Please merge mksh 39.1-3 (universe) from Debian testing (main)

[Thorsten Glaser]

Binary package hint: mksh

Binary package hint: mksh

The new version contains a *iot* of important bugfixes and has even entered
testing after 5 instead of the regular 10 days. The old Ubuntu specific patches
(one to disable debconf, as we cannot change /bin/sh away from dash, the
other to work around a gcc bug in jaunty and karmic) are still necessary.

I have tested it in a karmic chroot freshly a-g d-u’d today (otherwise not used).

I would kindly request this upload be sponsored (I am both mksh upstream author
and Debian package maintainer, and also had the last 2 uploads of mksh sponsored,
see LP#375604 and LP#385887). Debdiff attached as usual.

mksh (39.1-3ubuntu1) karmic; urgency=low

  * debian/mksh.postinst: hardcode debconf question if mksh should
    be installed as /bin/sh to false (see LP: #348872) but retain
    debconf infrastructure so that, if this was set to true in old
    installations, it will correctly be removed on upgrade
  * debian/rules: remove “-combine” option from Build.sh invocation
    to work around a gcc bug (see LP: #375604)

 -- Thorsten Glaser <email address hidden> Thu, 08 Oct 2009 08:23:59 +0000

mksh (39.1-3) unstable; urgency=low

  * Add support for using pax instead of cpio for extraction
  * Backport fix for return code bug (Closes: #548744)

 -- Thorsten Glaser <email address hidden> Tue, 29 Sep 2009 12:46:07 +0000

mksh (39.1-2) unstable; urgency=low

  * debian/rules: build mksh-small without floating point support
    also, because it’s ⓐ huge and â“‘ buggy in dietlibc
  * debian/diffs/backport-function-parens.diff: new, support the
    "function stop () {" bashism sometimes popular in sh scripts
  * debian/diffs/zz-version.diff: fix an oversight and use the
    actual 39.1-* version number instead of 38.9.yyyymmdd-* from
    when this patch was first being created in an experimental
    version; refresh diff against original files; use -U1 instead
    of -up for sh.h diff to not get fuzz at the RCS ID

 -- Thorsten Glaser <email address hidden> Sat, 26 Sep 2009 21:18:54 +0000

mksh (39.1-1) unstable; urgency=low

  * New upstream version R39 (despite #540512, I got fed up waiting for
    Guillem/Gerrit to contact me), Closes: #541617; complete ChangeLog:
    - [tg] Shut up a bogus gcc warning during configuration process
    - [tg] Spell AT&T consistently in the source code
    - [tg] Tweak mksh(1) manual page, from wbx@ and «lewellyn:#ksh»
    - [tg] dot.mkshrc: fix $@ vs. $* mix-up
    - [tg] dot.mkshrc: add DJB cdb hash function
    - [tg] Sync with oksh: fix Vi editing mode word erase handling, again
    - [tg] Skip whitespace between POSIX style shell function name and
      its definition parenthesÄ“s during detection if an alias of the same
      name already exists to be more robust (Debian Closes: #535970)
    - [tg] Build system improvements for ACK and nwcc, both on Debian sid
    - [tg] Fix spelling error in changelog discovered by Lintian
    - [tg] Aligh “set -o nounset” / “set -u” behaviour with future POSIX
      standard, as discussed with GNU bash maintainers, David Korn from
      AT&T ksh93, and The Open Group; prompted by use in Debian;
      Closes: #539538
    - [tg] add an unsupported way to make printf(1) a builtin
    - [tg] Build system and regression test code and comment improvements:
      better and more comments matching reality better; more reliability
      w.r.t. passed CPPFLAGS; more of the MKSH_SMALL changes may be
      overridden, all of them are now enumerated on the webpage; fixed
      some breakage; portability
    - [tg] MKSH_NOPWNAM and MKSH_SMALL will now both disable the ~fac/
      (homedir) expansion code wholly if defined, not just getpwnam(3)
      calls
    - [tg] shells without job control no longer define the standard
      â€œstop” and “suspend” aliases (they are pointless anyway);
      regression tests know
    - [tg] use system RCS ID macros on MirBSD if decent enough
    - [tg] shut up bogus gcc 4.5/trunk warnings caused by
      over-optimisation
    - [tg] restore ANSI C compilability broken in R38 (speed up, even)
    - [tg] use memcpy(3) ipv strlcpy(3) if possible and safe and secure
  * Integrate czech translation update, Dêkuji; Closes: #534788
  * Switch to debhelper 5, by suggestion of Patrick “aptituz” Schönfeld
  * Add (commented out) framework for building with klibc, pending
    bugfixes and enhancements I submitted to the Debian BTS; can be
    customised for with/without MKSH_SMALL and dynamic/static linkage
  * debian/copyright: remove setmode.c remark, it is never used:
    mksh-full has it provided by libbsd; mksh-diet, mksh-small, and
    mksh-klibc have the mknod builtin disabled and do not need it;
    strlcpy.c is provided by libbsd, dietlibc and klibc, so neither
    mksh-full nor mksh-diet (nor mksh-klibc) need it, only mksh-small
    in the glibc version (on platforms where dietlibc is unavailable)
  * debian/rules: Append -e to sub-make command line, to force it to
    use the correct build environment (mostly CFLAGS)
  * Apply policy compliance (-o posix if run as sh) to mksh-small
  * Disable duplication of -Wall on the compiler command line
  * debian/rules: improve comments
  * mksh-klibc: debian/rules can now build an MKSH_SMALL flavour
  * Let mksh have all builtins dash has; Closes: #532343
    â€£ New diffs/add-builtin-chdir.diff: add “chdir” builtin (= “cd”)
    â€£ New diffs/add-builtin-printf.diff: add “printf” builtin (manpage)
    â€£ debian/copyright, debian/rules: new source file printf.c (code),
      install and use it; make sure klibc has strtod(3) disabled
    â€£ New diffs/add-builtin_common.diff: common part of the diff
  * New diffs/zz-version.diff: use Debian specific ksh version number
  * debian/rules: reorder some assignments to make it better readable
    and add and improve comments
  * debian/watch: add RCS ID and comments
  * debian/mksh.menu: don’t create an entry for Diet mksh any more
  * debian/mksh.menu, debian/mksh*.xpm: add icons, long description
  * Update spanish translation; from asarch via IRC, ¡gracias!
  * Fix a spelling error and repetitiveness in the German translation
  * Use portuguese debconf translation for the missing parts of the
    brasilian-portuguese translations and mark them fuzzy
  * Translate the missing dutch parts myself (as good as I can)
  * Convert all debconf translation files to UTF-8
  * Remove VCS-CVS field override; the Lintian maintainers have
    agreed to adjust it so that anoncvs-over-ssh is not criticised
  * Switch to Debian Policy 3.8.3, no relevant changes

