bug-reports

一个开放源代码的 AT&T 的 Korn Shell 的复制品, 支持大多数 ksh89 的特性。

  (xxx-exec-environment-1 and 2) fail because printenv isn't found.
  [see Mail.XXX:XXX]
  [fixed in 5.2.9: changed test to use env instead]

x pdksh 5.2.8, - (reported by Stefan Dalibor): shell assumes 80 columns when
  it starts up if COLUMNS is set correctly in the environ.
  [see Mail.XXX:XXX]
  [fixed in 5.2.9: fixed so window size is checked at startup]

x pdksh 5.2.8, NeXT machines (reported by Kai Wong): clock_t, which lives
  in sys/time.h, isn't found by configure (causes warning messages).
  [fixed in 5.2.9: configure now checks in sys/time.h]

x pdksh 5.2.3, - (reported by Mike Jetzer): in vi, <ESC>= on word with ~
  but no /, beeps (or prints final path comonent?).
  [see Mail.9:66]
  [fixed in 5.2.9: fixed edit.c(add_glob) so no * is appended to ~username]

x pdksh 5.1.3, NeXT machines (reported by Jason Baugher): job control doesn't
  work on NeXT machines (both m68k and x86 based) in rlogin sessions.
  (caused by open("/dev/tty") failing - rlogin on NeXT doesn't set up
  controlling tty properly).
  [see Mail.7:29]
  [fixed in 5.2.9: added hack to main to get a controlling tty]

x pdksh 5.2.8, NeXT 3.2 (reported by Andrew S Townley): the output of the
  siglist.sh script fills the disk.  Also, the signal list generated (by the
  fixed script) is mostly empty.
  [see Mail.XXX]
  [fixed in 5.2.9: fixed siglist.sh script to avoid infinite loops.  Added
   comment to README warning of problem with NeXT's native cc -E]

x pdksh 5.2.9, - (reported by Loris Talpo): long prompts are messed up in
  vi mode.
  [see Mail.XXX]
  [fixed in 5.2.9: lex.c(pprompt) was broken]

x pdksh 5.2.9, - (reported by Will Renkel): a double backslash followed
  by a newline in an unquoted here document results in one of the backslashes
  and the newline being stripped.
  [see Mail.XXX]
  [fixed in 5.2.10: fixed backslash-newline processing in lex.c]

x pdksh 5.2.9, - (reported by Han Holl): read x?prompt doesn't work 
  on non-interactive shells when the input is from a tty.
  [see Mail.XXX]
  [fixed in 5.2.10: changed test in c_read() from FTALKING to isatty]

x pdksh 5.2.10, - (reported by Dale DePriest): expanding aliases causes
  extra input.
  [see Mail.XXX]
  [fixed in 5.2.11: fixed syn.c(c_list).]

x pdksh 5.2.10, - (reported by John Rochester): window size changes don't
  happen on Dec unix (osf) because TIOCGWINSZ is defined in <sys/ioctl.h>
  not in <termios.h>.
  [see Mail.XXX]
  [fixed in 5.2.12: sys/ioctl.h included with termios.h/termio.h if possible]

x pdksh 5.2.11, - (reported by Randy Bouzane): aliases in command substitutions
  result in code dumps.
  [see Mail.XXX]
  [fixed in 5.2.12: lex.c(getsc__): break if eof is read]

x pdksh 5.2.11, - (reported by Will Renkel): aliases containing ;<newline>
  or <newline><newline> aren't fully read/executed.
  [see Mail.XXX]
  [fixed in 5.2.12: syn.c(get_command): call inalias()]

x pdksh 5.2.11, SGI/IRIX 5.2 (reported by [e-mail]): pipelines
  containing built in commands hang forever.
  [see Mail.XXX]
  [fixed in 5.2.12: fixed pgrp sync code in jobs.c]

x pdksh 5.2.11, - (reported by Herbert Thielen via Larry Daffner): shell leaves
  temp files around when executing shell scripts that have functions with
  here documents which end in an exec.
  [see Mail.XXX]
  [fixed in 5.2.12: call main.c(cleanup_proc_env()) from exec.c(execute()).]

