/readline/histfile.c
+ - changed history_truncate_file to return 0 on success, errno
+ on failure
+ - changed history_do_write to return errno on a write error. This
+ affects write_history and append_history
+
+lib/readline/history.c
+ - max_input_history is now history_max_entries; the old variable is
+ maintained for backwards compatibility
+
+lib/readline/histexpand.c
+ - the list of delimiters that separate words for history_tokenize
+ is now a variable: history_word_delimiters; default value is as
+ before
+ - changed history_expand to return 0 immediately if its second
+ argument is NULL
+
+bashline.c
+ - changed to use history_max_entries
+
+lib/readline/history.h
+ - extern declaration for history_max_entries
+ - extern declaration for history_word_delimiters
+
+ 3/8
+ ---
+test.c
+ - changed unary_operator() to return FALSE for non-numeric arguments
+ to `-t'
+
+ 3/9
+ ---
+[bash-2.05-beta2 released]
+
+ 3/12
+ ----
+lib/readline/doc/rltech.texinfo
+ - documented rl_set_paren_blink_timeout()
+ - moved the example function to its own info node
+
+lib/readline/{hstech.texinfo,history.3}
+ - `$' is not in the default value of history_word_delimiters
+
+ 3/13
+ ----
+bashline.c
+ - in maybe_add_history, if the tests for HISTCONTROL succeed, but
+ the HISTIGNORE check fails, set first_line_saved to 0
+
+ 3/15
+ ----
+
+lib/glob/fnmatch.c
+ - fixed a bug in patscan that caused `[' to be treated as special
+ even inside a nested bracket expression. It's supposed to lose
+ it's special handling inside []
+ - fixed a bug in patscan that caused `]' to end an embedded bracket
+ expression even if it was the first character after the `[' (or
+ a leading `!' or `^')
+
+ 3/22
+ ----
+execute_cmd.c
+ - made a small change to shell_execve to report a more user-friendly
+ error if execve(2) fails because of a `#! interpreter' error with
+ `interpreter'
+ - in shell_execve, if HAVE_HASH_BANG_EXEC is not defined, make sure
+ the file length is > 2 characters before attempting to execute an
+ interpreter
+
+ 3/26
+ ----
+jobs.c, nojobs.c
+ - give_terminal_to now takes a second argument, FORCE. If that
+ second argument is non-zero, give the terminal away no matter
+ what job_control is set to
+
+jobs.h
+ - changed extern declaration of give_terminal_to
+
+{jobs,sig,subst}.c,parse.y
+ - changed all calls to give_terminal_to to add a second argument
+ of 0
+
+jobs.c
+ - changed second argument of call to give_terminal_to in
+ end_job_control to 1, to force this even if job_control is 0.
+ This fixes the problem of `exec other-job-control-shell' in a
+ startup file hanging, since run_startup_files sets job_control to 0
+
+subst.c
+ - fixed test of subshell_environment in command_substitute to test the
+ bit SUBSHELL_ASYNC rather than test equality, since it's now a
+ bitmask
+
+ 3/27
+ ----
+shell.c
+ - changed posix initialization calls to set POSIXLY_CORRECT=y and
+ then call sv_strict_posix, so $SHELLOPTS includes `posix' and
+ POSIXLY_CORRECT is set, for both the posix command line options
+ and when act_like_sh is non-zero
+
+redir.c
+ - the `>& filename' redirection now works in posix mode, redirecting
+ stdout and stderr, since POSIX.2 leaves it unspecified
+
+doc/bashref.texi
+ - removed item about `>& filename' from Bash POSIX Mode section
+
+parse.y
+ - fixed a problem in decode_prompt_string with `\W' not printing
+ anything if PWD == "//"
+
+lib/readline/input.c
+ - in rl_read_key, after calling (*rl_event_hook), check for rl_done
+ being set, and return '\n' if it is
+
+Makefile.in
+ - changed RELSTATUS to `release'
+
+ 3/28
+ ----
+support/bashbug.sh
+ - replace use of $(...) with `...` for the benefit of old versions
+ of sh
+
+builtins/set.def
+ - changed print_all_shell_variables to not print shell function names
+ and definitions when in posix mode
+
+doc/bashref.texi
+ - noted change in `set' behavior in posix mode section
+
+ 4/2
+ ---
+lib/readline/readline.c
+ - changed default value of RL_LIBRARY_VERSION to `4.2'
+
+ 4/5
+ ---
+[bash-2.05-release frozen]