Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-01 10:50:41 -0700 (Sun, 01 May 2005)
Revision: 924
Log message:

      Remove the soft-recursive .SCANNER dependencies I just added.
      Perparing to investigate using :value: dependencies instead.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-01 14:16:59 -0700 (Sun, 01 May 2005)
Revision: 925
Log message:

      Added more precise .SCANNER dependencies.
      The rules look like the following, where $& represents the results of the previous scan.
      
      .SCANNER: %.cmi: :value: $(search-path-exists $(OCAMLINCLUDES), $(replacesuffixes .cmi .cmo .cmx, .mli .ml .ml, $&))
      .SCANNER: %.cmo %.cmx: :value: $(search-path-exists $(OCAMLINCLUDES), $(replacesuffixes .cmi .cmo .cmx, .mli .ml .ml, $&))
      .SCANNER: %$(EXT_OBJ): :value: $(digest-path-exists $(INCLUDES), $&)
      
      The intent here is that omake uses the search path to find the dependent files,
      and the scanner is considered out-of-date if the location of a file moves.
      For OCaml files, it is just the location of the .ml and .mli files.
      For C files, it is the location *and* the contents of the files.
      
      There is a moderate cost to this (it adds about 1-2sec to a MetaPRL compile).
      Also, scanners will run more often.  In particular it will take 2 runs of
      omake to reach a scanner fixpoint because the value dependency is based
      on the results of the previous omake run.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-01 15:09:54 -0700 (Sun, 01 May 2005)
Revision: 927
Log message:

      Fixed an issue where .omc files were getting compiled twice if they were
      included twice from different directories.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-01 15:35:39 -0700 (Sun, 01 May 2005)
Revision: 928
Log message:

      Bumping the version number (as I prepare to be using the digest function in
      a MetaPRL OMakefile)
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-01 16:01:50 -0700 (Sun, 01 May 2005)
Revision: 929
Log message:

      Added .omakedb locking.  This probably won't work on Windows,
      so I'll try that next.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-01 17:04:46 -0700 (Sun, 01 May 2005)
Revision: 930
Log message:

      Implemented file locking for Win32.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-01 18:56:00 -0700 (Sun, 01 May 2005)
Revision: 931
Log message:

      Added file locking for .omc etc. files.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-01 19:09:37 -0700 (Sun, 01 May 2005)
Revision: 932
Log message:

      Unix.ftruncate doesn't work on Win32.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-01 20:04:41 -0700 (Sun, 01 May 2005)
Revision: 933
Log message:

      Recompute the .SCANNER digest if the scanner dependencies changed
      and the rule has any value dependencies.  This helps reach the
      .SCANNER fixpoint earlier.
      
      Here is the comment in omake_build.ml:
      
       * XXX: HACK: Recompute the command digest if the scanner dependencies
       * have changed.
       *
       * This is probably a reasonable thing to do, but it means that the
       * rule text may be computed twice for .SCANNER rules.
       *
       * This could be wrong in two cases:
       *    1. If the .SCANNER body performs a side-effect while computing
       *       the rule text.
       *    2. If the .SCANNER body depends non-trivially on the scanner
       *       dependencies $&.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-01 20:20:00 -0700 (Sun, 01 May 2005)
Revision: 934
Log message:

      Added a CHANGELOG prepareing for release.  Aleksey should take a look
      and add things at the end that we need to do before 0.9.5.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-02 17:24:37 -0700 (Mon, 02 May 2005)
Revision: 935
Log message:

      In 1-argument version of "split", split along any white space boundaries
      (using Lm_string_util.tokens_std).
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-02 18:01:25 -0700 (Mon, 02 May 2005)
Revision: 936
Log message:

      Omake_value.key_of_value was erroneously wrapping the key
      in an array.
      
