Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-01 14:14:24 -0800 (Thu, 01 Mar 2007)
Revision: 10132
Log message:

      Maximal prefix is based on the _minimum_ length, not the maximum.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-01 23:18:56 -0800 (Thu, 01 Mar 2007)
Revision: 10135
Log message:

      At least get 0.9.9 to compile.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-02 09:48:35 -0800 (Fri, 02 Mar 2007)
Revision: 10136
Log message:

      - Added a section on running OMake from a subdirectory (including a subsection
        on the hierarchy of the relative phony targets) - largely based on my email
        to Benjamin Pierce -
        http://lists.metaprl.org/pipermail/omake/2006-June/001043.html
      
      - Regenerated the documentation.
      
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-04 08:00:57 -0800 (Sun, 04 Mar 2007)
Revision: 10152
Log message:

      Rename some case-insensitive conflicts.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-04 08:28:53 -0800 (Sun, 04 Mar 2007)
Revision: 10153
Log message:

      I'd like to be able to use GIT for a little while, in the hope that I
      get better branching and merging (to be seen).
      
      For this, I need to the branches directory to have the normal form:
      every directory in omake-branches is a branch.  I apologize for
      this renaming, but it seems easier than hacking git-svn for our
      special case.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-04 08:32:37 -0800 (Sun, 04 Mar 2007)
Revision: 10154
Log message:

      Two cases were missed in the branch rename.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-04 12:15:36 -0800 (Sun, 04 Mar 2007)
Revision: 10157
Log message:

      Copy the branches omake-branches/closed to omake-closed-branches.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-05 11:10:24 -0800 (Mon, 05 Mar 2007)
Revision: 10159
Log message:

      Added the FASE paper to the OMake home page.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-05 15:35:26 -0800 (Mon, 05 Mar 2007)
Revision: 10163
Log message:

      I am creating a "fake" branch off the revision 9407 of 0.9.8.x. This can be
      used to "subtract" some of the patches from the 098-to-099 patch.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-05 15:37:45 -0800 (Mon, 05 Mar 2007)
Revision: 10164
Log message:

      Backported (for the diff tracking purposes only) Jason's hash-cons patch.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-05 20:12:38 -0800 (Mon, 05 Mar 2007)
Revision: 10170
Log message:

      Copying the root of the jumbo features.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-05 20:22:39 -0800 (Mon, 05 Mar 2007)
Revision: 10172
Log message:

      Move Aleksey's hashnode port into the hashnode jumbo branch.
      
      Also, remove 0.9.9.port for now.  We'll port at the very end.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-05 20:34:45 -0800 (Mon, 05 Mar 2007)
Revision: 10173
Log message:

      Added the "var1" jumbo branch.
      
      This is the base port, no changes yet.
      
      var1:
         Convert the source code in 0.9.8 so that it matches the
         source structure in 0.9.9.
      
         There should be no real semantical changes to the OMake
         language.  This is a code change only.
      
         Major architectural changes:
            - The Scope* is folded into the variable, it is
              no longer a separate field.
              
              The right way to think is that there are _not_
              3 environments, there are 3 kinds of variables
              and a single environment.
       
              There are 4 classes of variables (the "any"
              replaces the 0.9.8 default variable).
                  VarPrivate: private, statically scoped
                  VarThis: current object, dynamically scoped
                  VarVirtual: global, dynamically scoped
                  VarAny: the usual default 
      
            - Use venv_find_var_virtual to get a global var.
           
            - Omake_ir_ast now structurally looks like 0.9.9.
      
         There is an exception:
            - export variables are computed at compile time
              in Omake_ir_ast, instead of at runtime.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-05 20:46:55 -0800 (Mon, 05 Mar 2007)
Revision: 10177
Log message:

      Redo the "hashnode" branch.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-05 20:51:02 -0800 (Mon, 05 Mar 2007)
Revision: 10178
Log message:

      Re-add the "var1" branch.
      
      I'm making some small changes to 0.9.9 too.
      
      Comments specific to the backport are marked with ZZZ.
      
