Fresh Fish 6/FreshFish_September1994.bin/bbs/gnu/pdksh-4.9-src.lha/GNU/src/amiga/pdksh-4.9/sh/ChangeLog

Text File | 1994-05-04 | 13.0 KB | 396 lines

Tue Aug 31 20:32:54 1993  Simon J. Gerraty  ([e-mail])

    * patch from [e-mail] to avoid problem in findcom().

Tue Jun  1 23:36:36 1993  Simon J. Gerraty  ([e-mail])

    * Quick hack to "handle" multi-line prompts better.
    While in x_vi or x_emacs we use only the last line of the prompt.

Tue Jun  1 20:52:04 1993  Devetir Unix section <[e-mail]>

    * Serveral bug fixes from: Devetir Unix section
    <[e-mail]>

    Sorry for the delay in these, but they were stuck in transit
    because zen was dead for nearly 3 months!  --sjg

    1) In history.c:hist_init(), an extra file descriptor to the
    history file is created which can remain open to the old history
    file when the history needs shrinking.  Thus, the old history file
    survives with no name, and caused me trouble under 386BSD when
    shutting down suddenly (fsck had to clean it up and have another
    re-boot).

    2) In jobs.c:exchild() changed order of setpgid() and tcsetpgrp()
    as 386bsd at least does not allow tcsetpgrp() to an unused
    process group.

    3) bug in jobs.c:j_reapchld(), waitpid() can return 0 if there
    are no children, and even though there was code there to exit the
    loop on pid == 0, the variable sigchld_caught was not being set to
    0, so j_reapchld() would be continually called from j_waitj()
    wasting considerable cpu time.

    4) In jobs.c:j_resume(), fg of a job that has exited, but not yet
    been flushed from pdksh's list of jobs caused the controlling
    terminal's process group to be stuffed up, and pdksh to exit, on
    some machines.


Sat May 22 21:58:49 1993  Simon J. Gerraty  ([e-mail])

    * mail.c, expr.c: bug fixes from: Carsten Heyl
    <[e-mail]>

    * eval.c: bug fix from: Peter Collinson <[e-mail]>


Sat May  8 15:48:16 1993  Simon J. Gerraty  ([e-mail])

    * Ported to 386bsd (new base system :-), only sh directory
    involved though.  This mainly involved re-arranging some #if's so
    that if _POSIX_TERM is defined it will take precedence over _BSD

Sat Dec  5 23:54:40 1992  Simon J. Gerraty  (sjg@zen)

    * finally fixed the fd leakage that caused pipes to not terminate
    immediately.

Sat Nov 28 23:01:38 1992  Simon J. Gerraty  (sjg@zen)

    * Makefile: added install target

Wed Nov 25 09:10:11 1992  Simon J. Gerraty  (sjg@zen)

    * c_cd(c_ksh.c): fixed bug causing cd ., cd .. to look at $CDPATH.

Thu Nov 19 23:32:42 1992  Simon J. Gerraty  (sjg@zen)

    * For now bind emacs mode complete-list to ESC=

Tue Sep 15 15:24:38 1992  Simon J. Gerraty  (sjg@zen)

    * main.c: call init_editmode() between processing .profile and
    .kshrc so that edit mode can be explicitly set in .kshrc
    regardless of FCEDIT, EDITOR and VISUAL values.

Tue Sep 15 15:08:06 1992  Philippe Michel ([e-mail])

    * main.c: fixed bug in command line argument processing.

Fri Aug 21 15:24:37 1992  Bruce Momjian  ([e-mail])

    * exec.c: replace all calls of execute(tree, 0) with
    execute(tree, flags & XXWHL) to ensure it is passed to children.

Thu Aug 20 23:18:45 1992  Simon J. Gerraty  (sjg@zen)

    * fixed bug in complex_history: histsize not initialized before
    call to alloc().

Thu Aug 13 00:00:42 1992  Simon J. Gerraty  (sjg@zen)

    * setstr(), varsub(): be robust when given dud args.