      In Omake_eval.strings_of_value, the ValSequence should not
      be treated as a quoted block.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-02 18:50:51 -0700 (Mon, 02 May 2005)
Revision: 937
Log message:

      Added a TETEX2_ENABLED variable that controls whether to use the advanced
      latex options that were added in TeTeX v.2. The default value for this
      variable is determined by configure.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-02 18:57:51 -0700 (Mon, 02 May 2005)
Revision: 938
Log message:

      Fixed a typo.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-02 19:05:04 -0700 (Mon, 02 May 2005)
Revision: 939
Log message:

      Fixing a documentation typo
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-02 19:12:43 -0700 (Mon, 02 May 2005)
Revision: 940
Log message:

      Merged Changes.txt, CHANGELOG into CHANGELOG.txt
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-02 19:35:54 -0700 (Mon, 02 May 2005)
Revision: 941
Log message:

      Added a message to the lock file.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-02 19:46:49 -0700 (Mon, 02 May 2005)
Revision: 942
Log message:

      Changes.txt -> CHANGELOG.txt
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-02 20:39:50 -0700 (Mon, 02 May 2005)
Revision: 943
Log message:

      Add finer control over cached status.  This addresses #403.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-02 20:40:29 -0700 (Mon, 02 May 2005)
Revision: 944
Log message:

      Ignore the .omakedb.lock file.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-02 20:58:04 -0700 (Mon, 02 May 2005)
Revision: 945
Log message:

      ignore omake.fls
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-03 14:39:22 -0700 (Tue, 03 May 2005)
Revision: 946
Log message:

      Added a Target object to get all the information for a specified
      target.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-03 17:41:50 -0700 (Tue, 03 May 2005)
Revision: 947
Log message:

      - (Bug 400) Updated the way the "totals" in the "done" message are computed.
        For scans and rule executions, only include the rules with non-empty
        commands in the totals. For digests, exclude non-existing files and
        directories from the total count.
      
        Building omake from scratch:
           before: 191/550 scans, 210/550 rules, 565/1961 digests
           now   : 191/191 scans, 210/341 rules, 565/780 digests
      
        Building MetaPRL from scratch:
           before: 1929/4560 scans, 1685/4560 rules, 4707/14410 digests
           now   : 1928/2245 scans, 1685/3088 rules, 4708/6168 digests
      
        There is still some unaccounted stuff here, will have to investigate.
      
      - Merged a bit of duplicated code in omake_build.ml and omake_cache.ml
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-03 21:01:45 -0700 (Tue, 03 May 2005)
Revision: 948
Log message:

      (Bug 463) added a "where" function similar to tcsh's one. Also added
      Shell.where and Shell.which.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-03 21:06:14 -0700 (Tue, 03 May 2005)
Revision: 949
Log message:

      Added Windows versions of some of the generated files.
      

Changes by: ( at unknown.email)
Date: 2005-05-03 21:06:14 -0700 (Tue, 03 May 2005)
Revision: 950
Log message:

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

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-04 10:55:45 -0700 (Wed, 04 May 2005)
Revision: 951
Log message:

      Looking into new scanner policy, but its easier to work on Unix.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-04 11:38:51 -0700 (Wed, 04 May 2005)
Revision: 952
Log message:

      Augmented environment with explicit scanners.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-04 13:21:47 -0700 (Wed, 04 May 2005)
Revision: 953
Log message:

      (Bug 400) Fixed the issue with progress bar not being at 100% when the beuild
      completes. This was caused by the fact that the optional targets were created
      with "Succeeded" status right away and they were counted towards the total,
      but not towards the progress. I've added a separate counted for these
      commands, and subtract it from totals.
      
      Also, when the build fails, I added a message that tells how far we've gotten.
      

Changes by: Nathaniel Gray (n8gray at cs.caltech.edu)
Date: 2005-05-04 15:25:00 -0700 (Wed, 04 May 2005)
Revision: 954
Log message:

      Modified to support LDFLAGS.  Also should work with BSD readline now
      (hopefully).
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-04 15:29:57 -0700 (Wed, 04 May 2005)
Revision: 955
Log message:

      Changed the recursive scanner dependencies to apply only for .o compiled from
      .c, instead of to arbitrary .o.
      
      Addede recursive scanner dependencies to the .SCANNER rule for LaTeX.
      

Changes by: Nathaniel Gray (n8gray at cs.caltech.edu)
Date: 2005-05-04 18:04:33 -0700 (Wed, 04 May 2005)
Revision: 956
Log message:

      Search for \bibdata instead of \citation to trigger bibtex.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-04 18:48:08 -0700 (Wed, 04 May 2005)
Revision: 957
Log message:

      Rearranged the sections on rules and implicit rules a bit. Mentioned scoping
      of implicit rules. Minor spellchecking.
      
