Patch Upstream Report: OpenSUSE

The following OpenSUSE patches have now been upstreamed:

• cpp.patch --> e79c2929: iffe depends on cc -E not inserting newlines between tokens
• ksh93-aso.dif --> e2c2753e: libast: Fix detection of GCC 4.1+ 64/32-bit atomic operations model
• ksh93-builtin.dif --> 5a2e7dae: pty: Fix signal handling (re: 1ca9286a)
• ksh93-compat.dif --> be339424: Some more build fixes from OpenSUSE
• ksh93-dttree-crash.dif --> cc492752: libast: Update cdt(3): Allow empty strings in (dt)trees
• ksh93-edpredict.dif --> db5621db: Fix editor prediction code garbling input
• ksh93-env.dif --> 3ad43070: name.c: fix possible crash in attstore()
• ksh93-filedefined.dif --> 51670c76: Fix build failures caused by update in glibc
• ksh93-foreground-prgrp.dif --> 4645e6ad: job_reset(): fix for resetting foreground process group
• ksh93-gcc.dif --> 627df2b1: libast: optget(3): fix some uninitialised variables
• ksh93-heredoc.dif --> 674a0c35: Another lexical fix for here-documents (re: 6e515f1d)
• ksh93-heredoclex.dif --> 6e515f1d: Fix command substitutions run on the same line as a here-doc (#91)
• ksh93-ia64.dif --> 0684806d: feature tests: signbit detection fix for ia64
• ksh93-joblock.dif --> 52067c3d: Backport atomic job locking from ksh 93v- beta
• ksh93-jobs.dif --> 88e8fa67: Avoid crash due to broken optimisation in job locking [OpenSUSE]
• ksh93-limit-name-len.dif --> 9f980ba6: sh_setmatch(): fix node size calculation
• ksh93-limits.dif --> 7a6980a3: libast/comp/conf.sh: apply limits detection fixes for Linux
• ksh93-longenv.dif --> 05ac1dbb: Fix crash upon running many subshells (#113)
• ksh93-no-sysctl.dif --> 8a58166b: Add OpenSUSE fixes for the uname builtin on Linux
• ksh93-nvtree-free.dif --> 10b6ba80: Fix memory corruption when a compound variable is unset (#49)
• ksh93-optimizeleak.dif --> 9b5ff0f8: Fix leak in optimize processing
• ksh93-pathtemp.dif --> db71b3ad: libast: robustness fixes for temporary files and streams
• ksh93-redirectleak.dif --> db72f41f: Fix subshell file descriptor leak
• ksh93-reg.dif --> 69e18de5: edit.c: Fix history generation if there are zero arguments
• ksh93-sfio.dif --> 1bd06207: libast: sfio(3) fixes from ksh 93v- beta
• ksh93-signals.dif --> fa3a3d79: sh_ntfork(): handle SIGTSTP along with SIGTTIN and SIGTTOU
• ksh93-spawnlock.dif --> 6d135269: Fix locking error in spawn implementation
• ksh93-stkalias.dif --> 4604df9a: Stack robustness fixes from OpenSUSE
• ksh93-stkfreeze.dif --> 4604df9a: Stack robustness fixes from OpenSUSE
• ksh93-stkset-abort.dif --> 4604df9a: Stack robustness fixes from OpenSUSE
• ksh93-test.dif --> c5bd6874: test -p: fix bug due to wrong parentheses
• ksh93-typedef.dif --> b12be093: typeset: print_value(): save/restore last_table
• ksh93-uname.dif --> 8a58166b: Add OpenSUSE fixes for the uname builtin on Linux
• ksh93-uninitialized.dif --> 288b6c65: Fix various possible uses of uninitialised variables
• ksh93-unset-f.dif --> c52cb939: sh_funscope(): Fix possible dereference of null pointer
• ksh93-vi.dif --> 129614b9: edit/vi.c: sanity checks from OpenSUSE
• ksh93-vm.dif --> 3002c411: vmalloc: better detect default region size on Linux

The following have NOT been upstreamed, because:

• astksh_builtin_poll20120806_001.diff
  I'm unsure if users require this, and there are some regression test
  failures when building this on macOS [TODO: test other OSs].
  TODO: make an updated version of this patch so OpenSUSE can keep it.

