MakeMaker 5.91_02 -> 5.92_01

[p5sagit/p5-mst-13.2.git] / lib / ExtUtils / Changes

5.92_01 Mon Apr 29 23:09:38 EDT 2002
    - Fixing case of modules with no .pm files on VMS.
    - LDLOADLIBS fix for NetBSD and easier overriding (bleadperl 16233)
    * syncing in MM_MacOS from pudge.
    - syncing in NetWare fixes (16266 & 16190)
    - Cleaning up MM_NW5.pm somewhat.

5.91_02 Wed Apr 24 01:29:56 EDT 2002
    - Adjustments to tests for inclusion in the core.

5.91_01 Wed Apr 24 00:11:06 EDT 2002
    [[ API Changes ]]
    * A failing Makefile.PL in a subdir will now kill the whole
      makefile making process.
    * "make install PREFIX=something" will no longer work.  Sorry.
    - Now supporting the usevendorprefix %Config setting
    - Tests now guaranteed to run in alphabetical order.
    - Allowing $VERSION = 0.

    [[ Bug Fixes ]]
    - Missing prerequisite warning malformatted.
    - INSTALL*MAN*DIR and INST_MAN*DIR weren't allowed on the command 
      line.
    * For years now skipcheck() has been returning a different
      value than what was documented.
    - Partially reversing Ken's "speed up ExtUtils::Manifest" patch
      from 5.51_01 so MANIFEST overrides MANIFEST.SKIP.
    * Fixed PREFIXification so it works on Win32.
    * Fixed PREFIXification so it works on VMS.
    - Fixed INSTALLMAN*DIR=none on VMS.
    * NetWare fixes (bleadperl@16076)
    - Craig Berry fixed some macro corruption on VMS.
    - Systems configured to not have man pages now honored thanks to
      Paul Green
    - Hack to allow 5.6.X versions of ExtUtils::Embed use MY implicitly.
    - Moved use of glob out of MM_Unix so MacPerl could build

    [[ Test Changes ]]
    - Shortening directory levels to accomodate old VMS's
    - was using a slightly wrong prefix for the prefix tests

    [[ Doc Fixes ]]
    - Documenting VERBINST

5.90_01 Thu Apr 11 01:11:54 EDT 2002
    [[ API Changes ]]
    * Implementation of the new PREFIX logic.
    * Added new INSTALL targets:
        INSTALLSITEBIN
        INSTALLSITEMAN1DIR   INSTALLSITEMAN3DIR 
        INSTALLVENDORMAN1DIR INSTALLVENDORMAN3DIR
        INSTALLVENDORBIN INSTALLVENDORLIB INSTALLVENDORARCH
        SITEPREFIX VENDORPREFIX
    * INSTALLDIRS=site now properly using $Config{siteprefixexp}
      (Thieved from Debian.  Thanks!)
    * Added INSTALLDIRS=vendor & support for vendor directories
      (Thieved from Debian.  Thanks!)

    [[ Bug Fixes ]]
    - nmake syntax nits from Mattia Barbon
    - ExtUtils::Packlist fix for files with spaces in them from Jan Dubois
      (bleadperl@15800)
    - Old, old, old dmake bug in MM_Win32->pm_to_blib fixed by Nick
      (bleadperl 15847)

    [[ Internals ]]
    - Purging leftover PDP compiler flags.

5.55_03 Sat Apr  6 21:57:01 EST 2002
    - Reversing "fix" for RT 444.  It wasn't really fixed and it
      caused havoc with Compress::Zlib and Tk.

5.55_02 Sat Apr  6 03:09:15 EST 2002
    [[ Bug Fixes ]]
    - Craig Berry fixed install on VMS (again, I munged the last patch)
    - MakeMaker might not be recognizing it's in PERL_SRC on Win32.
      Made safer.
    - For some reason MM_VMS was using $Config{arch} instead of
      $Config{archname}
    - Fix (well, hack) for creating ABSPERL on VMS.
    - Quieting some warnings revealed by ExtUtils::Constant
    - test_via_script had a typo and forgot INST_ARCHLIB

    [[ Test Fixes ]]
    - INST.t has to tell MakeMaker it's in the PERL_CORE
    - Fixing expected values of INST_* when building the core.

