Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-04 12:17:07 -0700 (Wed, 04 Aug 2004)
Revision: 476
Log message:

      Fixed naming of the omake_ast_{lex,parse} files.
      

Changes by: ( at unknown.email)
Date: 2004-08-04 12:17:07 -0700 (Wed, 04 Aug 2004)
Revision: 477
Log message:

      This commit was manufactured by cvs2svn to create branch 'shell_begin'.

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-04 12:55:06 -0700 (Wed, 04 Aug 2004)
Revision: 479
Log message:

      NOTICE: this is a branch commit, it isn't fully implemented yet.
      
      This adds an internal shell to omake, so we can get some consistency
      across platforms.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-05 09:49:58 -0700 (Thu, 05 Aug 2004)
Revision: 480
Log message:

      Fixed a problem with threads and pipes.  For example, the command
      
      echo 1 | cat
      
      would block, because the handle duplicates used by the thread
      for the echo would be inherited by the cat.  So cat would have
      a handle to its own input, and the pipe would never get closed.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-15 20:49:35 -0700 (Sun, 15 Aug 2004)
Revision: 482
Log message:

      This nearly finishes this round of changes.  We get several things, mainly
      a shell and data structures.  I have to complete a manual section describing
      objects.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-16 07:54:48 -0700 (Mon, 16 Aug 2004)
Revision: 483
Log message:

      Fixed the Not_found exception Ron reported.  The problem was with
      expanding effects sets.  As the effects are computed, the set
      may expand to include targets for which commands have not been
      constructed.  This change starts the commands in the effects
      fixpoint loop.
      
      On Sat, 7 Aug 2004 07:22:46 -0400, Yaron Minsky <yminsky@gmail.com> wrote:
      > I've been playing around with 0.8.3 and have found a situation where
      > omake dies with a Not_found exception whenever I try to compile
      > something that doesn't have an mli file.  I've attached my OMakefile.
      > I'm just trying to produce a simple OMakefile that cooperates with
      > ocamlfind.  I'm assuming that omake shouldn't ever just die with an
      > exception like this.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-16 08:52:26 -0700 (Mon, 16 Aug 2004)
Revision: 484
Log message:

      Added BYTE_ENABLED and NATIVE_ENABLED checks.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2004-08-19 22:36:19 -0700 (Thu, 19 Aug 2004)
Revision: 485
Log message:

      If the first argument to filter-out does not have a wildcard character,
      then just take it to be a literal string (instead of raising Invalid_argument).
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-23 09:53:13 -0700 (Mon, 23 Aug 2004)
Revision: 487
Log message:

      Added the lexer and parser generators.  Eventually, these are
      intended to replace the Phobos modules.
      
      This branch is nearly complete, I just need to add pseudo-awk
      and sed functions, then make sure it all works on Win32.
      And document it: printf, lexer, parser, objects.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-23 13:34:18 -0700 (Mon, 23 Aug 2004)
Revision: 488
Log message:

      Added the simple awk/sed functions.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-24 16:38:58 -0700 (Tue, 24 Aug 2004)
Revision: 489
Log message:

      Reorganized the code and added documentation.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-26 12:06:22 -0700 (Thu, 26 Aug 2004)
Revision: 492
Log message:

      Added a lot of documentation and cleaned up generally.
      
      Added dependency scanning for .c files on Win32 (we never had it before).
      Grunge.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-26 18:58:43 -0700 (Thu, 26 Aug 2004)
Revision: 494
Log message:

      Hmmm, its a little bit strange to name a file or directory "core"
      on a Unix system.  By default this is in the set of cvs ignored
      files.  I removed it from the default list in lm_glob.ml
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-26 19:21:33 -0700 (Thu, 26 Aug 2004)
Revision: 495
Log message:

      Just get things to compile on Win32.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-27 20:19:25 -0700 (Fri, 27 Aug 2004)
Revision: 496
Log message:

      This gets MetaPRL to compile in Win32.
      These are various bigfixes.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-27 20:39:03 -0700 (Fri, 27 Aug 2004)
Revision: 498
Log message:

      Updated documentation.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-28 19:02:32 -0700 (Sat, 28 Aug 2004)
Revision: 499
Log message:

      Add line number information to the channels.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-29 16:07:42 -0700 (Sun, 29 Aug 2004)
Revision: 500
Log message:

      Changed scoping rules to prefer dynamic scoping everywhere.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-29 17:41:49 -0700 (Sun, 29 Aug 2004)
Revision: 501
Log message:

      For ASOUT, -o should be a separate argument.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-29 19:00:10 -0700 (Sun, 29 Aug 2004)
Revision: 502
Log message:

      This is a merge from the shell_begin branch.
      Let me know if there are problems!
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-29 19:14:50 -0700 (Sun, 29 Aug 2004)
Revision: 504
Log message:

      More changes to .cvsignore files.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2004-08-30 02:45:34 -0700 (Mon, 30 Aug 2004)
Revision: 505
Log message:

      Fixing a comment (for some reason if ".INCLUDE" was used at a beginning
      of the line inside a verbatim block in comments, the ".IN" still ended up
      being at the beginning of the line in omake.1 and would end up being
      interpreted as a macro).
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-30 08:42:05 -0700 (Mon, 30 Aug 2004)
Revision: 506
Log message:

      Fixed the major problems with building documentation.
      
      However, there is a problem lurking here.  Consider the
      following rule:
      
         foo: bar
            if $(predicate ...)
               command
      
      Currently, expressions in rules are evaluated directly, not in
      a subjob.  That means that the select loop is not operating while
      the command is being executed.  If it generates lots of output,
      it will block.
      
      I'll fix this later today, but I've added a workaround in the meantime
      by diverting stdout.
      
         foo: bar
            if $(predicate ...)
               stdout = $(fopen foo.out, w)
               command
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-30 09:31:54 -0700 (Mon, 30 Aug 2004)
Revision: 507
Log message:

      Corrected the FamErrlist[] name in the pseudo-modules.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-30 18:12:38 -0700 (Mon, 30 Aug 2004)
Revision: 508
Log message:

      Completely turn off process group management unless in interactive
      mode.  This should address Aleksey's problem with documentation
      generation.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-30 18:59:36 -0700 (Mon, 30 Aug 2004)
Revision: 509
Log message:

      Execute inline rule expressions in a separate thread/process.
      This fixes the problem of blocking while executing rules with
      a body that is an expression.
      
         target: src
            f()
      
      It used to be that omake would block if f() generated lots of output.
      Now f() is evaluated in a separate thread (on Win32) or process (forked
      in Unix).  This allows omake to continue normal processing.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2004-08-30 20:20:08 -0700 (Mon, 30 Aug 2004)
Revision: 510
Log message:

      Fixed an off-by-one error in Lm_channel.squash_text \r\n->\n
      line-ending transformation.  Win32 can be annoying:(