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。

  nothing.
  [see Mail.3:76]
  [fixed in 5.1 - vi.c(classify[]) table got changed by accident.]

x pdksh 5.0.10, - (reported by Danial Quinlan): forward-word and
  delete-word-forward functions in emacs don't go to the right place.
  [see Mail.3:79]
  [Fixed in 5.1 - changed order of loops in emacs.c(x_fword())]

x pdksh 5.0.10, - (reported by David Tamkin): eof in multiline command
  causes shell to exit, even if ignoreeof is set.
  [see Mail.3:76]
  [Fixed in 5.1 - reset eof after longjmp() in main.c(shell)]

x pdksh 5.0.9, Ultrix 4.2 (reported by Matthew Nethook): type-ahead while
  shell is waiting for a command to finish is temporarily lost until a
  program that reads from stdin or goes a stty/gtty is run.
  [see Mail.3:61,62]
  [Fixed in 5.1 - changed aclocal.m4 to not define HAVE_TERMIOS_H on ultrix]

x pdksh 5.0.10, - (reported by David Tamkin): if INT is trapped, ^C in
  vi/emacs won't flush buffer/re-issue new prompt.
  [see Mail.3:5,76]
  [Fixed in 5.1 - use unwind() in vi/emacs to get back to shell()]

x pdksh 5.0.10, - (reported by Dale DePriest): in emacs mode, file completions
  resulting in long names (>256) cause core dumps
  [see Mail.3:72]
  [Fixed in 5.1 - use dynamically sized buffers in emacs code]

x pdksh 5.0.10, - (reported by Dale DePriest): in emacs mode, command
  completions (^[=) resulting in multiple hits caused internal memory error.
  [see Mail.4:8]
  [Fixed in 5.1 - don't call list_stash() twice in compl_command]

x pdksh 5.0.10, - (reported by Dave Hatton): autoloading functions fail
  on the first attempt, then work.
  [see Mail.4:10]
  [Fixed in 5.1 - in findcom(), check for include() returning non-0 (was 0)]

x pdksh 5.0.10, - (reported by Art Pina via Dale DePriest): when SECONDS
  parameter is assigned, it always acts as if 0 were assigned.
  [see Mail.4:12]
  [Fixed in 5.1 - set internal seconds variable to time - assigned value]

x pdksh 5.1.0 - (reported by Larry Bouzane): for/select loops don't allow
  {..} to be used instead of do...done.
  [see Mail.4:16]
  [Fixed in 5.1.1 - changed syn.c(dogroup) to allow {/} instead of do/done]

x pdksh 5.1.0 - (reported by Andrew Moore and Larry Bouzane): a command ending
  in ; or & that is not followed by a newline causes a syntax error.
  [see Mail.4:126,128]
  [Fixed in 5.1.1 - don't call syntaxerr() in get_command() if EOF is read]

x pdksh 5.1.0, - (reported by Simon J. Gerraty): ksh died reading history
  file (complex history, in hist_skip_backup()).
  [see Mail.4:24]
  [Fixed in 5.1.1 - hist_skip_back(): don't start past the end of the buffer]

x pdksh 5.1.0 BSDI 1.1 (reported by Karl Denninger): after receipt of SIGHUP,
  shell waits for foreground process to complete.
  [see Mail.4:50,57]
  [Fixed in 5.1.1 - added fatal_trap flag, check in jobs.c(j_waitj)]

x pdksh 5.1.0 - (reported by Bob Manson): a leading non-white-space IFS
  character does cause a field to be delimited.
  [see Mail.4:68]
  [Fixed in 5.1.2 - changed expand() to do the right thing.]

x pdksh 5.1.2, -: ^c during $ENV or .profile kills shell; should just go
  to prompt.
  [see Mail.5:14]
  [fixed in 5.2.4 - added intr_ok flag to main.c(include)]

x pdksh 5.1.2, - (reported by Dan Quinlan): when shell prints out
  execution trees (typeset -f), if botches elif statements.
  [see Mail.5:17]
  [fixed in 5.1.3 - changed tree.c(ptree) to deal with elif.]