5.55_01 Thu Apr  4 23:58:06 EST 2002
    [[ API Changes ]]
    * Long deprecated INST_EXE completely removed.
    - Removing TEST_LIBS constant (never seen in a stable MM anyway)
    - Added $default argument to prefixify() for safer prefixification.

    [[ Bug Fixes ]]
    - $mm->{PERM_*} were not being set
    - fixin() redundantly chmod'ing scripts.  Let the Makefile do that.
    - The above means MM_Unix will now compile cleanly under strict
    * init_dirscan's search for Makefile.PL's will no longer look
      inside a distdir.
    * Fixed running tests when there are t/'s in subdirs.
    - MM_DOS inheriting from MM_Unix instead of MM_Win32.  DJGPP
      appears to be unix-like (bleadperl@15650)
    - Escaping constants with # in them so they're not confused as
      make comments (bleadperl)
    * ExtUtils::MM_Win95 didn't return a true value
    - Fixing disttest when perl is a relative path.
    * Fixed disttest & tests in subdirs on Win95
    - Fixed recursive Makefile.PL scan on VMS so it skips the distdir
    - Fixed a bug when an alternative MAKEFILE is specified and you're
      using something other than a Makefile.PL (RT 444)
    - Quieting uninit warning when there are no tests.

    [[ Doc Fixes ]]
    - Documented INSTALLMAN*DIR == 'none' feature.

    [[ Test Fixes ]]
    - Paul Green's which_perl patch to build a proper Perl on systems
      that use command extensions.
    - basic.t's Makefile.PL PREFIX call wasn't properly quoted for VMS
      (Craig Berry)
    - little initialization glitch in MM_Win32.t

    [[ Internals ]]
    - INST_* constants moved to init_INST
    - INSTALL* constants moved to inst_INSTALL
    * Internal warning supression removed.
    - Temporarily turning off SelfLoader in MM_Unix so warnings
      have the right line numbers.

5.54_01 Sat Mar 30 02:32:44 EST 2002
    [[ New Features ]]
    * Added FULLPERLRUN and FULLPERLRUNINST

    [[ Bug Fixes ]]
    - hint files were made a little too noisy in 5.53_01

    [[ Test Fixes ]]
    - adjusting for NFS time drift
    - basic.t was finding the wrong perl in the core

    [[ Internals ]]
    - FULLPERL* PERL* setup moved to init_PERL

5.53_02 Fri Mar 29 04:47:44 EST 2002
    - Adjusted for Perl core.  Synced into bleadperl.

5.53_01 Fri Mar 29 00:13:02 EST 2002
    [[ Bug Fixes ]]
    - Removed duplicate NAME macro definition
    - Fixed makemakerdflt target for VMS
    * bleadperl@11460 introduced a bug with recursive Makefile.PL's not
      having '.' in @INC.  PDL should build now.
    * MANIFEST.SKIP only working on file names, not full relative paths.
      Only since 5.51_01  (RT 413)
    * make test TEST_VERBOSE=1 was busted by 5.50_01 (RT 423)
    * Error messages from dying Makefile.PL's and hint files were
      accidentally supressed in 5.48_04.  Fixed.
    * Makefile.PL's are supposed to be run in package main.  5.48_04
      broke this.  Fixed.
    * Fixing installing on VMS.

