source: trunk/essentials/dev-lang/perl/lib/ExtUtils/Changes

6.30  Fri May 20 16:05:38 PDT 2005
    * PL_FILES behavior tweak again to restore old behavior.  Sometimes its 
      supposed to run before pm_to_blib, sometimes after.
    - Some tests shipped with 'no_plan' which will break on older 
      Test::Harness.

6.29  Thu May 19 14:15:21 PDT 2005
    * The behavior of PL_FILES is restored to its pre-6.26 behavior as several
      CPAN modules depend on this.  PL programs run via PL_FILES have 
      INST_LIB and INST_ARCH in their @INC and so can load any just built 
      modules.
    - Now honors PERL_CORE environment variable.
    - Testing to ensure FIRST_MAKEFILE is honored.

6.28  Tue Apr 12 16:17:07 PDT 2005
    - Fix realclean so it cleans up files installed from ext/ in the core
    - Fix dir_target() so it doesn't warn should any of the INST_* paths
      be the same (as with the ext/ modules in the core)
    - Fix MANIFEST.SKIP so it skips not just _darcs/ but everything inside
      it and any which happen to be in subdirs.
    - MM_AIX forgot to import neatvalue() from E::MakeMaker. 
      (bleadperl@24185)
    - Fixed a minor C<<>> POD nit (Scott Lanning)

6.27  Mon Apr  4 16:36:14 PDT 2005
    * Added _darcs to the list of revision control administrative 
      directories skipped both in libscan and in MANIFEST.SKIP.

6.26_01 Mon Mar 28 21:34:39 PST 2005
    * PL_FILES was broken in the last release.  The .PL files were not being
      passed the file they were to generate.
    * How PL_FILES runs the programs and what it does with the value is 
      now documented.
    * The default behavior of PL_FILES is now documented.

6.26  Mon Mar 21 20:36:22 PST 2005

6.25_12 Fri Mar 18 16:15:35 PST 2005
    - Use the generic dmake search.cpan.org URL in the README so the newest
      version is found.
    - Fixed pm_to_blib trouble on VMS introduced in 6.25_10.

6.25_11 Tue Mar 15 02:00:20 PST 2005
    - Fixed minor parent class dependency loop in ExtUtils::MM_Any.
    - Fixed a circular dependency on processed PL files introduced as part
      of RT 6460 in 6.25_02.

6.25_10 Sun Mar 13 16:14:04 PST 2005
    - pm_to_blib.ts is gone, back to pm_to_blib (except on VMS).  I 
      couldn't make it work and still retain backwards compatibility.
      This fixes the recompilation problems that plagued _09 and various
      previous alpha releases.

6.25_09  Sat Mar 12 10:10:34 PST 2005
    - Stubbing out PASTHRU on VMS.  $(MMSQUALIFIERS) should do the equivalent
      and better.
    - Added test to check for the existence of a compiler so XS tests are
      now possible.

6.25_08  Tue Feb  8 09:04:36 EST 2005
    - Fixing xsubpp on VMS.  Had a foo/bar path in it.
    - Turning XSUBPP into a runable command by the addition of PERLRUN
      made it unusable as a dependency.  Added XSUBPPRUN for that and left
      XSUBPP alone.
    - Removed DESTDIR from PASTHRU as install is not recursive and it can
      make commands too long on VMS.
    - Older versions of MMS cannot handle directory targets at all.  Using
      .exists files everywhere.
    - Restore missing *.Opt clean on VMS.
    - Fix .exists files accidentally getting caught in static builds.
    - Older versions of MMS cannot handle directory targets at all.
    - Add Interix support (bleadperl@33892)
    - distsign target changed to distsignature to avoid clash with
      Module::Install

6.25_07  Fri Dec 31 03:47:20 EST 2004
    - perllocal on VMS was inserting executables twice.
    - No longer using $(IGNORE) macro.  Turns out MMS/K was not honoring
      it.  Using "-$(NOECHO) command" which seems to make everybody happy.
    - Executables with no extension weren't getting installed on VMS due to 
      a bug in rename().  Broken sometime in this series of alphas.