x pdksh 5.1.2, - (reported by Dale DePriest): fc -l -- -40 fails if there
  are fewer than 40 commands.
  [see Mail.5:19]
  [fixed in 5.1.3 - changed history.c(histget) to allow out of range numbers]

x pdksh 5.1.2, - (reported by Art Mills): file completion in command mode
  doesn't work on a single character.
  [see Mail.5:13]
  [fixed in 5.1.3 - in vi.c(vi_cmd) call complete_word() with 1 not 0]

x pdksh 5.1.2, - (reported by Dan Quinlan): an error in a let statement
  causes shell to exit function/script.  at&t ksh just prints error and
  returns from let.
  [see Mail.5:17]
  [fixed in 5.2.3 - added error_ok arg to evaluate() and v_evaluate()]

x pdksh 5.1.2, - (reported by Art Mills): if markdirs option is set, file
  completion in vi adds two slashes to directories.
  [see Mail.5:35]
  [fixed in 5.1.3 - vi.c(complete_word), don't add / if file ends in one]

x pdksh 5.1.2, - (reported by Dale DePriest): history read from history file
  have negitive numbers and can't be accessed (fc thinks neg numbers are
  relative).
  [see Mail.5:39]
  [fixed in 5.1.3 - EASY_HISTORY/hist_init: increment line for each line]

x pdksh 5.1.2, - (reported by David Tamkin): FPATH isn't searched if PATH
  search can't find command (undocumented at&t ksh feature).
  [see Mail.5:45]
  [fixed in 5.1.3 - exec.c(findcom) search FPATH if PATH search fails]

x pdksh 5.1.2, - (reported by Dan Quinlan): output typeset -f isn't
  very pretty (no indenting done).
  [see Mail.5:17]
  [fixed in 5.1.3 - indenting added to ptree routines]

x pdksh 5.0.9, ISC 3.2 (reported by [e-mail]): Running the
  following script with pdksh crashes the machine:
    cat > /tmp/foobar
  The same command in an interactive pdksh does not cause a crash.
  [see Mail.3:21,Mail.5:62]
  [Fixed by Interactive - it is caused by an OS bug for which there is a patch]

x pdksh 5.1.3, linux - (reported by Dan Quinlan): doesn't compile under new
  linux due to declaration conflict between basename() in unistd.h and
  pdksh'd basename.
  [see Mail.5:90]
  [fixed in 5.2.0 - changed basename() to arrayname()]

x pdksh 5.1.3, - (reported by William Hudacek): very long prompts cause
  vi command line editor grief.
  [see Mail.6:2]
  [fixed in 5.2.0 - initial part of prompt is stripped if its too long]

x pdksh 5.1.3, - (reported by Roberto Zacheo): when set -u, variable trimming
  with always causes an error.
  [see Mail.6:21]
  [fixed in 5.2.0 - fixed varsub() to test if variable is null]

x pdksh 5.1.3, - (reported by David Tamkin): when a fucntion is autoloaded,
  ksh complains the definition file didn't define the function, even if it did.
  [see Mail.6:52]
  [fixed in 5.2.0 - exec.c(comexec): when checking if defined, use cp,

x pdksh 5.1.3, ICS unix 3.2 (reported by Robert Clark): auto configuration
  test for memmove doesn't work
  [see Mail.6:65]
  [fixed in 5.2.0 - special cases added for memmove, bcopy, memset]

x pdksh 5.1.3, Unixware (Intel-SVR4.2) (reported by Thanh Ma): auto
  configuration test for memset doesn't work; same for rlimit type.
  [see Mail.6:67]
  [fixed in 5.2.0 - special cases added for memmove, bcopy, memset; rlim_t
   configuration stuff re-arranged]

x pdksh 5.1.3, - (reported by Mike Jetzer + fix): . in vi doesn't work
  after history motion or after one command is completed and another is being
  edited.
  [see Mail.6:85]
  [fixed in 5.2.0 - fix up classify table, special case for empty initial
   insert]

x pdksh 5.1.3, - Janjaap van Velthooven: ^v (version) missing in vi mode.
  [see Mail.6:98]
  [fixed in 5.2.0 - added]

x pdksh 5.1.3, - : y% on or before right bracket/paren/brace doesn't yank the
  brackets - just what is in the brackets...
  [fixed in 5.2.0 - changes to vi.c(domove,vi_cmd)]