Wed Aug 12 23:56:29 1992  Simon J. Gerraty  (sjg@zen)

    * Fixed build problem when COMPLEX_HISTORY is not defined.
    * don't make COMPLETE_LIST the default, add
        bind '^[^['=complete-list
    in .kshrc if desired.


Mon Aug  3 22:41:17 1992  Simon J. Gerraty  (sjg@zen)

    * emacs.c: correctly bind <ESC><erase>.
    * var.c: treat COLUMNS and FCEDIT as special.

Sat Aug  1 17:17:02 1992  Simon J. Gerraty  (sjg@zen)

    * Incorporated massive contribution from Peter Collinson
    includes new features (refer to ../Changes.pc) and support for
    BSDI's BSD/386.

    * emacs.c: new command complete-list provided by
    [e-mail] this nicer than the standard ksh
    file completion.  Define COMPLETE_LIST to bind this to <ESC><ESC>
    by default.

Sat May 30 15:44:56 1992  Simon J. Gerraty  (sjg@zen)

    * added flag XXWHL in tree.h, used by execute() when calling
    exchild() to indicate that stdin should not be invalidated.  This
    corrects handling of:
         ls | while read f; do ls -l $f; done

    Thanks to Bruce Momjian for tracking down the fault.

Tue May 12 19:23:17 1992  Simon J. Gerraty  (sjg@zen)

    * Fix bug in init_editmode() if EMACS and VI are not both defined.

Sun May  3 17:47:54 1992  Simon J. Gerraty  (sjg@zen)

    * sigact.c:  allow force build if USE_* defined.
    * main.c: call init_editmode() _after_ processing /etc/profile.
    * jobs.c: ensure SA_RESTART is defined.

Sat Apr 25 00:20:51 1992  Simon J. Gerraty  (sjg@zen)

    * Implemented a simple history file.
    The default file is "$HOME/.pdksh_hist" but can be changed by
    setting HISTFILE in any of /etc/profile,.profile or .kshrc.
    The format is trivial - one line per history item.
    You can creat a set history file, by making it read-only.
    History is only saved during wrap-up and only if the file is
    writeable.  NOTE: if writeable the file is OVERWRITTEN.

Fri Apr 24 22:22:04 1992  Simon J Gerraty  (sjg@zen)

    * sigact.c: New file.
    An implementation of sigaction() and freinds, which simply
    interfaces to BSD's setsigmask() et al, BSD4.1's setsig() or plain
    old signal(2).  It attempts to use the most useful available.
    If it thinks the system really has sigaction() it compiles to
    noting, and its header sigact.h does nothing.
    At present all use of sigaction is bounded by USE_SIGACT, if/when
    this approach appears to work, the alternatives will be withdrawn.

Fri Apr 24 10:42:40 1992  Simon J Gerraty  (sjg@taureau)

    * Added support for sigaction and friends!, turns out sigset()
    et al just don't handle job control.  Define USE_SIGACT if you
    have sigaction().  Job control now works on this System V machine
    (Bull DPX/2).

    * getsc_(lex.c), x_getc(edit.c):  Added goto's! to allow the
    read() calls to be retried (on systems that don't do it
    automatically) if the read() was interrupted by a SIGCHLD
    (sigchld_caught > 0).  This was the cause of the problem with an
    async child terminating the shell on System V.

Wed Apr 22 14:57:01 1992  Simon J Gerraty  (sjg@taureau)

    * exchild(jobs.c): block SIGCHLD when fork()ing, until safe to
    receive.

    * Added support? for sigset() and friends.  Allow JOBS to work on
    System V machines.  This does NOT work right yet.
    With JOBS defined, an async process terminates the shell when it
    (the child) exits, and ^Z does nothing!

Tue Apr 21 15:18:08 1992  Simon J. Gerraty  (sjg@sun0)

    * removed const qualifyer from setctypes() 1st arg, as gcc-2.1
    generates incorrect code for this function.  This is a temperary
    hack until gcc is fixed.