6.25_06  Sun Dec 26 17:21:37 EST 2004
    - Forgot to define BOOTDEP macro.
    - .exists files are back.  Directories cannot be used directly as 
      targets as their mod time changes too frequently.
    * Added INSTALLBASE as an alternative to PREFIX but haven't documented
      it yet.  I'll do that next release.

6.25_05  Wed Dec 22 07:59:02 EST 2004
    - One of the 6.25 alphas broke BSD make.  It doesn't like "- @ command".
      Fixed by adding an $(IGNORE) macro.
    - 6.25 alphas caused a Makefile to be added to the dist.  Fixed.
    - The new cd() code needed to be dependent on dmake or nmake for
      Windows.  Not Win9x vs WinNT/XP.

6.25_04  Tue Dec 21 00:53:06 EST 2004
    - 6.25_03 was always rebuilding XS modules.

6.25_03  Mon Dec 20 23:04:22 EST 2004
    - dir_target() is back.  Now each directory to be created has its own 
      target like before, but no more .exists or blibdirs.ts files.  This
      ensures that each blib directory is created as necessary and fixes
      things like SVN's perl bindings.

6.25_02  Mon Dec 20 03:31:49 EST 2004
    - Set PM_FILTER as late as possible so it can see all the earlier
      macro definitions.  Necessary for challenged make implementations
      like nmake.  Should fix Mail::SpamAssassin installs on Win32.
      [rt.cpan.org 4545]
    - clean and realclean are now more careful about accidentally deleting
      directories instead of files.  [rt.cpan.org 6851]
    - small fix for parallel builds, make sure pm_to_blib has run before
      we try to use stuff in blib. [rt.cpan.org 6460]
    - MAKEFILE=foo appears to have been broken for recursive builds and
      several other things.  I think this was broken by 6.18.

6.25_01  Fri Dec 17 21:29:04 EST 2004
    * *.bak added to the default MANIFEST.SKIP.
    * META.yml will no longer be generated in the build directory.  It will
      only appear in the distdir.  This should make it easier on developers,
      they don't have to worry about checking the file in all the time.
    * Similarly, the SIGNATURE file will not be updated in the build 
      directory.  It will only be generated in the distdir.
    - A bunch of redundant Win9x and VMS code removed.
    - 'make test' on Windows no longer pre-expands its list of test files.
      This caused problems on large distributions like bioperl.  Thanks to
      Tim Bunce for suggesting the obvious fix.

6.25  Wed Dec 15 06:59:46 EST 2004
    - Build.PL was being considered like Module_pm.PL.  Build.PL is now 
      ignored.  [[email protected]] [rt.cpan.org 8809]
    - Devel::Cover cover_db/ directory now ignored by MANIFEST.SKIP

6.24_01 Thu Dec  9 00:44:48 EST 2004
    - Docs encouraged possibly insecure use of /tmp [CAN-2004-0976 via Debian]
    - Remove empty rpath from .so files [bleadperl@23183]
    - At long last the core man page title has its apostrophe.
    - BeOS tests fixed [Ingo Weinhold] [perlbug #32717]
    - Putting a pm_to_blib target back (alias of pm_to_blib.ts) in case
      anyone was depending on it being there (mod_perl was).

6.24 Tue Nov 30 15:35:13 EST 2004
    - dir_target() was accidentally named dir_targets() causing problems
      installing some Tk modules.

6.23 Fri Nov 26 16:01:50 EST 2004
    - MM_MacOS lost its $VERSION causing some CPAN indexing problems.  Enough
      people noticed this to warrant a quick release.