5.52_01 Tue Mar 26 00:24:26 EST 2002
    [[ Bug Fixes ]]
    * ActivePerl 5.6.1/build 631 now 100%
    - Fixed some SelfLoader warnings
    * ExtUtils::MM_Win32 not subclassed off of ExtUtils::MM_Any properly
    - Bug in local $ENV{FOO} was causing failures in MM_Win32 tests.
      Compensating for the bug.
    - $Config{prefixexp} is sometimes bad on ActivePerl, compensating
    - Accidentally left htmlifypods tests in MM_Win32 tests

    [[ Doc Changes ]]
    - Expanding Known Good list
    - Adding Known Programs to README

5.51_01 Mon Mar 18 01:37:02 EST 2002
    [[ API Changes ]]
    - Removing xsubpp and typemap from the distribution.  These are not
      Perl version independent files and should not be upgraded.
    - Removing ExtUtils::Embed.  Version specific module and should
      not be updated.
    - Removing ExtUtils::Constant.  Not directly tied to MakeMaker.
      Will be distributed seperately by Nick Clark.

    [[ New Features ]]
    * realclean now deletes 'dist' directory.

    [[ Bug Fixes ]]
    * Fixing ExtUtils::Installed for VMS
    * Fixed it so MakeMaker can build itself without needing an eariler
      version installed
    * Fixed ExtUtils::Installed so packlists work on VMS
    * ExtUtils::MM_VMS test had a stupid typo that prevented most
      of the tests from running.
    - Fixing VMS so 'mmk' is always 'mmk all' (bleadperl 15218)
    - ExtUtils::MM_Any->catfile was calling catdir
    - Added Ken William's "speed up ExtUtils::Manifest" patch.
    - Added Nick Clark's return value of manifest routines patch
      (bleadperl@14978)
    - Merging in bleadperl changes (14438, 14453, 14611, 14612, 14622,
      14655, 14677, 14688, 14902, 15023, 15187, 15367)
    - bleadperl change 15325 (VMS 'mmk all' hack) rejected.
    - ExtUtils::MM_Any->test_via_harness() now using -MExtUtils::testlib
      instead of -I's.  This keeps the command line shorter on VMS, but
      it means TEST_LIB doesn't work anymore.  May prove problematic.
    - PERLRUN now uses FULLPERL instead of PERL.  This avoids 
      accidental use of miniperl.  May cause problems in the core.
    - Fixed test_via_harness() on VMS so it uses PERLRUN.
    - ExtUtils::Manifest wrongly handling MANIFEST on VMS.
    - ExtUtils::Manifest::maniskip broken due to misuse of /o on a regex.

5.50_01 Mon Mar  4 23:44:48 EST 2002
    [[ API Changes ]]
      htmlifypods and all HTML targets have been removed.  It didn't
      work properly.  A target for HTMLifying of docs will be
      added sometime in the future (read: send a patch).
    - Deprecated ROOTEXT variable removed (its been deprecated for
      more than five years).
    - Removed ExtUtils::Miniperl from the CPAN distribution.  Can't
      see how its useful outside the core.

    [[ New Features ]]
    * Emacs backup files (ie. *~) are no longer copied into blib
      (this means you won't get Foo.pm~ accidentally installed anymore).
    - prefixify() now returns if the prefixification worked or not.
    - added the Perl patchlevel info to the description embedded in DLLs
      (Ilya Zakharevich perl change 14810)

    [[ Bug Fixes ]]
    * 5.49_01 broke anything that depended on the MM class to be loaded
      with ExtUtils::MakeMaker, like CPAN.  Temporarily fixed.
    * Many places wrongfully assume MM methods can be called as class
      methods!  Inside and outside of MakeMaker.
    * DOS now acts like Win32 instead of Unix.  This should be less wrong.
    - Netware tweak from Ananth Kesari (perl change 14780)

    [[ Doc Changes ]]
    * made the docs about the behavior of PREFIX a bit more vague
      reflecting its odd behavior.    
    - Replaced references to %Config with things people are more likely
      to understand in PREFIX & LIB docs.
    - Put PREFIX before LIB in the docs.

    [[ Internal Refactorings ]]
    - File::Spec wrappers consolidated in MM_Any
    - test_via_harness/script consolidated in MM_Any
    * Added ExtUtils::Command::MM to replace large -e "one-liners"
    * Added ExtUtils::MM_UWIN, took UWIN specific code out of MM_Unix.
    * Added ExtUtils::MM_DOS, took some DOS specific code out of MM_Unix
    - Added a dist() override to MM_OS2.
    - Took the OS/2 specific code out of MM_Unix's dist()
    - Starting to quote as many command line switches as possible
      to make VMS have to override less of MM_Unix.