      Jason, you might want to review this.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-05 12:08:22 -0700 (Thu, 05 May 2005)
Revision: 958
Log message:

      - (Bug 470) Exclude the .bbl from the generated depepndencies list for LaTeX.
        We need to do this since we handle bibtex internally in the LaTeX build rule.
      
      - Also, changed the "makeindex" call to be a part of the latex build rule as
        well.
      
      - Updated the pdflatex rule to match the recent changes in the latex rule.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-05 14:20:03 -0700 (Thu, 05 May 2005)
Revision: 959
Log message:

      One of the magic numbers have changed since the "early days" of 0.9.4.9,
      increasing the version number.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-05 14:35:32 -0700 (Thu, 05 May 2005)
Revision: 960
Log message:

      Added an OMAKE_VERSION variable.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-05 16:58:43 -0700 (Thu, 05 May 2005)
Revision: 961
Log message:

      Added $(target-is-proper ...) and $(filter-proper-targets ...).  A
      target is "proper" if it can be built by the current project.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-05 17:07:45 -0700 (Thu, 05 May 2005)
Revision: 962
Log message:

      Added a longer comment after $(filter-proper-targets ...)
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-05 17:58:59 -0700 (Thu, 05 May 2005)
Revision: 963
Log message:

      Added $(dependencies-all ...) for recursive dependencies,
      and $(dependencies-proper ...) for dependencies that are not leaves.
      
      Note that the following is probably not what you want, because it
      requires that the project be built before it can be cleaned.
      
         clean: omake.opt
            rm $(dependencies-proper omake.opt)
      
      We could get this right with a rule something like the following.
      
         clean: :scanner: omake.opt
            rm $(dependencies-proper omake.opt)
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-05 17:59:43 -0700 (Thu, 05 May 2005)
Revision: 964
Log message:

      Acidentally committed a test target.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-05 18:40:40 -0700 (Thu, 05 May 2005)
Revision: 965
Log message:

      On Win32 use the SHGetFolderPath function to get the location
      of the HOME directory.
      
      Instead of using Unix.access to check whether a file is
      writable or not, just open the file in write-mode.  This
      is certainly more expensive, but is also much more reliable.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-06 15:33:26 -0700 (Fri, 06 May 2005)
Revision: 966
Log message:

      Allow using $(if test, exp) in place of $(if test, exp, $(EMPTY))
      

Changes by: ( at unknown.email)
Date: 2005-05-06 15:33:26 -0700 (Fri, 06 May 2005)
Revision: 967
Log message:

      This commit was manufactured by cvs2svn to create branch 'new_scanner2'.

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-07 13:03:19 -0700 (Sat, 07 May 2005)
Revision: 968
Log message:

      Folded together normal and scanner rules in the env.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-07 19:26:57 -0700 (Sat, 07 May 2005)
Revision: 969
Log message:

      Basic separation is finished.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-08 12:31:39 -0700 (Sun, 08 May 2005)
Revision: 970
Log message:

      Handle :effects: correctly.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-08 13:09:18 -0700 (Sun, 08 May 2005)
Revision: 971
Log message:

      (Bug 473) in $&-based SCANNER dependency generation for OCaml, remove files
      other than .ml and .mli from the returned list. In MetaPRL case, this should
      protect from having the SCANNER depend on things like .cmiz.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-08 15:21:43 -0700 (Sun, 08 May 2005)
Revision: 972
Log message:

      Changed :effects: as follows.
         - The targets after an :effects: are just for concurrency control.
         - If a rule can build a target, you should list the target on
           the left-hand-side.
      
      This changes the rules for ML files, which now look something like
      this:
      
         %.cmx: %.ml
            ...
            %.cmx %.o: %.ml %.cmi
               ...
      
      For :effects: we should probably add some control over the ordering.
      For example, suppose we have
      
         foo: :effects: boo
            ...
         boo: moo
            ...
      