x pdksh 5.1.3, - (reported by Rob Mayoff): [[ ]] command doesn't do lazy
  evaluation.
  [see Mail.7:2]
  [fixed in 5.2.1 - test routines re-arranged to deal with this]

x pdksh 5.1.3, - (reported by Will Renkel): "r | more" doesn't work (nothing
  is sent to more).
  [see Mail.7:13]
  [fixed in 5.2.0 - history commands now done in c_fc, not pushed onto input
   stack]

x pdksh 5.1.3, - (reported by Rod Byrne, John Rochester): if a program leaves
  the non-blocking (O_NONBLOCK) flag set after it exists, the shell
  exits (multiple eofs).
  [see Mail.7:15,16,51]
  [fixed in 5.2.0: O_NONBLOCK is reset if read fails with EAGAIN,EWOULDBLOCK]

x pdksh 5.1.3, - (reported by Dale DePriest + fix): emacs: can't delete chars
  from pattern in incremental search mode.
  [see Mail.7:17]
  [fixed in 5.2.0 - handle it]

x pdksh 5.1.3, Linux 1.2.2 (reported by Fritz Heinrichmeyer + fix): siglist.sh
  doesn't work due to bug in bash 1.4.3 (trap is called incorrectly in
  subshell causing temp file to be removed prematurely).
  [see Mail.7:21]
  [fixed in 5.2.0 - clear all traps in subshell so file isn't removed]

x pdksh 5.1.3, - (reported by Dale DePriest + fix): emacs: can't prefix
  commands with more than single digit; many commands don't use nnumber
  prefix.
  [see Mail.7:26,40]
  [fixed in 5.2.0 - x_set_arg reads sequence of numbers, other commands
  changed to use x_arg]

x pdksh 5.1.3, - (reported by Dale DePriest): fc command line parsing
  (and its interaction with history alias) doesn't act like at&t ksh:
  history -40 gives bad option 4 error.
  [see Mail.7:41,49]
  [fixed in 5.2.1 - kludge parsing of -40 (numbers are option letters)]

x pdksh 5.1.3, - (reported by Dale DePriest): if PS1 contains paramaters that
  get expanded, and if those parameters contain any ! characters, the !'s get
  changed to history numbers.
  [see Mail.7:44]
  [fixed in 5.2.0 - substitution done after ! and !! substitution]

x pdksh 5.1.3, - (reported by Steve Wallis): set -a (set -o allexport) has
  no effect.
  [see Mail.7:47]
  [fixed in 5.2.0 - changes to c_read, c_getopts, and comexec]

x pdksh 5.1.3, - (reported by Alexander S. Jones): (sleep 10000&) waits for
  the sleep to complete.
  [see Mail.7:54]
  [fixed in 5.2.0 - execute() case TASYNC clears EXEC flag in call to execute]

x pdksh 5.1.3, - (reported by Will Renkel): positional parameters can't be
  accessed within temporary variable assignments (eg, "FOO=$1 blah" doesn't
  set FOO to $1.
  [see Mail.7:57]
  [fixed in 5.2.0 - var.c(newblock) - copy argc/argv from previous environment]

x pdksh 5.1.3, SCO unix ? (reported by Sean Hogan): job control stuff doesn't
  work as sco doesn't do job control operations on /dev/tty.
  [see Mail.7:30,43,69,70,74]
  [fixed in 5.2.0 - don't try opening /dev/tty if on SCO]

x pdksh 5.1.3, - (reported with fix by Mike Jetzer): vi globing tacks
  * at the end of files even if there are globing chars in last component
  of filename (at&t ksh does not).
  [see Mail.7:71]
  [fixed in 5.2.0 - don't append * if there are unescaped globing chars]

x pdksh 5.1.3, - (reported with fix by Gabor Zahemszky): typoes in acconfig.h,
  sh.h uses SVR3_PGRP insteda of SYSV_PGRP.
  [see Mail.7:87]
  [fixed in 5.2.0]

x pdksh 5.1.3, - (reported by Gabor Zahemszky): emacs doesn't have ^[^].
  [see Mail.7:87]
  [fixed in 5.2.0 - added search-char-backward]

x pdksh 5.2.0, - (reported by David Tamkin): pwd -P doesn't strip .. and .
  properly.
  [see Mail.7:98]
  [fixed in 5.2.0 - include ksh_stat.h in c_ksh.c]