5.49_01 Mon Feb  4 00:42:40 EST 2002
    - Default TEST_LIBS now contains absolute paths, so tests can
      chdir without breaking @INC.
    - Some bug fixes in MM_Win32 were missing in MM_NW5.
    - LLIBPERL was not being set.
    - Gisle Aas fixed a warning in prompt when the user hits ctrl-D
      or pipes STDIN to /dev/null.
    - VMS's test is now a proper ->can check.
    - Stray newline in Command.t causing havoc (bleadperl 14441)

    * Lots of internal changes.  Everything from here on is an internal
      change.
    - Broke ExtUtils::Liblist::Kid out into its own .pm.  Temporary.
    - Broke MM and MY out into their own .pm's.  Possibly temporary.
    - Broke ExtUtils::MM_Win95 out of ExtUtils::MM_Win32.  MM_Win95
      is now an MM_Win32 subclass.
    - Removed wrapper methods from ExtUtils::Liblist that were
      defering File::Spec loading.  Almost everything uses File::Spec
    - Changed most of the 'our's to 'use vars' in prep for 5.005_03
      backporting
    - Changed ExtUtils::MakeMaker->import(...) cargo-cultery in MM_* 
      modules to proper "use ExtUtils::MakeMaker qw(...)"
    - All non-Unix MM_* modules now directly inherit from MM_Unix.
      They did this before, but in a round-about way.
    - MM_* modules no longer directly muck with @MM::ISA.  Instead
      @MM::ISA does that itself.
    - Removed unnecessary require of Exporter in many MM_* modules.
    - MM_Cygwin was using an MM_Unix function directly without have
      explicitly required MM_Unix.
    - Most of MM_NW5 was redundant with MM_Win32.  So MM_NW5 is now
      a subclass of Win32 and the reduendant code has been deleted.
    - Replaced lots of calls to File::Spec->updir, curdir and rootdir
      with a global in MM_Unix.  Should make things a bit faster.
    - Untabified ExtUtils::MakeMaker.  I hate tabs.
    - "Which MM subclass should I use" logic moved from EU::MakeMaker
      to EU::MM.
    - Deprecated EU::MakeMaker::Version_check deleted.

5.48_04 Mon Jan 21 19:28:38 EST 2002
    - No longer requiring Test::More
    - Now tested to work with a clean 5.6.1
    - Stripped out $Revision based $VERSIONs.
    - Rolled in patches from p5p, vmsperl & VOS (14325, 14338, 14360, 14364)
    * hint files now run in a safer context.  Only $self is available.
    - ExtUtils::testlib now provides absolute paths into @INC.
      No longer obsolete
    - Little test fixes

5.48_03 Thu Jan 17 23:44:38 EST 2002
    * moved bin/inst to bin/instmodsh
    * Some Win32 backporting fixes.  The -x switch doesn't seem to
      work on Win32/5.6.1.
    * Bug on Win32.  MAKEMAKER variable not set properly.
    * _02 broke hints files.  Now fixed.
    - Minor prereq error formatting glitch
    - ExtUtils::testlib no longer a thin wrapper around blib, now a 
      thin wrapper around lib like it was.  blib is noisy on 5.6.1.
    - Nick and chromatic found and fixed some warnings in the tests

5.48_02 Wed Jan 16 19:11:26 EST 2002
    - Fixed some Win32 warnings.  Needs more work.

5.48_01 Wed Jan 16 15:10:28 EST 2002
    * Adapted from bleadperl@14303
    * Backported to 5.6.1
    - inst and xsubpp moved out of lib/ and into bin/
    - ExtUtils::testlib obsoleted.  Now a thin wrapper around blib.pm
    ***--> Non-Unix platforms not tested at all!  Please test and report
           back.  Tests probably only need minor fixes.