6.22 Tue Nov 23 22:22:22 EST 2004
    - Minor glitch in Command.t test on Win32 [thanks Steve Hay]
    * Fixed MANIFEST.SKIP so it avoids all the new *.ts files.
    - Don't spuriously define pasthru arguments

6.21_03 Sat Apr 3 2004
    - pm_to_blib and blibdirs stamp files now pm_to_blib.ts and blibdirs.ts
      everywhere because some filesystems insist you have a dot in the 
      filename.  This eliminates a bunch of special case code.
    - Minor potential glitch in the version check logic.
    - Making Makefile rebuild logic more resistant to inconsequential
      errors when cleaning up things that are going to be rebuilt anyway.
    - Adding SCCS to the list of directories in MANIFEST.SKIP.
    * Fixing blibdirs so it doesn't rebuild every make run.
    * Using a more conservative MAX_EXEC_LEN for Win98 to allow building
      large modules like Encode. [Greg Matheson]
    * test target mangled by dmake.  Doesn't seem to like q{}.
      [Greg Matheson]
    - instmodsh now using Archive::Tar to create archives if available.
      [Slaven Rezic]
    - instmodsh now using more portable tar flags to create archives.
      [Slaven Rezic]
    - Added FAQ entry for "How do I use a module without installing it?"
      and "How do I keep from installing man pages?" [Joe Cromie]
    - Fixed finding the default MANIFEST.SKIP if its on a different volume
      [Ilya Zakharevich]
    - When building Perl, we had a chance of picking up the old installed
      xsubpp from @INC [Andreas Koenig]
    - makeaperl() now a bit more intellegent about ignoring perl libraries
      not called libperl.  [Ilya Zakharevich bleadperl@22032]
    * MacOS Classic (MacPerl) is no longer supported.  Please use
      Module::Build instead.
    * Restored pm_to_blib.ts so pm_to_blib generation will be quiessent on
      VMS. [rt 4675]
    - For simplicity, dir_target() is now a no-op.  Should not effect any
      existing uses.
    - Converted utility commands from using $(PERLRUN) to $(ABSPERLRUN).
      This allows distclean to work with SDBM_File. [rt 5616]
    - realclean was using "rm -f" instead of "rm -rf" meaning directories
      wouldn't be properly cleaned.  Caused by bleadperl@7952. [rt 5208]
    - fixpath() could undo wraplist() on VMS [rt 4955]
    - ensure ExtUtils::Command::rm_f deletes all versions of a file on VMS
      [rt 4687]
    - hint.t generating improperly named hint files on OS's where $^O
      contains a '.'.  They should be converted to underscores.  Test
      mistake, not a bug.  [rt 5365]
    - Fixed ExtUtils::Command::chmod() so it will work on directories on
      VMS. [rt 4676]
    - parse_version.t was using no_plan causing trouble on older T::Hs.
      [rt 5633]

6.21 Tue Nov 11 00:12:56 PST 2003
    - NetBSD was looking in INSTALLARCHLIB/CORE for libperl instead of 
      PERL_ARCHLIB/CORE.  Would cause problems if INSTALLARCHLIB was changed
      (ie. LIB or PREFIX used). [Jochen Eisinger]
    - Turns out a handful of modules use dir_target().  Restored a version
      for backwards compatibility.
    - Moved blibdirs target from top_targets() to its own section.  Lots of
      modules rewrite top_targets() so blibdirs wouldn't be written.

6.20 Thu Nov  6 02:25:58 PST 2003
    - Fixing dos2unix on Cygwin.  In-place editing doesn't work 100% so we
      take a more conservative approach.
    - Small postamble.t test temp file cleanup fix for Cygwin and Windows.
    - Small Command.t test fix for 5.5.3.  No real bug [rt 4290]
    - Small Liblist fix for Cygwin and import libraries [Gerrit P. Haase]
    - metafile error message slightly mangled.  Bare $! mistake confusing 
      nmake [rt 4285 4301]

6.19 Mon Nov  3 22:53:56 PST 2003
    - Removed 00setup_dummy.t and zz_cleanup_dummy.t.  Tests now setup and
      teardown the dummy modules as needed.
    - Little test glitch on systems without $Config{usevendorprefix} set
    - INSTALL(SITE|VENDOR)MAN*DIR now fall back to $(INSTALLMAN*DIR)
      instead of copying its value.  This lets "perl Makefile.PL 
      INSTALLMAN1DIR=/some/man/man1" work like expected.

