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。

  [see Mail.8:98]
  [fixed in 5.2.3 - undef COMPLEX_HISTORY if flock not available]

x pdksh 5.2.2, - (reported by Gabor Zahemszky): tilde expansion not preformed
  in word part of ${foo[-+=?} substitution.
  [see Mail.9:7]
  [fixed in 5.2.3 - allow ~foo to end in a close brace]

x pdksh 5.2.2, - (reported by Gabor Zahemszky): "fc 30" edits from 30 to
  most recent history (should be just 30).
  [see Mail.9:7]
  [fixed in 5.2.3 - if !-l and no last given, use first]

x pdksh 5.2.2, - (reported by Gabor Zahemszky): [many problems with man page]
  [see Mail.9:12]
  [fixed in 5.2.3 - fixed problems]

x pdksh 5.2.2, - (reported by Gabor Zahemszky): #else followed by non-comment
  in sigact.c.
  [see Mail.9:13]
  [fixed in 5.2.3 - turn it into a comment]

x pdksh 5.2.2, - (reported with fix by Gabor Zahemszky): two argument form of
  cd doesn't work.
  [see Mail.9:14]
  [fixed in 5.2.3 - in c_cd(), use current_wd not path]

x pdksh 5.2.2, - (reported with fix by Gabor Zahemszky): command -V doesn't
  report reserved words.
  [see Mail.9:30]
  [fixed in 5.2.3 - in c_whence(), look for reserved words if vflag set]

x pdksh 5.2.3, - (reported by Dale DePriest): at&t's tbl wants space
  between font specification and end of table descrption (ie, fB . not
  fB.).
  [see Mail.9:41]
  [fixed in 5.2.4 - put spaces in]

x pdksh 5.2.3, - (reported by David Tamkin & Claus L{gel Rasmussen): PS1
  isn't imported from environment anymore.
  [see Mail.9:43,76]
  [fixed in 5.2.4 - main: don't set PS1 if it is already set]

x pdksh 5.2.3, - (reported by Gary Rafe): If PS1 contains newlines, vi
  editing mode dones't redraw lines properly.
  [see Mail.9:63]
  [fixed in 5.2.4 - added prompt_skip stuff to vi/emacs]

x pdksh 5.2.3, - (reported & fixed by Mike Jetzer): cd: error message if
  directory didn't exist was wrong.
  [see Mail.9:66]
  [fixed in 5.2.4 - print correct string in error message]

x pdksh 5.2.3, - (reported & fixed by Mike Jetzer): vi: <ESC>* shouldn't append
  a * if word contains a $.
  [see Mail.9:66]
  [fixed in 5.2.4 - vi.c(glob_word): check for $ in word, check for null
   expansion]

x pdksh 5.2.3, - (reported & fixed by Mike Jetzer): vi: <ESC>= doesn't
  list expansions in column form.
  [see Mail.9:66]
  [fixed in 5.2.4 - use pr_menu to print things nicely]

x pdksh 5.2.3, - (reported Larry Bouzane): should be a way of installing
  binary/man page as pdksh instead of ksh.
  [see Mail.9:100]
  [fixed in 5.2.4 - use the --enable-shell=pdksh option to configure]

x pdksh 5.2.3, - (reported by Gabor Zahemszky): [many problems with man
  page]
  [see Mail.10:20]
  [fixed in 5.2.4 - fixed problems]

x pdksh 5.2.3, - (reported by Gabor Zahemszky): exec 1<&9 reports
  error with ">&9" in it.
  [see Mail.10:20]
  [fixed in 5.2.4 - changed iosetup()]

x pdksh 5.2.3, - (reported by Gabor Zahemszky): man page doesn't document
  /dev/fd/N
  [see Mail.10:20]
  [fixed in 5.2.4 - updated manual]

x pdksh 5.2.3, - (reported by Ted Coady): [[ foo/bar = foo* ]]
  fails; should succeed.
  [see Mail.10:32]
  [fixed in 5.2.4 - fixed problem in exec.c(dbteste_getopnd)]

