Please sync mksh (40.4-1) from Debian unstable into universe

[Thorsten Glaser]

The Ubuntu armel and armhf and the Debian armhf (funnily enough, not armel) build of mksh 40.3-1 contain a fully broken /bin/mksh-static binary, which was not caught by the regression testsuite due to not good enough failback behaviour. It is now caught, and the Debian armhf 40.4-1 build lets the dietlibc mksh-static fail and falls back to using eglibc, which produces a working (albeit somewhat huge) binary.

Please do the sync from Debian now, to supersede the broken versions from the Ubuntu archive early.

Everything else will have to be done on the dietlibc side (i.e. reducing testcases and investigating why ARM platforms starting from certain compiler flags produce broken executables). The effect of this sync (asides from pulling a new upstream bugfix release) is to recognise these broken executables during build, fail them and fall back to eglibc.

Thanks for your effort.

[Micah Gersten (micahg)]

Taking a look

[Micah Gersten (micahg)]

Thank you for requesting this sync. This builds fine on amd64, so I will sync it shortly. I appreciate you following up on the previous regression.

[Micah Gersten (micahg)]

This bug was fixed in the package mksh - 40.4-1

---------------
mksh (40.4-1) unstable; urgency=low

  * The “not-so-little-any-more brother” upload
  * Update to mksh R40d (release):
    - [tg, Jilles Tjoelker] test(1) built-in behaves exactly as POSIX says
    - [tg] Move compile-time assertions to Build.sh from misc.c#ifdef DEBUG
    - [tg] Invocation documentation is at the bottom of Build.sh
    - [tg] test.sh: verbosely look for perl(1) interpreter to use
    - [tg] New tests for integers (base 1‥36, base unspecified, base OOB)
    - [tg] Correct error paths for typeset -n global state
    - [tg] Deprecate interpreting "010" as octal number, will go
    - [tg] Improvements re. integer handling; more explicit manpage text
    - [tg] Do not use caddr_t on Linux, so dietlibc stops bitching
    - [tg, Jilles Tjoelker] Catch division/modulo overflow 0x80000000/-1
    - [tg] Emacs mode ^O regression fix when the fetched lines are edited
  * Add missing upstream code change explanations to the changelog
    entries 40.3-1, 40.2-5, 40.2-4 (in lieu of upstream-changelog)
  * Fix two oversights that might (finally) get us line-buffered I/O
  * Disable use of <libutil.h> to avoid needing to B-C libbsd-dev
  * Use hardening=+all in DEB_BUILD_MAINT_OPTIONS querying buildflags
  * mtest.t: new brkcontin to catch breakage failing test.sh operation
  * Also, scan the test log and fail the build if the testsuite
    appears not having run at all (Closes: #651615)

 -- Thorsten Glaser <email address hidden> Sun, 11 Dec 2011 18:51:39 +0000