    * do not install std/stdc/stdarg.h in std/h, any compiler that
    will use stdarg.h should have one.

Sun Apr 19 20:16:32 1992  Simon J. Gerraty  (sjg@zen)

    * added support of stdargs to tree.c and _fixed_ the use of
    varargs.

Sat Apr 18 16:35:48 1992  Simon J. Gerraty  (sjg@zen)

    * x_read(edit.c): added logic to check $EDITOR etc.


Fri, 10 Jan 92 13:13:52      Bullseye Software  (sumax!polari!bullseye)

    * alloc.c:afreeall() references memory it has already deallocated.

Wed, 1 Jan 92 20:03:55    Bert Gijsbers ([e-mail])

    * expand(eval.c):  PS1='${PWD##/*/}[!]% ' prints also characters
    with ascii = 128

    * main(main.c): Testing whether the first letter of argv[0] is a
    '-' does not work when ksh reads .kshrc.

Sun, 29 Dec 91 20:05:02  Bert Gijsbers ([e-mail])

    * Support for Minix.

Wed, 11 Dec 91 12:41:31    Kees J. Bot ([e-mail])

    * fixes for sun3 with 4.1.1

Wed, 4 Dec 91 15:53:43  Reg Quinton ([e-mail])

    * fixes for SGI

Mon Nov 25 12:36:42 1991  Simon J. Gerraty  (sjg at zen)

    * stdh.h:  make sure FD_CLEXEC is defined if we support F_SETFD.

    * emacs.c: Attempt to make alloc() of x_tab acceptible to more
    compilers.  Not 100% yet.

Sat Nov 23 14:31:44 1991  Simon J. Gerraty  (sjg at zen)

    * Improved fd_clexec handling for systems that don't have
    an F_SETFD fcntl().  The new arrangement will not blow up if an
    attempt is made to fd_clexec a fd above MAXFD (64 by default).
    main.c:main()  and io.c:savefd() now simply call
    exec.c:fd_clexec().

Fri Nov 22 11:24:57 1991  Simon J. Gerraty  (sjg at zen)

    * Since many people had problems using the headers and libraries
    in std/*, modified sh/* so that they can be compiled in the
    absence of std/*.  This requires putting in some fixes that I had
    left out from my 3.2 version.  Particularly, catering for systems
    that do not have F_SETFD.

    * exec.c:
    Added fd_clexec array for tracking fd's to close in child after
    fork().  This is avoid wasting fd's on systems that don't have
    F_SETFD.

    * jobs.c:
    Allow Sun's and perhaps other BSD systems to define WAIT_T to be
    union wait and thus use their native status handling for children.

    * The file sh/MACHINES now tracks systems the shell has been
    compiled on.


Sat Nov  9 14:57:30 1991  Simon J. Gerraty  (sjg at zen)

    * Release version 4.1 as a new base line.

Thu Nov  7 23:11:25 1991  Simon J. Gerraty  (sjg at zen)

    * John R MacMillan suppied a fix for a bug in yylex() that was the
    cause of several odd problems such as:
        $ foo=echo
        $ $foo bar
        bar: not found
        $ pwd
        $ /local/src/pdksh
        $ $foo bar
        bar
        $

Sun Sep 15 23:19:27 1991  Simon J. Gerraty  (sjg at zen)

    * emacs.c:
    Added function x_lastcp() which returns a pointer to that char in
    the edit buffer that will be the last displayed on the screen.
    Thus:

          cp = x_lastcp();
      while (cp > xcp)
            x_bs(*--cp);

    Will correctly position the cursor on the screen (regardless of
    TABs etc in the buffer).  The previous method got out of sync if
    there were any TABs to display.