x pdksh 5.2.12, - (reported with fix by Eric J. Chet via Thomas Gellekum):
  . can cause core dump when cleaning up.
  [see Mail.XXX]
  [fixed in 5.2.13: call shf_close() before quitenv()]

x pdksh 5.2.12, - (reported by Bruce Burns): test with a single -t option
  always returns true (does a string test instead of isatty(1)).
  [see Mail.XXX]
  [fixed in 5.2.13: do the isatty(1) unless in posix mode]

x pdksh 5.2.12, - (reported by David Tamkin): aliases ending in "; "
  cause continuation prompt to be printed.
  [see Mail.XXX]
  [fixed in 5.2.13: syn.c(c_list) now handles blank lines, removed cf=CONTIN
   in syn.c(get_command)]

x pdksh 5.2.12, - (reported by Herbert Thielen via Larry Daffner): set
  does not allow +o and -o options in the same command line.
  [see Mail.XXX]
  [fixed in 5.2.13: changed ksh_getopt to remove exclusion code.]

x pdksh 5.2.12, - (reported by Han Holl <[e-mail]>):
  set -A foo -- bar doesn't skip the --.
  [see Mail.XXX]
  [fixed in 5.2.13: changed misc.c(parse_args) so A takes an option]

x pdksh 5.2.12, - (reported by David Tamkin): -e (errexit) should be ignored
  when processing profile and ENV.
  [see Mail.XXX]
  [fixed in 5.2.13: errexit saved/cleared/restored in main()]

x pdksh 5.2.12, - (reported by Han Holl <[e-mail]>):
  [ -x foo ] succeeds for root if foo exists.
  [see Mail.XXX]
  [fixed in 5.2.13: changed c_test.c(test_access) to use stat in this case]

x pdksh 5.2.4, - (reported by Gabor Zahemszky): echo ${foo[*]#/} generates
  bad substsitution error, newer ksh's don't (older ones do);
  error includes {...#@(/)}.
  [fixed in 5.2.13: moved the @(..) hack from yylex() to expand()]

x pdksh 5.2.8, - : extended pattern globing doesn't handle nested parens (),
  e.g., [[ aby = +(a|b(x|y)) ]]

x pdksh 5.2.12, - (reported by Marc Olzheim <[e-mail]>):
  "echo a | (echo b | echo c)" causes a core dump.
  [see Mail.XXX]
  [fixed in 5.2.13: clear XPIPEI,XPIPEO at start of jobs.c(exchild)]

x pdksh 5.2.12, - (reported by Curt Finch <[e-mail]>): in an interactive
  shell, globbing isn't done on redirections if the command is part of
  a pipeline.  e.g., in "cat < /tmp/*gz | grep foo", the /tmp/*gz is
  not expanded.
  [see Mail.XXX]
  [fixed in 5.2.13: clear XPIPEI,XPIPEO at start of jobs.c(exchild)]

x pdksh 5.2.12, - (reported with fix by Greg A. Woods <[e-mail]>):
  in emacs, ^[_ (aka ^[.) gets the word from the previous line relative to
  where the current line came from in the history.  It should get it from
  the last executed line ala at&t ksh.
  [see Mail.XXX]
  [fixed in 5.2.13: use absolute last command]

x pdksh 5.2.12, - (reported by Gabor Zahemszky): MAILCHECK and MAIL
  don't report `new mail' unless MAIL is re-assigned.
  [see Mail.XXX]
  [fixed in 5.2.13: mail code was using variables memory, which was later
   trashed by exporting the MAIL variable.  Fixed by saving copy of value.]

x pdksh 5.2.12, - (reported by Gabor Zahemszky): memory gets badly fragmented
  when reversing a (large) file using a simple while read loop and variable
  contatenation.
  [see Mail.XXX]
  [fixed in 5.2.13: alloc.c changed to allow malloc() to deal with large
   blocks.]

x pdksh 5.2.12, - (reported by Bernd Eggink <[e-mail]>)
  ksh style functions don't save/reset/restore OPTIND.
  [see Mail.XXX]
  [fixed in 5.2.13: save and restore user_opt for ksh-style functions.]