      This branch is completely untested currently.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-06 08:28:51 -0800 (Tue, 06 Mar 2007)
Revision: 10180
Log message:

      May be better to track the trunk and propagate changes.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-06 13:12:12 -0800 (Tue, 06 Mar 2007)
Revision: 10183
Log message:

      0.9.8 assumes that the forced mode *always* takes precendence.
      Consider.
      
         public. =
            Foo. =
      
         protected. =
            Foo. +=
               ...
      
      The resulting Foo is protected.  This reverts to that behavior,
      with a slight change.
      
      In old 0.9.8 the += expands to
      
            protected.Foo. =
                extends $(protected.Foo)
                ...
      
      Now, we generate this code.
      
            protected.Foo. =
                extends $(public.Foo)
                ...
      
      
      I'm not claiming that either of these really makes much sense.
      The intent is to preserve 0.9.8 semantics.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-06 13:13:41 -0800 (Tue, 06 Mar 2007)
Revision: 10184
Log message:

      The "man" directory should be ignored I assume.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-06 16:35:30 -0800 (Tue, 06 Mar 2007)
Revision: 10187
Log message:

      Removed all the sub-environment functions like
         venv_find_var_virtual
      etc.  No shortcuts, you have to construct an entire
      variable before asking for its value.  Standard variables
      are in omake_var.ml
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-07 10:11:39 -0800 (Wed, 07 Mar 2007)
Revision: 10190
Log message:

      Added keyword arguments feature.
      See the DEPENDENCIES file for details.
      
      Note that I am using git to manage these branches.
      The subversion stuff is for safekeeping, and making
      the features available for review.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-07 11:50:51 -0800 (Wed, 07 Mar 2007)
Revision: 10193
Log message:

      Note: my summary of the keyword-argument model is incorrect.
      I'll fix the write-up when I get a chance.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-08 11:33:50 -0800 (Thu, 08 Mar 2007)
Revision: 10196
Log message:

      This is a simplified keyword-argument model.
      See DEPENDENCIES for details.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-08 15:31:57 -0800 (Thu, 08 Mar 2007)
Revision: 10197
Log message:

      Changed the HRL bullet a bit.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-08 15:32:37 -0800 (Thu, 08 Mar 2007)
Revision: 10198
Log message:

      Oops, made a typo.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-08 15:41:42 -0800 (Thu, 08 Mar 2007)
Revision: 10199
Log message:

      Rephrased things a bit again...
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-08 16:24:04 -0800 (Thu, 08 Mar 2007)
Revision: 10200
Log message:

      [Bug 641] Added a helper function Omake_env.venv_find_object_or_empty. Patch
     by Erick Tryzelaar (erickt at dslextreme.com).

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-09 11:10:44 -0800 (Fri, 09 Mar 2007)
Revision: 10202
Log message:

      [Bug 641] Added getpwnam, getpwuid, getpwents, getgrnam, and getgrgid built-in
     functions. Based on a patch by Erick Tryzelaar (erickt at dslextreme.com).

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-12 09:47:01 -0700 (Mon, 12 Mar 2007)
Revision: 10204
Log message:

      Removed the AC_* style functions, replacing them with better named ones. I've
     added a short documentation section outlining the correspondences between the
     some of the AC_* functions in autoconf and the functions we provide, but I do
     not think it would be worth it to actually implement the AC_* functions.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-12 09:56:14 -0700 (Mon, 12 Mar 2007)
Revision: 10205
Log message:

      Regenerated the documentation (required a few minor fixes in a few places).

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-12 11:00:50 -0700 (Mon, 12 Mar 2007)
Revision: 10206
Log message:

      - (Bug 640) use /Fo: and /Fe: in C compiling and linking on Windows, as appropriate. Not fully tested yet.
     
     - Updated the Makefile.nt to use omake-boot, not src\main\omake for building src\main\osh.exe. This is needed because osh.exe may now be a hardlink to omake.exe, and updating it may be impossible when src\main\omake.exe is running...

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-12 11:55:32 -0700 (Mon, 12 Mar 2007)
Revision: 10207
Log message:

      [Bug 640] My previous commit was not quite correct. This one I've tested some more and it appears to work.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-12 11:57:42 -0700 (Mon, 12 Mar 2007)
Revision: 10208
Log message:

      Regenerated the documentation.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-12 12:47:59 -0700 (Mon, 12 Mar 2007)
Revision: 10209
Log message:

      Filling in the changelog in preparation for 0.9.8.1

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-13 12:10:27 -0700 (Tue, 13 Mar 2007)
Revision: 10212
Log message:

      Added tgetstr, xterm_escape_begin, xterm_escape_end built-in functions.
     