Wed Aug  7 11:26:55 1991  Simon J. Gerraty  (sjg at sun0)

    * jobs.c:
    The Sun SPARCstation 2 was proving extremely unreliable using ksh.
    After puting traces in jobs.c, it turns out the problem was
    multiple SIGCHLD events happening too quickly causing the handler
    to be interupted and thus not recognise the job that it had just
    reaped was one of its own.  Having done the waitpid(), but not
    adjusted the job table entry, j_waitj() would loop forever waiting
    for a job to finnish (that had already done so!)
    Solution was to have the SIGCHLD handler simply record the events
    by inrcrementing sigchld_caught.  The actual reaping is now done
    in a new funtion j_reapchld(), which does what the old signal
    handler did but blocks SIGCHLD while scanning the job table.
    j_waitj() calls j_reapchld() when sigchld_caught is non-zero.
    The SS2 is now much more reliable...

    * trace.c:
    Added my simple _TRACE facility (used to track the j_waitj
    problem).  Simply -DUSE_TRACE for it to have effect.  If USE_TRACE
    is undefined, calls to _TRACE() expand to while(0) which an
    optimizer will usually remove.  sh.h now includes trace.h

Mon Jun 10 10:27:14 1991  Simon J. Gerraty  (sjg at zen)

    * emacs.c:
    A couple of assignments (xbp = xbuf) were not migrated from the
    3.2 version.  Caused an anoying bug when retrieving history
    commands.

Mon May 27 12:50:20 1991  Simon J. Gerraty  (sjg at sun0)

    * added fixes supplied by Mike Jetzer:
    These relate mainly to vi mode.  See Changes.mlj

    * c_sh.c c_exit():
    Modified behavior to not imediately exit if there are stopped
    jobs.  A subsequent exit will kill any jobs left and terminate the
    shell.

Fri May 24 15:20:10 1991  Simon J. Gerraty  (sjg at sun0)

    * edit.h:
    Cleaned up prototypes.  Built shell on sun3.
    While gcc-1.39 builds the ksh ok on the 386i, on the sun3 jobs
    don't work correctly - any non-builtin command gets stopped and
    put into the background.  Had same problem with 3.2, using
    /usr/bin/cc works fine.

Thu May 23 13:45:20 1991  Simon J. Gerraty  (sjg at sun0)

    * migrated my 3.2 edit.c changes to the new shell.
    Affects edit.c, emacs.c
    Added edit.h which is now included by edit.c,emacs.c and vi.c

    * vi.c:
    Fixed handling of '!' in prompt by using pprompt() as in emacs.c

    * std/stdc/vprintf.c:
    Fixed bug in output of left '0' padded unsigned numbers was
    always padding with ' ' which left a space in ksh's temp file
    names.  This prevented fc -l from working.

Here is my 3.2 ChangeLog:
Fri Mar 22 16:50:14 1991  Simon J. Gerraty  (sjg at sun0)

    * edit.c:
    Added x_set_arg() and x_prev_histword().
    x_set_arg() handles 'ESC''0-9' type args which are used by word
    related commands.
    x_prev_histword() recovers the last (default) or sepcified arg
    word from the previous command line.  Bound to ESC. and ESC_ to be
    compatible with real ksh.

Tue Feb 26 14:16:17 1991 Simon J. Gerraty  (sjg at zen)

    * edit.c:
    Changes to handle editing of command lines longer than $COLUMNS in
    a manner compatible with real ksh.

Mon Feb 25 12:20:36 1991 Simon J. Gerraty  (sjg at sun0)

    * var.c,table.h:
    Implemented $RANDOM
    Some scripts use [ "$RANDOM" != "$RANDOM" ] to check for ksh.

Wed Feb 20 12:20:36 1991 Simon J. Gerraty  (sjg at sun0)

    Changes so that shell will compile on sun386i.

    * exec.c,main.c,io.c:
    Handle the case where FD_CLEXEC isn't defined.

    * jobs.c:
    SunOS has its own ideas about job status etc.

    * tree.c:
    Fixed conflict between varargs and stdarg.