x pdksh 5.2.12, - (reported by Marc Olzheim <[e-mail]>)
  ${..%..} stuff don't work in SH mode.
  [see Mail.XXX]
  [fixed in 5.2.13: removed ifdef KSH from misc.c(do_gmatch)]

x pdksh 5.2.12, - (reported with fix by George Robbins
  <[e-mail]>)
  in sh, "exec 3>&1" does not keep fd 3 open in executed commands.
  [see Mail.XXX]
  [fixed in 5.2.13: c_sh.c(c_exec): added ifdef KSH around fd_clexec() call]

x pdksh 5.2.12, - (reported with fix by George White
  <[e-mail]>)
  in remove_temps(main.c), space for name field is not allocated correctly.
  [see Mail.XXX]
  [fixed in 5.2.13: initialize t->name in new structure]

x pdksh 5.2.12, - (reported with fix by Marc Olzheim <[e-mail]>):
  when at (past) end of the line, word/command completion will skip back
  [see Mail.XXX]
  [fixed in 5.2.13: edit.c(x_locate_word): delete special handling of
   end-of-buffer]

x pdksh 5.2.12, - (reported by Mike Kelly <[e-mail]>):
  doting a directory (or a empty path) is allowed.
  [see Mail.XXX]
  [fixed in 5.2.13: exec.c(search_access): don't do non-regular files for R_OK]

x pdksh 5.2.13, - (reported by Mike Kelly <[e-mail]>):
  typeset -f FUNC doesn't print follows command (and expression) substitutions.
  [see Mail.XXX]
  [fixed in 5.2.14: tree.c(tputS): add wp++]

x pdksh 5.2.13, - (reported by Thomas Gellekum):
  make check fails on freebsd with "chmod 644 abcx failed - Inappropriate ...".
  [see Mail.XXX]
  [fixed in 5.2.14: make test/th convert permissions to octal]

x pdksh 5.2.13, - (reported with fix by David E. Wexelblat):
  when re-allocating memory, too much may be copied from old memory.
  [see Mail.XXX]
  [fixed in 5.2.14: use min old old size and new size]

x pdksh 5.2.13, - (reported with fix by David E. Wexelblat):
  set -o printed some options sans names.
  [see Mail.XXX]
  [fixed in 5.2.14: use 0 instead of null in options[] table]

x pdksh 5.2.13, - (reported by Gabor Zahemszky):
  emacs mode: <esc>. in very fist command causes core dump.
  [see Mail.XXX]
  [fixed in 5.2.14: ring bell if no history in emacs.c(x_prev_histword)]

x pdksh 5.2.13, - (reported by Keith S McCabe): pdksh dumps core
  after a cd command.
  [see Mail.XXX]
  [fixed in 5.2.14: exec.c(flushcom) was setting bits in table entry, instead
   of clearing a single bit]

x pdksh 5.2.12, - (reported by Jaime A. Urquidi): typeset -i reports
  on array elements that have no value (at&t ksh reports on array
  base name - no index).
  [see Mail.XXX]
  [fixed in 5.2.14: hack to c_ksh.c(c_typeset) to generate the ksh88 style
   output]

x pdksh 5.2.13, - (reported with fix by Todd C. Miller):
  ulimit -ctn unlimittttted kills shell (resource exceeded).
  [see Mail.XXX]
  [fixed in 5.2.14: hacked c_ulimit to generate error val is 0 and expr doesn't
   start with a digit]

x pdksh 5.2.13, - (reported with fix by Theo de Raadt):
  ". /dev/null" says access denied.
  [see Mail.XXX]
  [fixed in 5.2.14: exec.c(search_access): allow non-regular file if reading]

x pdksh 5.2.13, - (reported with fix by Eric Youngdale): flag field in aliases
  incorrectly changed (all flags set instead of clearing ISSET) in
  exec.c(flushcom).
  [see Mail.XXX]
  [fixed in 5.2.14: exec.c(flushcom): change = ~ISSET to &= ~ISSET]

x pdksh 5.2.13, - (reported by Andre Delafontaine): ${#array[*]} prints
  largest index instead of number of (set) elements in an array (ksh88 does
  the former).
  [see Mail.XXX]
  [fixed in 5.2.14: eval.c(varsub): count number of elements]

x pdksh 5.2.13, - (reported by Clifford Wolf): sys_siglist[] doesn't
  always have NSIG non-null entries...
  [see Mail.XXX]
  [fixed in 5.2.14: trap.c(inittrap): check for null sys_siglist entries.]