6.18 Mon Nov  3 20:09:51 PST 2003
    - parse_version() was blowing over $_ (Ilya Zakharevich)
    - 5.6.2 has the same Cygwin shared lib name fix as 5.7.0. (Gerrit Haase)
    * When $Config{install(site|vendor)man*dir} aren't set it will now
      fall back to $Config{installman*dir}.  Similar for 
      $Config{install(site|vendor)bin}.  This preserves behavior for older
      Perls.
    * Eliminated the dependency on perl.h for pure-perl builds.
    * Eliminated .exists files.  Added blibdirs target using MKPATH instead.
    - Removed now unused dir_target() method.
    - Added a little documentation to instmodsh.
    * Made sure PREFIX always has something in it.  Module authors like to use
      it when extending MakeMaker.  Currently using one of 
      PERL/SITE/VENDORPREFIX based on the value of INSTALLDIRS.
    * Added SIGN to generate module signatures with cpansign (Autrijus Tang)
    - Slight tweaks to MAKEFILE_OLD and MAKE_APERL_FILE restoring a mistake
      made around 6.06 to fix static builds on OS/2
    - Added ExtUtils::Command::dos2unix()
    * manicopy() would make the copied files read-only if 'cp' was used.
    * Fixed prereq check for modules which are also keywords like if.pm and
      open.pm (Autrijus Tang)
    - Made ExtUtils::Command mv and cp return whether or not they succeeded.
    * metafile_addtomanifest now mentions what its doing.
    * metafile will not die if it can't write to META.yml, just emit a 
      warning.  This lets 'make dist' continue with a read-only META.yml
    - Small fix to identify UWIN on NT (Randy Sims)
    * metafile will not touch the META.yml if it does not need to be changed
      (Thanks to Tim Bunce for this idea)
    * Fixed 'make uninstall' which looks like its been broken since last
      November and nobody noticed.
    - Fixing minor bug in VMS prefixification logic when PREFIX was not set
      by the user.
    - Fixing potential bug in prefixification for relative directories
    * Documented UNINST=1

6.17 Sun Sep 14 20:52:45 PDT 2003
    - Fixing LD so it can be set properly on Win32. (Orton Yves)
    * Fixing the init_PERL() "tack $Config{exe_ext} onto $^X" logic so 
      it works on VMS allowing $^X to be found.
    * Fixing make install UNINST=1 on Windows so it can search PERL5LIB
      (Orton Yves)
    - Fixing "could not open MANIFEST" test for OS/2 (and probably others) 
      where a read-only file cannot be deleted.
    - Hacking around the fact that when built for debugging, VMS creates
      ndbgperl.exe.
    - Eliminating the never implemented "help" argument.

6.16 Mon Aug 18 01:39:51 PDT 2003
    * Fixing the max exec length for Windows to account for old
      versions of nmake (the one you can download for free from MS).
    - Hack to normalize //foo//bar paths in ExtUtils::Installed.
      [Sreeji K Das]
    * Adding Module::Build utility files to MANIFEST.SKIP
    - PREREQ_PRINT is supposed to exit after dumping. (Wolfgang Friebel
      bleadperl@20530)
    - maniadd() was trying to open the MANIFEST even if it didn't need to
      causing unnecessary errors for read-only MANIFESTs (rt.cpan.org 3180)

6.15 Sun Aug  3 16:41:05 PDT 2003
    - Make FIXIN on Win32 use the new pl2bat rather than an already
      installed one when building the core (Mattia Barbon)

6.14 Sun Aug  3 16:23:27 PDT 2003
    - Adding a reference to the Module::Build META.yml spec in the META.yml
      itself and the metafile documentation.
    - Removed Begin/End comments from manifypods that were getting in the
      way of users adding code to the manifypods target.
    - Putting a POD2MAN macro back as an alias to POD2MAN_EXE and using
      it instead of POD2MAN.  This allows people to generate their own man 
      installation targets in a backwards compatible way.
    * Adding reference to the Module::Build META file spec in the docs.