      Then we should probably add a dependency:
      
         boo: foo
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-08 15:43:11 -0700 (Sun, 08 May 2005)
Revision: 973
Log message:

      Added explicit :scanner: dependencies for ML files.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-08 16:17:02 -0700 (Sun, 08 May 2005)
Revision: 974
Log message:

      Removed a useless field.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-08 16:18:57 -0700 (Sun, 08 May 2005)
Revision: 975
Log message:

      Increased the version number.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-09 15:47:38 -0700 (Mon, 09 May 2005)
Revision: 976
Log message:

      - Added a function "set-diff", bumped the version number
      
      - (Bug 470) When generating the TeX dependencies fron the .fls "access log",
        do not include the "OUTPUT" files in the dependency list.
      
      - (Bug 473) My previous commit had a typo, fixing it.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-09 15:57:55 -0700 (Mon, 09 May 2005)
Revision: 977
Log message:

      Set the TEXINPUTS/BIBINPUTS environment variables using "setenv" instead of
      giving it explicitly on every latex/bibtex/pdflatex/etc command invocation.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-09 18:10:07 -0700 (Mon, 09 May 2005)
Revision: 978
Log message:

      - The progress bar broke again, fixing it in a somewhat more maintainable way
        (update the counter in reclassify_command instead of manually chaising all
        the calling reclassify_command points that may need a counter
        increment/decrement)
      
      - Bumping the version number (to keep it higher than the trunk one).
      
      - Merged a small scanner-related change from the trunk.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-10 10:44:14 -0700 (Tue, 10 May 2005)
Revision: 979
Log message:

      Scanner was being run before dependencies were built.  This fixes the
      .fls problem.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-10 14:24:34 -0700 (Tue, 10 May 2005)
Revision: 980
Log message:

      Added a "export rules" hack.  This is to get around the problem in
      MetaPRL where scoping is not handled correctly (support/tactics,
      support/shell, theories/itt).
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-10 17:05:25 -0700 (Tue, 10 May 2005)
Revision: 981
Log message:

      Added the $(project-directories) function to get all the directories
      in the project.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-10 17:09:27 -0700 (Tue, 10 May 2005)
Revision: 983
Log message:

      Forgot to remove the functions that were moved to Omake_builtin_target.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-10 17:30:33 -0700 (Tue, 10 May 2005)
Revision: 984
Log message:

      Minor simplification to fold together some command states as Aleksey
      asked.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-11 15:21:56 -0700 (Wed, 11 May 2005)
Revision: 985
Log message:

      Renamed to find-in-path, find-in-path-optional, etc.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-12 14:39:48 -0700 (Thu, 12 May 2005)
Revision: 987
Log message:

      Made configure's --enable-bytecode option do something at least
      semi-reasonable.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-12 16:06:43 -0700 (Thu, 12 May 2005)
Revision: 988
Log message:

      This is a significant change to .SCANNER rules, where .SCANNER rules
      are treated much more like normal rules.
      
      Externally, a .SCANNER rule has the usual rule form:
      
         .SCANNER: target: dependencies...
            ...scanner commands...
      
      In this commit, the scanner target is now decoupled from the
      build target, allowing a scanner result to be used for multiple
      build targets.  For example, ocamldep produces dependencies
      for .cmo and .cmx files simultaneously.  They can share
      the scanner rule by specifying an explicit :scanner: dependency.
      
          .SCANNER: scan-ocaml-%.ml: %.ml
              ocamldep $<
      
          %.cmo: %.ml :scanner: scan-ocaml-%.ml
              $(OCAMLC) ...
      
          %.cmx %.o: %.ml :scanner: scan-ocaml-%.ml
              $(OCAMLOPT) ...
      
      The current convention is that scanner targets should be named
      scan-<language>-<source-file>.
      
         -- If a rule has multiple :scanner: dependencies, the actual
            dependencies will be the union of the scanner results.
      
         -- The .SCANNER targets use a different namespace than
            normal targets, so it is valid to have overlapping rules.
      
                .SCANNER: foo:
                    echo "foo: boo"
      
                foo: :scanner: foo
                    ...
      
         -- For backwards compatibility, if a rule has no :scanner:
            dependencies, then omake will try to find a scanner with
            the same name as the target.  So in the example above,
            the :scanner: foo is actually unnecessary.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-12 16:23:25 -0700 (Thu, 12 May 2005)
Revision: 990
Log message:

      Fixed the problem with things like
         echo ""
      being a syntax error.  The issue was that the shell was using multi-quoting,
      so a string like ""Hello "xxx" world"" would be a single string with a
      quoted "xxx" in it.  I've removed that feature.
      
