bash/CWRU/old-changelogs/CWRU.chlog.v8

[ Work begins after network release of version 1.11 ]

				     1/11
				     ----
nojobs.c
	- Posix systems now reap zombie children and retry a fork() once if
	  it fails
	- Posix systems should use waitpid() rather than wait() wherever
	  possible
	- Posix systems do not need to validate a pid before calling waitpid()
	  in wait_for_single_pid(); waitpid takes a pid argument

				   1/13
				   ----
execute_cmd.c
	- get_next_path_element can return NULL, so the code that calls it
	  must take that possibility into account

jobs.c
	- there was an extra parameter in the select() call

builtins/type.def
	- successful_finds needed to be initialized to 0

machines.h
	- fixed a typo (_D -> -D) in the DG/UX machine description

				   1/14
				   ----
execute_cmd.c
	- fixed extract_colon_unit to return "" in the case of a trailing
	  colon in the path

INSTALL
	- note that on SCO Xenix 386 one must use cc -E rather than /lib/cpp
	  to process cpp-Makefile

cpp-Makefile
	- fixed the problem of a single quote in a makefile comment

machines.h
	- Xenix 386 machines need -DUSG in SYSDEP_CFLAGS

lib/readline/readline.c
	- changed the includes around so that the Xenix 386 support is
	  in the same section as the USGr3 code, and the Xenix 286
	  support is in the same section as the plain USG code

shell.c
	- split the `int code = setjmp (top_level)' statement in
	  reader_loop into two statements -- some compilers don't like
	  it

parse.y
	- changed the overloaded `yy_input_dev' to a `union STREAM',
	  where a `union STREAM' is

		typedef union STREAM {
			FILE	*s_file;
			char	*s_string;
		} STREAM;

	  and changed the parameter to init_yy_io and all the functions
	  that call it to use a STREAM argument instead of casting back
	  and forth between a (char *) and a (FILE *)

builtins/times.def
	- If hpux or USGr4, #undef HAVE_RESOURCE, rather than try to fit
	  all the special cases onto a single line deciding whether or
	  not to include <sys/resource.h>

				   1/15
				   ----
bashline.c
	- changed the bindable name for the ksh-style ^O function from
	  operate_and_get_next to operate-and-get-next

execute_cmd.c
	- some systems (e.g SGI) allow the result of alloca to be assigned
	  only to a `simple' variable, so I introduced a dummy one instead
	  of assigning to redirectee->word directly

shell.c
	- fixed a typo: PENDANTIC -> PEDANTIC

machines.h
	- took -DREVERESED_SETVBUF_ARGS out of the entries for Xenix 386

cpp-Makefile
	- added double quotes around the RHS of the assignment to
	  SYSTEM_NAME to avoid further cpp processing if the system name
	  happens to be a cpp define

bashline.c
	- added `search-forward', `search-backward' functions that perform
	  non-incremental history searches using the vi-mode code

				   1/16
				   ----
builtins/ulimit.def
	- fixed a typo in a comment
	- added parens around the object in an #if defined preprocessor
	  statement

machines.h
	- the Xenix 386 machine descriptions need to #undef HAVE_GETWD

builtins/read.def
	- Fixed read to treat backslash as an escape character unless
	  -r is given, as per Posix.2

lib/readline/readline.c
	- Fixed up the maze of dire[cn]t includes and defines so that
	  they're correct for Xenix (finally)

lib/glob/glob.c
	- ditto.  Now the defines/includes are correct for Xenix 386

execute_cmd.c
	- a loop of the form
			while : ; do echo -n 1; done
	  can run a machine using Gwyn's alloca() emulation out of memory
	  because alloca() keeps getting called at the same stack level
	  and never frees anything up.  Added a call to alloca(0) after
	  the call to execute_simple_command in execute_command_internal
	- added Karl Kleinpaste's workaround for the AFS file creation
	  bug, dependent on AFS_CREAT_BUG

parse.y
	- \s in a prompt string needs to decode to the shell basename, as
	  per the documentation, rather than the full shell name

				   1/17
				   ----
nojobs.c
	- fixed a typo in one of the WAITPID calls, thanks to Bruce Evans

[At this point, all of these fixes were sent along to bfox]

support/getcppsyms.c
	- now recognizes cpp symbol __hppa for use on HP precision
	  architecture machines

machines.h
	- M_MACHINE is now `hppa' for HP Precision Architecture machines, so
	  to make sure hpux is defined, it's now in SYSDEP_CFLAGS
	- Make sure that the RS6000/AIXv3 description defines M_OS as AIX,
	  not "AIX".  The quotes are added in cpp-Makefile

				   1/27
				   ----
parser.h
	- new file, contains structures necessary to parse commands

input.h
	- new file, contains definitions and structures used to read input

bashline.c
	- removed the definitions for search-backward, search-forward

cpp-Makefile
	- removed the definition of DESTDIR; now passed down from Makefile

error.c, error.h
	- new files to isolate the error-handling functions

command.h
	- new file including structures used internally to represent
	  commands

parse.y
	- include input.h, parser.h
	- changed all the code that deals with input streams to fit the
	  new framework defined in input.h

make_cmd.c
	- moved the error reporting code to error.c
	- added cpp code to ensure that alloca is defined correctly

shell.h
	- removed the code that was moved to input.h, parser.h, command.h

builtins/common.c
	- moved a call to setjmp out of an if statement
	- changed the input code to work with the new input framework

builtins/getopt.c
	-include config.h if compiling for the shell

lib/readline/readline.c
	- change representation of characters > 127 to octal
	  rather than M-c
	- changed representation of characters < ' ' to ^X rather
	  than C-x
	- fixed a memory leak in parser_if () by freeing `tname'

machines.h
	- removed MIPS_CFLAGS from the ultrix Decstation entry, since the
	  limit no longer needs increasing

variables.c
	- move the initialization of PWD before the environment is read, so
	  an exported version of PWD takes precedence

readline.c, funmap.c, emacs_keymap.c
	- removed support for rl_arrow_keys -- it should now be done by the
	  inputrc file

documentation/bash.1
	- only non-job-control shells start background jobs ignoring SIGINT
	  and SIGQUIT

builtins/umask.def
	- don't print the octal representation of the umask if the symbolic
	  representation has already been displayed
	- a bad option to umask should return EXECUTION_FAILURE, not -1

shell.c
	- new function, init_signal_handler (sig), to initialize the SIG
	  signal hander, call it in main loop

nojobs.c
	- call init_signal_handler (SIGINT, sigint_sighandler) so the correct
	  thing happens on Posix systems

general.c
	- updated the comment above the fallback implemenation of killpg()

machines.h
	- machine entry for the Unix PC
	- new variable HAVE_DIRENT

				   1/28
				   ----
machines.h
	- SCO needs to #undef HAVE_GETCWD so that the getcwd() function
	  supplied with bash is used

subst.c
	- USG machines should include <string.h>, others <strings.h>
	- sub_append_number should use itos() rather than xmalloc and
	  sprintf

builtins/ulimit.def
	- changed the resource limit struct to one that has no explicit
	  size to account for systems with differing numbers of resource
	  limits

lib/readline/readline.c
	- there was an off-by-one error in the code that counts the number
	  of items to list when doing completion listing