x pdksh 5.2.3, - (reported by Ruei-wun Tu): make on NeXT/NeXTSTEP 3.3
  doesn't understand .PRECIOUS target and so does nothing.
  [see Mail.10:43]
  [fixed in 5.2.4 - moved .PRECIOUS after all in Makefile.in]

x pdksh 5.2.3, - (reported & fixed by Paul Borman): shell doesn't kill
  foreground process when SIGHUP received; Also, CONT sent before HUP.
  [see Mail.10:44]
  [fixed in 5.2.4 - j_exit now sends HUP to foreground process]

x pdksh 5.2.3, AIX 3.2.5 (reported by Ian Portsmouth): C compiler compains
  about sigtraps[] being re-declared in trap.c.
  [see Mail.10:73]
  [fixed in 5.2.4 - use cpp define to avoid bogus re-declaration error]

x pdksh 5.2.3, - (reported by Michael Haardt): ENV should not be
  included if shell is compiled as sh and posix option not set.
  [see Mail.10:83]
  [fixed in 5.2.4 - only include ENV if POSIX, if compiled as sh]

x pdksh 5.2.3, - (reported & fixed by DaviD W. Sanderson): case statements
  don't allow {/} in place of IN/ESAC.
  [see Mail.10:77,78]
  [fixed in 5.2.4 - allow {/} in case statements]

x pdksh 5.2.3, - (reported by Larry Daffner): $? is incorrectly zero'd
  at start of traps.
  [see Mail.11:9]
  [fixed in 5.2.4 - don't clear exstat in main.c(shell)]

x pdksh 5.2.3, - (reported by Frank "Crash" Edwards): configure on linux XXX
  doesn't detect the presence of lstat().
  [see Mail.11:36]
  [fixed in 5.2.4 - change configure to include <sys/stat.h> in lstat() test]

x pdksh 5.2.3, - (reported by Gabor Zahemszky): typeset -f dumps core
  in the after using autoload functions.
  [see Mail.11:74?]
  [fixed in 5.2.4 - c_typeset no longer traverses the array link for functions]

x pdksh 5.2.3, - (reported by Gabor Zahemszky): typeset -f does not report
  undefined autoload functions
  [see Mail.11:74?]
  [fixed in 5.2.4 - c_typeset: don't ignore unset functions]

x pdksh 5.2.3, - (reported by Dale DePriest): alias -t -r does not
  reset aliases.
  [see Mail.11:99]
  [fixed in 5.2.4 - c_alias: call ksh_getopt_reset() before calling c_unalias]

x pdksh 5.2.3, - (reported & fixed by Jason Tyler): 'echo abc^Jfc -e - a=b e'
  echos b, not bbc.
  [see Mail.11:100?]
  [fixed in 5.2.4 - hist_replace: use s, not last]

x pdksh 5.2.3, - (reported by Jason Tyler): 'fc -e -' when there is
  no history causes infinite loop.
  [see Mail.11:100?]
  [fixed in 5.2.4 - histbackup: allow histptr to go below history]

x pdksh 5.2.4, - (reported by David Tamkin): jmp_buf is used instead of
  sigjmp_buf.
  [see Mail.XXX:XXX]
  [fixed in 5.2.5 - added ksh_jmp_buf and defined appropriately]

x pdksh 5.2.4, - (reported by Stephen Coffin): /<RETURN> in vi mode does not
  repeat last search.
  [see Mail.XXX:XXX]
  [fixed in 5.2.5 - vi.c(vi_hook) - make it repeat last search]

x pdksh 5.2.4, - (reported by Gabor Zahemszky): functions containing select
  commands aren't printed correctly by typeset.
  [see Mail.XXX:XXX]
  [fixed in 5.2.5 - tree.c(ptree) - add case for TSELECT]

x pdksh 5.2.4, - (reported & fixed by Stefan Dalibor): COLUMNS isn't set on
  shell start up (and window size is ignored) 'cause tty_fd isn't valid when
  x_init() is called.
  [see Mail.XXX:XXX]
  [fixed in 5.2.5 - call x_init() after j_init() is called]