x pdksh 5.2.13, - (reported with fix by Todd C. Miller): waitfor in jobs.c
  can cause core dump if j_lookup fails.
  [fixed in 5.2.14: jobs.c(waitfor): return if j_lookup fails]

x pdksh 5.2.13, - (reported by Martin Bond): if shell is started several
  times in quick succession, echo $RANDOM produces the same results.
  [fixed in 5.2.14: main.c(main): seed RANDOM using time, pid, ppid]

x pdksh 5.2.13, - (reported by Martin Bond): repeating "echo `echo $RANDOM`"
  will always produce the same number.
  [fixed in 5.2.14: call var.c(change_random) from jobs.c(exchild)]

x pdksh 5.2.13, hpux 10.x (reported by Mike Kelly): pwd will dump core if
  current directory is not readable.
  [fixed in 5.2.14: config test & code to work around hpux C library bug]

x pdksh 5.2.13, linux (reported by Mike Jetzer): getwd warning from linker
  [fixed in 5.2.14: configure.in/misc.c: check for getcwd, use it over getwd]

x pdksh 5.2.13, (reported by Dmitri Kulginov): "trap exit 1" does not set a
  trap for HUP (exit is mistaken as a signal name, not a command).
  [fixed in 5.2.14: c_sh.c(c_trap): use case sensitive lookup for first arg]

x pdksh 5.2.13, (reported by Mark Funkenhauser): eval "$(false)" does not
  result in $? being set to 1 (is 0).
  [fixed in 5.2.14: c_sh.c(c_eval): set exstat to subst_exstat before shell()]

x pdksh 5.2.13, (reported with fix by Kevin Schoedel): word boardaries in
  file completion are only spaces - at&t ksh uses ()<>&| and spaces.
  [fixed in 5.2.14: edit.c(IS_WORDC): changed macro to be LEX1 + quotes]

x pdksh 5.2.13.5, (reported with fix by Martin Lucina <[e-mail]>):
  exit status parsing in exit command incorrect (sets status to random
  value if no argument given).
  [fixed in 5.2.14: c_sh.c(c_exitreturn): only set exstat if arg given]

x pdksh 5.2.13.5, (reported with fix by Martin Lucina <[e-mail]>):
  KSH_CHECK_H_TYPE in aclocal.m4 has too many [] around the patterns.
  [fixed in 5.2.14: aclocal.m4(KSH_CHECK_H_TYPE): remove two pairs on []]

x pdksh 5.2.13.5, (reported by Charles M. Hannum <[e-mail]>): An exit
  trap set in a subshell is not executed (unless explicit exit used).
  [fixed in 5.2.14: exec.c(execute): changed exit(rv) to unwind(LEXIT).]

x pdksh 5.2.12, - : MAILCHECK isn't preserved from the envirnment on startup.
  [fixed in 5.2.14: changed main's initcoms[].]

x pdksh 5.2.13, (reported by Marc Olzheim): time at the end of a pipeline
  doesn't print anything.
  [fixed in 5.2.14: exec.c(execute): clear XEXEC when calling timex().]

x pdksh 5.2.13, (reported by David J. McMahon): here documents in subshells
  don't work if the parent exits before the subshell.
  [fixed in 5.2.14: heredocs now saved in memory, written to temp when needed.]

x pdksh 5.2.13, (reported by Seiichi Namba): emacs: keys bound in .profile/$ENV
  are overridden by stty settings (eg, binding ^U in .profile has no effect).
  [fixed in 5.2.14: emacs.c: added x_bound array to track what user has set.]

x pdksh 5.2.13: vi: failed redo (.) commands caused line to be returned to the
  shell (eg, "echo hi/there^[Bdf/.").
  [fixed in 5.2.14: vi.c(vi_hook): removed !=0 from VREDO switch expression]

x pdksh 5.2.13, (reported by Arthor Pool): man page: (a) the time reserved
  word is not described; (b) description of command line wrapping is in vi
  section only (not in emacs); (c) limit on array indicies not mentioned;
  (d) ignoreeof ignored if eof read 13 times.
  [fixed in 5.2.14: man page updated]

x pdksh 5.2.13, (reported by Arthor Pool): set -u causes loss of stdout
  when command substitution with undefined parameter reference is run
  in an interactive shell.
  [fixed in 5.2.14: jobs.c(fill_command): don't eval TCOM arguments]

  XXX fd 1 lost (general fd pool handler?, error handler for comsub?)

  AP messages:
      time not descr
      vi/emacs <+>
      trap
      typeset -f ... "$(jasdsjh)" ...
      array
      os2 interrupts + pattern 
      ignoreeof
      set -A --
      set -u -> comsub errors & fill_command eval

x pdksh 5.2.13, (reported by Dave Hillman): test -nt
  and test -ot do not succeed if file2 (file2) does not exist.
  [fixed in 5.2.14: c_test.c(test_eval): return true if appropriate stat fails]