     I can now have the following in my ~/.oshrc:
     
     hostname = $(shell hostname)
     match $(hostname)
     case $'^\([^.]*\)[.]'
        hostname = $1
        export
     
     prompt =
        basic_prompt = $"[osh] $(hostname):"$`(homename .)
        bold_prompt = $(tgetstr bold)$(basic_prompt)$"> "$(tgetstr sgr0)
        if $(xterm_escape_begin)
           value $(xterm_escape_begin)$(basic_prompt)$(xterm_escape_end)$(bold_prompt)
        else
           value $(bold_prompt)

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-13 14:35:02 -0700 (Tue, 13 Mar 2007)
Revision: 10213
Log message:

      Partially backported the variable part of 0.9.9.
     See the DEPENDENCIES file for details.
     
     var3 builds, but is untested.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-13 16:19:01 -0700 (Tue, 13 Mar 2007)
Revision: 10215
Log message:

      Typo fix: EXT_OBJ already contains a ., so I should not be using ".$(EXT_IBJ)"

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-13 16:31:58 -0700 (Tue, 13 Mar 2007)
Revision: 10216
Log message:

      Remove ValFunValue (removed at "keyword" branch).

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-14 09:57:36 -0700 (Wed, 14 Mar 2007)
Revision: 10217
Log message:

      Use "-" instead of "_" in xterm-escape-begin and xterm-escape-end function
     names; implemented xterm-escape wrapper (that would return an empty string 
     TERM does not have an "xterm-like" value).

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-14 12:21:02 -0700 (Wed, 14 Mar 2007)
Revision: 10218
Log message:

      Added support for marking the "invisible" escapes in the shell prompt. The
     "invisibility" markers are used both:
      - to tell the readline library that these parts should not be included in
        "line length after the prompt" calculations, and
      - in our own MAX_PROMPT_LENGTH handling, where we cut off the prompts that
        are two long.
     
     When the prompt is too long, add "..." at the end (making sure that only the
     "visible" parts of the prompt are affected).
     
     All the above changes are Unix-only.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-14 12:36:07 -0700 (Wed, 14 Mar 2007)
Revision: 10219
Log message:

      Fixed a bug in common prefix computation that caused the common prefix to
     always be equal to the string already entered (so a single tab never did
     anything).

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-14 14:39:35 -0700 (Wed, 14 Mar 2007)
Revision: 10220
Log message:

      Merged Aleksey's changes from the 0.9.8.x trunk.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-15 13:15:37 -0700 (Thu, 15 Mar 2007)
Revision: 10221
Log message:

      Fixed the issue with the PDFLATEXFALGS variable never being used. Now if
     pdflatex is used, PDFLATEXFALGS would be used in place of LATEXFLAGS. For
     backwards-compatibility, the default value of PDFLATEXFALGS is $`(LATEXFLAGS).
     
     Thanks to Luis O'Shea (loshea at gmail.com) for reporting this!

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-15 13:48:24 -0700 (Thu, 15 Mar 2007)
Revision: 10222
Log message:

      Regenerated the documentation.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-15 14:01:40 -0700 (Thu, 15 Mar 2007)
Revision: 10223
Log message:

      Bumping the version number to 0.9.8.1 (release 1) in preparation for the
     release (which I am planning for tomorrow, to have time to build all the
     binaries before announcing the release).

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-15 14:18:07 -0700 (Thu, 15 Mar 2007)
Revision: 10225
Log message:

      Disable some of the warnings on Win32

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-15 14:55:49 -0700 (Thu, 15 Mar 2007)
Revision: 10226
Log message:

      Windows installation updates:
     - Updated the installer file for version 0.9.8.1
     - The latest version of the installer uses the relative paths for the source files, not the absolute ones like it used to, so I've updated it to take files from ../omake-install
     - Do not clobber .config from Makefile.nt, if it already exists!

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-15 15:18:29 -0700 (Thu, 15 Mar 2007)
Revision: 10227
Log message:

      On machines that are very fast, boot/Makefile.dep would not be created
     correctly ("Makefile.dep: $(ALLFILES)" dependency would not mean much if
     $(ALLFILES) were created within 1 sec after the Makefile.dep was touched and
     the filesystem only keeps the time stamps with a 1-sec precision).
     