x pdksh 5.2.4, - (reported by Will Renkel): "echo -" just prints a blank
  line - should print the minus.
  [see Mail.XXX:XXX]
  [fixed in 5.2.5 - c_ksh.c(c_print): don't do argument parsing on lone -]

x pdksh 5.1.3, - (reported by Gabor Zahemszky): emacs doesn't have ^[*.
  [see Mail.7:87]
  [fixed in 5.2.5]

x pdksh 5.2.3, - (reported by Mike Jetzer): in vi, <ESC>= doesn't append
  a / after directories.
  [see Mail.9:66]
  [fixed in 5.2.5]

x pdksh 5.2.0, - (reported by Gabor Zahemszky): can set readonly variables
  via command assignments (eg, "readonly x=y; x=z /bin/echo hi" should
  fail and doesn't).
  [see Mail.8:50,65]
  [fixed in 5.2.5 - LOCAL_COPY flag passed from comexec() down to local()]

x pdksh 5.2.4, - (reported by Tom Karches): history: "r old=new", with
  no commands prefix given, prints "fc: too mnay arguments" - it should
  do the subst on the previous command.
  [see Mail.XXX:XXX]
  [fixed in 5.2.5]

x pdksh 5.2.3, - (reported by Vigen Pogosyan): assignments in $(( ... ))
  remember the base that was assigned in pdksh - does not in at&t ksh.
  [see Mail.10:54]
  [fixed in 5.2.5: uset setint() in expr.c(evalexpr)]

x pdksh 5.2.4, - (reported by Gabor Zahemszky): emacs: ^O steps down
  two lines (should be 1).
  [see Mail.XXX:XXX]
  [fixed in 5.2.5: convert history line to command number, then convert back]

x pdksh 5.2.3, - (reported by David Gast(? [e-mail])): fc -ln -1 -1
  reports the current command, not the previous command.
  [see Mail.10:49]
  [fixed in 5.2.5]

x pdksh 5.2.3, - (reported by Matthew Green): foo=`^Jecho bar` doesn't
  set foo to bar (foo is empty).
  [see Mail.XXX:XXX]
  [fixed in 5.2.5: syn.c: set multiline.on when source is SSTRING]

x pdksh 5.2.5, - (reported by Gabor Zahemszky): continue/break: if n
  is too big, shell prints internal error message.
  [see Mail.XXX:XXX]
  [fixed in 5.2.6: fix c_brkcont to use last loop if n is too big]

x pdksh 5.2.5, - (reported by Gabor Zahemszky): set: +o in ksh93
  prints command that sets various options.
  [see Mail.XXX:XXX]
  [fixed in 5.2.6: changed misc.c(printoptions)]

x pdksh 5.2.5, - (reported by Gabor Zahemszky): COLUMNS/LINES variables
  are not exported.
  [see Mail.XXX:XXX]
  [fixed in 5.2.6: use typeset() in edit.c(x_init) to export COLUMNS/LINES]

x pdksh 5.2.5, - (reported by Gabor Zahemszky): emacs: <ESC><ESC> puts
  space after completed directories.
  [see Mail.XXX:XXX]
  [fixed in 5.2.6: check for single/non-directory match in emacs.c(do_complete)]

x pdksh 5.2.5, - (reported by Gabor Zahemszky): vi: # removes comment
  and executes if command already commented.
  [see Mail.XXX:XXX]
  [fixed in 5.2.6: added vi.c(do_comment)]

x pdksh 5.2.7, - (reported by Adrian Marsh): test doesn't have == operator.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: added == to c_test.c operator table]

x pdksh 5.2.7, - (reported by Mike Jetzer): pdksh sets/exports COLUMNS/LINES
  which causes applications not to respond to window size changes.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: COLUMNS/LINES no longer exported automatically]

x pdksh 5.2.7, - (reported by Gabor Zahemszky): getopts sets OPTIND differently
  that at&t ksh when a bad option is given.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: OPTIND not set if option was bad - fragile fix - may go away]