6.13 Thu Jul 31 16:48:01 PDT 2003
    - Generated META.yml now has the prereqs in sorted order (Andy Lester)
    - Arguments to child Makefile.PLs were still having .. prepended.  This
      was exposed by the earlier fix of rt.perl.org 4345.
    - extralibs.ld (for static builds) wasn't being cleaned up.

6.12 Tue Jul 29 22:19:38 PDT 2003
    - Should any of your prefixes be /, MakeMaker might lose this and
      prepend your prefix as "prefixyour/directory" rather than
      "prefix/your/directory" (Ed Moy bleadperl 20283)
    - Reversing "realclean was duplicating work already done by clean"
      (CVS 1.124 MM_Unix.pm).  Turns out it wasn't a duplication and this
      was causing INST_* files to be left behind if your INST_LIB wasn't
      blib/* as in the perl core.  realclean now removes the files copied
      to INST_*.
    - Eliminating xsubpp_version().  No longer relevant.  It was checking
      for a feature added to xsubpp before 5.003_07 and doing it in a
      really bizarre way.

6.11 Thu Jul 24 01:24:37 PDT 2003
    - Small grammatical doc touch up by Sean Burke
    - Making basic.t clean up better after itself so it can be run
      twice without setting up and cleaning up Big-Dummy again.
    * Noting where to find make in the README
    * Noting that MakeMaker does not produce a GNU make compatible Makefile
      on Windows.

6.10_08 Mon Jul 21 18:17:06 PDT 2003
    * $Config{installvendorman*dir} was being ignored and our default
      always used. [rt.cpan.org #2949]
    * Setting NORECURS to true was still resulting in a Makefile which
      tried to recurse into subdirectories which contained a Makefile.PL
      (it simply wouldn't have run them to generate Makefile's).
      This appears to be a rather old bug.
      [rt.cpan.org #2951]
    * [rt.perl.org #4345] Arguments passed to WriteMakefile() would get
      stopped upon by $self->{PARENT} in recursive builds in the process
      of prepending a .. onto it.
    - Command.t now using its own temp directory when testing 
      expand_wildcards() so no stray files confuse it.
    * No longer warning about 'postamble' attribute.
    - Added installvars() to list INSTALL* variables so we don't keep
      repeating the list all over the code.
    - Added init_DEST() to create DESTINSTALL* variables for installing
      with DESTDIR set.
    * DESTDIR + PREFIX now does not duplicate DESTDIR.  [Redhat #91892
      and part of rt.cpan.org #2954]
    * DESTDIR is now not directly prepended to the INSTALL* variables.
      installation targets now use DESTINSTALL* instead.  This prevents
      the DESTDIR from showing up in places it shouldn't like perllocal.pod.
      DESTDIR is still appearing in .packlist (bug). [rt.cpan.org #2954]
    - MM_VMS->prefixify() wasn't honoring the %Config_Override causing
      some MakeMaker generated config variables to be ignored.
    - MM_VMS->prefixify() was only VMSifying part of its filepaths causing 
      internal comparisions to get confused and think it had to prefixify 
      when it didn't.

6.10_07 Sat Jul  5 16:12:52 PDT 2003
    * Fixing location of perllocal.pod so its always in INSTALLARCHLIB
      instead of one for each perl, site and vendor.
    - Sun's make treats escaped newlines oddly.  Switching oneliner() back
      to using multiple -e's.
    * Bug in libscan() meant that any .pm containing RCS or CVS or SCCS