     Added a "sleep 1" to address this problem on Unix. It may remain on Windows,
     "fortunately", on Windows nmake is way too slow to his this.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-15 15:21:27 -0700 (Thu, 15 Mar 2007)
Revision: 10228
Log message:

      Updated the BuildRequires

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-15 15:21:59 -0700 (Thu, 15 Mar 2007)
Revision: 10229
Log message:

      Updated the BuildRequires

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-15 15:32:12 -0700 (Thu, 15 Mar 2007)
Revision: 10230
Log message:

      rpmbuild wants to be able to strip the binaries, but we install them
     read-only, so the build fails.
     
     Trying to work around this problem.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-15 15:34:32 -0700 (Thu, 15 Mar 2007)
Revision: 10231
Log message:

      Previous commit was not quite right, trying again.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-15 16:55:57 -0700 (Thu, 15 Mar 2007)
Revision: 10232
Log message:

      Tagging the release.

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-15 19:08:39 -0700 (Thu, 15 Mar 2007)
Revision: 10233
Log message:

      Added autoloading, see bug #651.
     
     The problem has to do with cyclic open commands.  For example, C.om opens Configure.om,
     and vice-versa.  This fix performs eager linking, but files are loaded only when
     their values are dereferenced.  [Also: need to force all opens before building a project.]
     
     Also added a "select" command.  This is intended to allow files that are configured for
     multiple versions.
     
     The select is based on Python's "future" directive, where parsing is conditional on
     a specific Python feature.
     
     Configure.om gives the following example.
     
        select feature
        case strictscope
            open build/C
        default
            open build/Common
     
     Unlike other constructs, the indentation does not specify a new scope.  This is more
     like an #ifdef.
     
     Also, I'm willing to consider other syntaxes and/or concepts.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-16 16:49:53 -0700 (Fri, 16 Mar 2007)
Revision: 10234
Log message:

      MANDIR is not loger relevant when building/installing OMake.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-16 16:55:58 -0700 (Fri, 16 Mar 2007)
Revision: 10235
Log message:

      Made it easier to build a Windows package with the correct default library directory.

Changes by: Nathaniel Gray (n8gray at caltech.edu)
Date: 2007-03-17 00:25:41 -0700 (Sat, 17 Mar 2007)
Revision: 10236
Log message:

      Update to the files for building the 0.9.8.1-1 release.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-19 10:15:46 -0700 (Mon, 19 Mar 2007)
Revision: 10240
Log message:

      Bumping the version to 0.9.8.1+

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-19 10:19:17 -0700 (Mon, 19 Mar 2007)
Revision: 10241
Log message:

      Committing Jason's "cons-hash" patch (with couple of minor changes).
     
     I have not looked at it much (yet) - hope to do it over the next few days. I
     will probably want to do some profiling on it too - for some reason, on my
     machine a "no-op" run of OMake on itself (after making a few iterations to
     make sure everything is in the FS cache) takes about 1s withous this patch,
     but 1.5s with it...

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-20 00:18:06 -0700 (Tue, 20 Mar 2007)
Revision: 10244
Log message:

      Intermediate commit--I messed up my git repos a little, so
     I think I should save to svn more often.
     
     Several small changes:
        - Autoloading should work.
        - If a file is to be used with "open", it must define
          it variables.  No more "X = foo".  If X is local,
          the first definition must be qualified "public.X = foo".

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-20 09:49:04 -0700 (Tue, 20 Mar 2007)
Revision: 10245
Log message:

      Auto-updated omake.spec with the new version number.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-20 10:24:32 -0700 (Tue, 20 Mar 2007)
Revision: 10247
Log message:

      Moved the profiling option to the end of the .config

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-20 10:45:59 -0700 (Tue, 20 Mar 2007)
Revision: 10248
Log message:

      - Dropped a custom Omake_command_digest implementation and changed it to using
       Lm_hash.HashDigest (which is almost identical).
     