x pdksh 5.2.7, - (reported with fix by Marc Olzheim): sh version shouldn't
  have mail check stuff, macro expansion in PS[0-9].
  [fixed in 5.2.8: added lots of ifdefs]

x pdksh 5.2.7, - (reported by Gabor Zahemszky): sub commands in PS1 cause
  a warning message to be printed.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: lex.c(set_prompt) - don't print the warning message]

x pdksh 5.2.7, - (reported by Tom Watson): some environment variables
  (eg, PATH) are converted to uppercase on 16-bit int machine.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: struct tbl.flag (32 bit thing) was being treated as an
   int in some places]

x pdksh 5.2.7, - (reported by Gabor Zahemszky): unset always returns 0 - should
  return 1 if variable/function is not set.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: fixed c_sh.c(c_unset)]

x pdksh 5.2.7, - (reported by Gabor Zahemszky): select should only print the
  menu the first time, if REPLAY is empty, or if a blank line is entered.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: fixed up exec.c(execute,do_selectargs)]

x pdksh 5.2.7, - (reported by Gabor Zahemszky): shell reports "cannot execute"
  error if file exists, even if . not in path.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: fixed up exec.c(comexec)]

x pdksh 5.1.3, - (reported with partial fix by [e-mail]): shell doesn't
  listen to sigwinch.
  [see Mail.7:7 and related]
  [fixed in 5.2.8: changed edit.c(x_init) to catch sigwinch]

x pdksh 5.2.7, - (reported by Gabor Zahemszky): typeset doesn't report
  variables that have attributes (like export) but no values.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: fixed up c_ksh.c(c_typeset)]

x pdksh 5.2.7, - (reported by Gabor Zahemszky): error message printed as
  a result of "set -o nounset" is different from at7t ksh.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: fixed error messges in eval.c]

x pdksh 5.2.7, - (reported by Gabor Zahemszky): vi/emacs: when listing
  command/file completions, should go back at most one space.  Also, should
  allow completions on zero length names.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: fixed edit.c(x_locate_word); now allows zero-length
   file completions (but not command)]

* pdksh 5.2.7, - (reported by Gabor Zahemszky): emacs: <esc># doesn't do
  the comment thing.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: added emacs.c(x_comment) et al.]

x pdksh 5.2.7, - (reported by Gabor Zahemszky): arithmatic expressions
  containing variables not expanded as in at&t ksh. eg, "x=1+2, let y=x"
  fails.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: added evaling/INEXPREVAL/ET_RECURSIVE code to expr.c]

x pdksh 5.2.7, - (reported by Gabor Zahemszky): unsetting the 0th element
  of an array kills the whole array.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: var.c(unset) - allow ARRAY to be preserved]

x pdksh 5.2.7, - (reported by Gabor Zahemszky): unsetting a function while
  it is being executed can result in core dump.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: table.c(texpand) - dont free if FINUSE is set]

* pdksh 5.2.7, - (reported by Gabor Zahemszky): exec 3<&p doesn't close
  shells copy of the coprocess file desc.
  [see Mail.XXX:XXX]
  [fixed in 5.2.8: coprocess stuff made to act like ksh93 co-processes]

x pdksh 5.2.8, - (reported with fix by Lars Hecking): doesn't compile as
  sh - c_ksh.c and jobs.c boom out.
  [see Mail.XXX:XXX]
  [fixed in 5.2.9: added ifdef KSH to appropriate places]

x pdksh 5.2.8, - (reported by Paolo Zeppegno): assignments containing brackets
  are treated as commands.
  [see Mail.XXX:XXX]
  [fixed in 5.2.9: fixed bug in vars.c(skip_wdvarname).]

x pdksh 5.2.5, - (reported by Adrian Marsh): configuration on Linux FT fails.
  Caused by configure script using -g flag - gcc passes -lg to ld, ld fails
  to find -lg (autoconf or Linux FT bug).
  [see Mail.XXX:XXX]
  [fixed in 5.2.9: changed autoconf's -g test to do linking as well.]

x pdksh 5.2.8, Solaris 2.5.1 (reported by Stefan Dalibor): 2 tests