x pdksh 5.2.0, - (reported by Dale DePriest): unistd.h config test
  doesn't include sys/types before dirent.h.
  [see Mail.8:2]
  [fixed in 5.2.0]

x pdksh 5.2.0, - (reported by Robert Gallant): emacs file/command completion
  code can clobber memory.
  [see Mail.8:11]
  [fixed in 5.2.1 - wrong variable being checked in buffer growing in
   emacs.c(compl_file,compl_command)]

x pdksh 5.2.0, - (reported by David Tamkin): when CDPATH set and cd'ing to a
  directory that doesn't exist, the error message contains the last element
  of the CDPATH.
  [see Mail.8:8]
  [fixed in 5.2.0 - fixed error message]

x pdksh 5.2.0, - (reported by David Tamkin): if PS1 has an error in it
  (eg, parameter expansion error), the shell loops forever printing
  the error.
  [see Mail.8:32]
  [fixed in 5.2.3 - create error handling environment while expanding PS1]

x pdksh 5.2.0, Coherent machines (reported by Gabor Zahemszky): insert after
  movement in emacs mode replaces all chars with first char on line.
  System's bcopy doesn't handle overlapping src/dst.
  [see Mail.8:38,43]
  [fixed in 5.2.1 - check for broken memmove/bcopy in aclocal.m4]

x pdksh 5.2.0, - (reported by Gabor Zahemszky): ^[= in vi prints empty
  strings for directory matches if markdirs is set.
  [see Mail.8:48]
  [fixed in 5.2.1 - skip trailing /'s before looking for last /]

x pdksh 5.2.0, - (reported by Gabor Zahemszky): <ESC>^H bound to del-back-char
  not del-back-word
  [see Mail.8:50-52]
  [fixed in 5.2.1 - fixed x_emacs_keys]

x pdksh 5.2.1, - (reported by David Tamkin): compile fails due to lack
  of c_test.h
  [see Mail.8:58]
  [fixed in 5.2.2 - fixed put c_test.h in distribution]

x pdksh 5.2.2, - (reported by Simon J. Gerraty): hist_source not being
  initialized in complex history.
  [see Mail.8:64]
  [fixed in 5.2.3 - set it in second hist_init()]

x pdksh 5.2.2, - (reported by Gabor Zahemszky): set -A does not reset
  the array contents.
  [see Mail.8:65]
  [fixed in 5.2.3 - changed var.c(unset) to unset whole array if appropriate]

x pdksh 5.2.2, - (reported by Gabor Zahemszky): getopts stops after an error;
  at&t ksh carries on with next option.
  [see Mail.8:65]
  [fixed in 5.2.3 - remove GI_DONE flag from ksh_getopt()]

x pdksh 5.2.2, - (reported by Gabor Zahemszky): getopts prints shell name
  twice in error messages.
  [see Mail.8:65]
  [fixed in 5.2.3 - added GI_NONAME flag]

x pdksh 5.2.2, - (reported by Gabor Zahemszky): pdksh's test doesn't know about
  /dev/fd/n.
  [see Mail.8:65]
  [fixed in 5.2.3 - added test_stat() and test_eaccess()]

x pdksh 5.2.2, - (reported by Thomas Gellekum): config test for memmove/bcopy
  missing semi-colon
  [see Mail.8:67]
  [fixed in 5.2.3]

x pdksh 5.2.2, - (reported by Donald Craig): fc string doesn't find string
  if it is the most recent command.
  [see Mail.8:76]
  [fixed in 5.2.3 - fixed off by one error in history.c(hist_get)]

x pdksh 5.2.2, - (reported by Gabor Zahemszky): pdksh doesn't do the
  "You have running jobs" when user attempts to log out.
  [see Mail.8:74]
  [fixed in 5.2.3 - added set -o nohup option with supporting code]

x pdksh 5.2.2, - (reported by Gabor Zahemszky): configure test for
  broken memmove/bcopy doesn't work.
  [see Mail.8:93]
  [fixed in 5.2.3 - fixed test to copy overlapping buffers]

x pdksh 5.1.3, - (reported by <[e-mail]>): doesn't compile on
  solaris 5.x with COMPLEX_HISTORY defined.