     - Added the omake_command_digest file to the list of magic files for the cache
       and marked the list of hash coded (type Omake_command_digest.code) with the
       "MAGICBEGIN"/"MAGICEND"

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-20 11:08:24 -0700 (Tue, 20 Mar 2007)
Revision: 10249
Log message:

      Removed an unused function.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-20 11:38:44 -0700 (Tue, 20 Mar 2007)
Revision: 10250
Log message:

      Minor inlining.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-20 12:25:52 -0700 (Tue, 20 Mar 2007)
Revision: 10252
Log message:

      Small optimizations.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-20 15:12:18 -0700 (Tue, 20 Mar 2007)
Revision: 10254
Log message:

      Optimized the create_node_or_phony - instead of maintaining a set of phony
     nodes (which meant that we had to try interning every time before we could
     test for membership), maintain a phony set in terms of an elt_node set.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-20 15:33:42 -0700 (Tue, 20 Mar 2007)
Revision: 10255
Log message:

      Minor optimization.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-20 16:35:04 -0700 (Tue, 20 Mar 2007)
Revision: 10256
Log message:

      - Small optimization in Omake_node
     - Added support for native profiling with -compact.

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-20 16:55:30 -0700 (Tue, 20 Mar 2007)
Revision: 10257
Log message:

      Linking policies: open is strict, include is lazy.
     
     This program will fail unless Foo defines X.
     
        open Foo
        X = 1
     
     That is, for open, all variables must be defined explicitly.
     
     This program will produce a warning unless Foo defines X.
     
        include Foo
        X = 1

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-20 17:06:04 -0700 (Tue, 20 Mar 2007)
Revision: 10258
Log message:

      Further optimized the path code, finally bringing the performance close to
     what it was before the hash-cons patch was applied.
     
     The problem was that when appending an extra path (which might contain ".."
     somehere inside) to a directory, it would fully reconstruct the directory and
     then construct it from scratch again. In presence of cons-hashing this is
     really expensive.
     
     The obvious solution is not to deconstruct and only pop one level every time
     the ".." is encountered.

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-20 18:26:57 -0700 (Tue, 20 Mar 2007)
Revision: 10260
Log message:

      Autoloading doesn't require an actual value.
     Whenever a value faults, load the object, then try again.

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-20 20:17:42 -0700 (Tue, 20 Mar 2007)
Revision: 10261
Log message:

      Fix -Wdeclare warnings.

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-21 13:15:57 -0700 (Wed, 21 Mar 2007)
Revision: 10264
Log message:

      Fixed a link warning.

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-21 19:42:15 -0700 (Wed, 21 Mar 2007)
Revision: 10265
Log message:

      var3 is nearly done.  Self-builds work.  I should probably try to make it
     more backward compatible.
     
     This commit changes "open" to include rules in the opened file (and
     all files it opens, etc.)
     
     In 0.9.8, "open" is about the same as "include".  Basically, "open"
     means "include once".
     
     In 0.9.9 open is different because opened files are evaluated in a
     pristine environment that contains only Pervasives (see bug #586).
     BTW, it may be fairly easy to specify a Site.om that specifies this
     pristine environment.  For example Site.om might set the PATH to
     some default value.
     
     This means that rules in opened files must be merged when the file
     is opened.  Autoloading broke it because the rule would get added
     at some delayed time.
     
     This commit changes the behavior.
        - "open" is eager, loading values+rules recursively.
        - "autoload" is lazy, but you don't get any rules from the autoloaded file.
     
     Configure.om has this code:
     
         open build/Common
     
         select feature
         case strictscope
             autoload build/C
     
     This has the usual bogus effect that if you only open Configure, then
     configuration is useless (this is the same in 0.9.8).  However, the
     cyclic dependency is gone, and the build works as it should.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-22 09:27:35 -0700 (Thu, 22 Mar 2007)
Revision: 10266
Log message:

      Oops, had a 2006 instead of 2007 for the 0.9.8.1 release!

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2007-03-22 14:11:44 -0700 (Thu, 22 Mar 2007)
Revision: 10269
Log message:

      Fixed a bug with MenhirMulti not adding the extensions properly.

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2007-03-22 17:29:37 -0700 (Thu, 22 Mar 2007)
Revision: 10271
Log message:

      Remove all Node.t from the IR, fixing the long-standing bug
     where absolute names are saved in the .omc file.
     
     See DEPENDENCIES for details.