• ksh93-gmt2utc.dif
  This patch is not quite correct. See: 8d8a8257
  To get ksh to prefer UTC over GMT in 'printf %T' output, only the
  change in format[] was needed. The corresponding change in zone[]
  made it prefer UTC for London time, even in summer time, which is
  wrong -- e.g.:

      $ LANG=C TZ=Europe/London arch/*/bin/ksh -c 'date; printf %T\\n now'    
      Tue Jun 30 01:39:09 BST 2020    
      Tue Jun 30 00:39:09 UTC 2020

• ksh-qemu.patch
  Your .spec applies this patch conditionally so it should not be upstreamed.
  Note that the race condition with vfork() or posix_spawn() is fixed by using
  fork() if job control is enabled.
  see: https://github.com/ksh93/ksh/commit/[f207cd57](https://github.com/ksh93/ksh/commit/f207cd57)
  discussion: https://github.com/ksh93/ksh/issues/79

• ksh93-alias-k.dif
  In fact this was not upstramed in 93v- as claimed. It's also wrong: alias
  and hash should not be declaration builtins as they do not modify variables.
  Please show me the contents of the non-public bug [bnc#824187] so I can
  verify if a fix is still needed.

• ksh93-backtick.dif
  Not applicable; command substitutions have been fixed by smaller Red Hat
  patches and it is working well now. Please verify against your reproducers
  from bnc#887320] [bnc#926172] [bnc#934437] [bnc#953533] [bnc#955221]
  (would be nice if I could see those bugs, too)

• ksh93-cdpwd.dif
  Already fixed in bb4745e8 using a patch from Solaris.

• ksh93-disable-vfork.dif - description: "disable vfork, use fork instead"
  Your .spec applies this patch conditionally so it should not be upstreamed.
  Note that the race condition with vfork() or posix_spawn() is fixed by using
  fork() if job control is enabled.
  see: https://github.com/ksh93/ksh/commit/[f207cd57](https://github.com/ksh93/ksh/commit/f207cd57)
  discussion: https://github.com/ksh93/ksh/issues/79

• ksh93-fdstatus.dif
  Causes various regression test failures and hangs. Do not apply. The
  associated problems have likely been fixed already, e.g. ‘read’ can read
  from fifos just fine even without a final newline due to fixes in libast
  sfpkrd(). Please show me the contents of the relevant non-public bugs
  [bnc#808449, bnc#814135] so I can verify whether there is anything there
  left to fix.

• ksh93-fs3d.dif
  Not applicable as both 3dfs and libcoshell have been removed from ksh 93u+m.

• ksh93-int16double.dif
  Very similar patch applied in 5135cf65 with an additional fix for nvdisc.c.

• ksh-locale.patch
  No-op change.

• ksh93-jpold.dif
  This is already part of a Red Hat patch that was applied in ce68e1be.

• ksh93-malloc-hook.dif - description: "fix build with std malloc"
  Not convinced this is needed; it builds just fine with standard malloc now.

• ksh93-path-skip.dif
  Different fix from ksh2020 already applied:
  https://github.com/ksh93/ksh/commit/[eee47df4](https://github.com/ksh93/ksh/commit/eee47df4)
  https://github.com/att/ast/commit/[3bc58164](https://github.com/ksh93/ksh/commit/3bc58164)

• ksh93-profile.dif
  I can't determine if this patch is necessary or correct, and the code
  looks a bit strange, strdup()'ing a string which it then doesn't use.
  Holding off on this pending further investigation. (It looks like it might
  allow a path in $ENV to override /etc/ksh.kshrc if the shell is compiled
  with SHOPT_SYSRC on, which may or may not be good.)

• ksh93-s390.dif
  Ifdef'ed as Linux-specific. Probably now unnecessary on Linux as well,
  after the fixes in libast's sfpkrd() function (although I do not have
  access to a Linux s390 box to test it). Test:
  mkfifo myfifo
printf no-final-linefeed >myfifo &
read myline
If that doesn't hang, this patch is not needed.

• ksh93-shift_ijs.dif
  The regression tests that come with this patch fail. In terms of
  multibyte locales, I'm okay with ksh 93u+m supporting UTF-8 only.
  If this needs to work, someone other than me will need to fix it.

• ksh93-subshellpwd.dif
  Not applicable; fixed differently.
  See https://github.com/ksh93/ksh/issues/153

• ksh93-suid_exec.dif
  Conflicts with build patches for CDE/dtksh.
  Perhaps best to keep this a local patch only. It will need updating.

• ksh93.dif
  Misc undocumented changes with unclear reasons, most look OS-specific.

• workaround-stupid-build-system.diff:
  Not applicable, should not be needed now.