 -- Thorsten Glaser <email address hidden> Wed, 16 Sep 2009 11:08:08 +0000

[Thorsten Glaser (mirabilos)]

(Side note: this is also the very version I’m going to submit to backports.org for Debian Lenny users,
due to the wide range of fixes and because Debian’s stable-update policy wouldn’t allow it, even
though it unbreaks a log of scenarios.)

[Fabrice Coutadeur (fabricesp)]

According to the changelog, it seems not to be a bug fixing only release:
    * [tg] Shut up a bogus gcc warning during configuration process
    * [tg] Spell AT&T consistently in the source code
    * [tg] Tweak mksh(1) manual page, from wbx@ and «lewellyn:#ksh»
    * [tg] dot.mkshrc: fix $@ vs. $* mix-up
    * [tg] dot.mkshrc: add DJB cdb hash function
    * [tg] Sync with oksh: fix Vi editing mode word erase handling, again
    * [tg] Skip whitespace between POSIX style shell function name and its
        definition parenthesēs during detection if an alias of the same name
        already exists to be more robust (Debian Closes: #535970)
    * [tg] Build system improvements for ACK and nwcc, both on Debian sid
    * [tg] Fix spelling error in changelog discovered by Lintian
    * [tg] Aligh “set -o nounset” / “set -u” behaviour with future POSIX
        standard, as discussed with GNU bash maintainers, David Korn from AT&T
        ksh93, and The Open Group; prompted by use in Debian; Closes: #539538
    * [tg] add an unsupported way to make printf(1) a builtin
    * [tg] Build system and regression test code and comment improvements:
        better and more comments matching reality better; more reliability w.r.t.
        passed CPPFLAGS; more of the MKSH_SMALL changes may be overridden, all of
        them are now enumerated on the webpage; fixed some breakage; portability
    * [tg] MKSH_NOPWNAM and MKSH_SMALL will now both disable the ~fac/
        (homedir) expansion code wholly if defined, not just getpwnam(3) calls
    * [tg] shells without job control no longer define the standard “stop” and
        “suspend” aliases (they are pointless anyway); regression tests know
    * [tg] use system RCS ID macros on MirBSD if decent enough
    * [tg] shut up bogus gcc 4.5/trunk warnings caused by over-optimisation
    * [tg] restore ANSI C compilability broken in R38 (speed up, even)
    * [tg] use memcpy(3) ipv strlcpy(3) if possible and safe and secure

Do you confirm? If it's the case, motu-release should be subscribed, for a FFe.

Thanks,
Fabrice

[Thorsten Glaser (mirabilos)]

Yes, it is an upgrade of the shell which fixes a couple of bugs also (as I am both upstream
and Debian Maintainer, I try to fix things upstream whenever possible), plus a number of
backports from mksh-current (will become R40, but that one will introduce some differences
in the scripting language and isn’t ready yer, whereas R38c→R39 should not break any
existing scripts).

[StefanPotyra (sistpoty)]

+1

[Scott Kitterman (kitterman)]

Ack from me. Approved. Someone with web access please mark confirmed.

[Fabrice Coutadeur (fabricesp)]

taking for sponsorship

[Fabrice Coutadeur (fabricesp)]

Uploaded. Thanks for your contribution!

I've just updated the maintainer field in debian/control, and recover some past changelog entries that has been lost in the merge process.

Thanks,
Fabrice

[Thorsten Glaser (mirabilos)]

@fabrice_sp: thanks (you didn’t necessarily have to change the maintainer though,
at least nobody did that for the last two uploads).

Sadly, #381332 re-appeared, will prepare a fix right now (or rather, integrate the
old fix for that bug).

[Fabrice Coutadeur (fabricesp)]

Hi,

I had to change the maintainer filed, otherwise, I wasn't able to upload it (rule enforce by debuild)

Thanks,
Fabrice