sens pdksh jako /bin/sh

[Arkadiusz Miskiewicz]

Witam,

Czy jest sens używania w AC jako /bin/sh pakietu pdksh? Pytam z prostej 
przyczyny - pakiet jest kompletnie niemaintainowany, nie wychodzą na niego 
łatki, a bugów jest od czorta... u nas nikt nie ma serca poprawiać tego...

Masę zmian jest za to w wersji OpenBSDowej pdksh, do poczytania tu:
http://www.openbsd.org/cgi-bin/cvsweb/src/bin/ksh/

Swego czasu w PLD były niesamowite wręcz przygody z allocatorem pamięci 
używanym przez pdksh (zakęconym jak ruski słoik) - co zrobili ludzie z 
OpenBSD? - ano wywalili tamten kilkuset linijkowy (~ ok 700 linii) alokator 
zastępując go własnym mieszczącym się w ... niecałych 100 linijkach.

Debian ciągnie ostro z OpenBSD, poprawek jest mnóstwo.

Propozycja jest zsh jako /bin/sh, mały fajny shell i co najważniejsze - 
aktywnie rozwijany. Ewentualnie kto chętny na sportowanie BSDowych poprawek 
dla pdksh?

Oto co Debian pozabierał z OpenBSD:
+  * More patches from OpenBSD:
+    + vi.c: Buffers are not strings so use memcpy(), not strlcpy() to copy
+      them. Also add some further bounds checks in the name of paranoia.
+    + exec.c: Unbreak parameter assignment when calling bourne style
+      functions.
+    + exec.c: For the >& and <& operators, add a check for "dup from" ==
+      "dup to" and just return success if they are the same.  Fixes the
+      "ls 2>&2" problem.
+    + eval.c, exec.c, io.c, jobs.c: If "from fd" == "to fd" don't call
+      dup2() or close "from fd".
+  * More patches from OpenBSD:
+    + c_test.c: The special case code for "test -x" over NFS was
+      incorrect.  The right thing to do is to try access(2) first
+      (since that occurs on the NFS server side) and only check for the
+      absence of an execute bit when access(2) succeeds.
+    + edit.c: in word location, fix forward scanning so it correctly
+      account for any escaped char and not only spaces.  for "foo
+      (bar.a)" and "foo (bar a)", cd foo\ \(bar.<tab> will correctly
+      expand to foo\ \(bar.a\).
+    + alloc.c: Don't allow alloc() and aresize() to fail.  Their return
+      value was only checked in two place (both in conjunction with
+      str_save).  Upon malloc/realloc failure we call internal_errorf()
+      which pops throws and error and pops back to the last good state.
+  * Don't do tilde expansion after `=' sign in non-assignments.
+    Patch stolen from the posh package (closes: #187839).
  * Merge many patches from OpenBSD:
+    + Introduce FSH flag, which is set when the shell is called as `sh'.
+    + alloc.c: Kill hand-made memory allocation code, that is definitely
+      buggy. Replace with simple wrapper around malloc, at least this works,
+      and it's easier to debug anyways.
+    + tree.c: Fix three off-by-one errors.
+    + c_sh.c: don't set close-on-exec flag on file descriptors in FSH mode
+      (closes: #154540). Documented the change in ksh(1).
+    + history.c: Compare the return from mmap with MAP_FAILED, do not cast it
+      to int and compare with -1.
+    + main.c: set edit mode to emacs by default, may be overridden by the
+      environment or the user.  Also, we want tab completion in vi by 
default.
+    + misc.c: use strtol() in getn().
+    + emacs.c:
+       - bind TAB (^I) to complete-list by default
+       - complete-list first completes; if that does not work, it lists
+       - fix a memleak in do_complete()
+    + edit.c:
+       - completion now works after '=' (dd), and ':' (ssh) and ` (backtick)
+       - add '#' to the list of escaped characters during vi/emacs filename
+         completion
+
+  * exec.c: Found and fixed yet another problem with `set -e' scripts
+    (see a changelog entry for 5.2.14-3), which caused `dpkg-buildpackage -B'
+    to fail on systems where /bin/sh is ksh.
+
+  * c_sh.c: Make `set' command return 0 always, not only in the POSIX mode.
+    According to Jeff Sheinberg <jeffsh w localnet.com>, this new behaviour
+    is more compatible with SUSv2 standard and other shells (esp. ksh93)
+    (closes: #118476). Documented the change in ksh(1) man page.

-- 
Arkadiusz Miśkiewicz     CS at FoE, Wroclaw University of Technology
arekm.pld-linux.org, 1024/3DB19BBD, JID: arekm.jabber.org, PLD/Linux