      Added back the $(replacesufixes ...) in OCamlScannerDeps.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-12 16:34:25 -0700 (Thu, 12 May 2005)
Revision: 991
Log message:

      Updating the changelog.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-12 16:51:12 -0700 (Thu, 12 May 2005)
Revision: 992
Log message:

      Handle exceptions more aggressively in Omake_exec_local, and
      print the exception on the error channel.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-12 17:02:19 -0700 (Thu, 12 May 2005)
Revision: 993
Log message:

      For efficiency, in OCamlScannerDeps apply the "set" function to the result of
      "replacesuffixes".
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-12 17:13:47 -0700 (Thu, 12 May 2005)
Revision: 994
Log message:

      The variable $& now refers to *all* the dependencies that were
      returned by the previous scan.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-12 17:46:43 -0700 (Thu, 12 May 2005)
Revision: 995
Log message:

      Added a SCANNER_MODE variable (defaults to enabled).
      
      # \item[enabled] Allow the use of default \verb+.SCANNER+ rules.  Whenever a rule does
      #    not specify a \verb+:scanner:+ dependency explicity, try to find a
      #    \verb+.SCANNER+ with the same target name.
      # \item[disabled] Never use default \verb+.SCANNER+ rules.
      # \item[warning] Allow the use of default \verb+.SCANNER+ rules, but print a warning
      #    whenever one is selected.
      # \item[error] Do not allow the use of default \verb+.SCANNER+ rules.  If a rule
      #    does not specify a \verb+:scanner:+ dependency, and there is a default
      #    \verb+.SCANNER+ rule, the build will terminate abnormally.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-12 18:01:30 -0700 (Thu, 12 May 2005)
Revision: 996
Log message:

      Re-raise some exceptions when they occur during a process creation.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-12 18:21:14 -0700 (Thu, 12 May 2005)
Revision: 997
Log message:

      Fixed an oversight that caused the directory for <value> commands
      to be printed incorrectly.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-12 18:40:20 -0700 (Thu, 12 May 2005)
Revision: 998
Log message:

      Fixed a problem with caching the wrong set of dependencies for the scanner
      rules.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-12 18:40:47 -0700 (Thu, 12 May 2005)
Revision: 999
Log message:

      Removed the $(name).aux etc rules from the LaTeXDocument function
      (Aleksey should look at this).
      
      Fixed the .fls .SCANNER.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-12 18:56:20 -0700 (Thu, 12 May 2005)
Revision: 1000
Log message:

      SCANNER_MODE = error.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-12 18:59:13 -0700 (Thu, 12 May 2005)
Revision: 1001
Log message:

      Updating a comment.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-12 19:04:58 -0700 (Thu, 12 May 2005)
Revision: 1002
Log message:

      The Rule object no longer has a scanned-deps field.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-12 20:04:16 -0700 (Thu, 12 May 2005)
Revision: 1003
Log message:

      Determine the Tetex config at omake run time, for better binary portability.
      
      The current model defines a Tetex_config object, so the first time you
      run omake, you determine the config to be used in all subsequent runs.
      
      I haven't remove the configure variable until we see how well this
      works.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-12 22:31:50 -0700 (Thu, 12 May 2005)
Revision: 1004
Log message:

      Instead of a Tetex_config, I've added a more generic OMConfigure file. In
      particular, I moved the test for ocamlfind presence into OMConfigure.
      
      OTOH I restored the delayed status of all the TETEX2_ENABLED accesses - I want
      it to be possible to override the default value in individual projects.
      
      Finally, I removed the TETEX2_ENABLED test from configure.in
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-12 22:32:57 -0700 (Thu, 12 May 2005)
Revision: 1005
Log message:

      Updating the Windows version.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-12 22:45:54 -0700 (Thu, 12 May 2005)
Revision: 1006
Log message:

      - I changed OMakeroot.src back into a CVS file, instead of it being a
      configure-generated one.
      
      - Updated the copyright from "2003-2004" to "2003-2005" in a few places.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-12 23:11:06 -0700 (Thu, 12 May 2005)
Revision: 1007
Log message:

      Actually, I have to take this all back :-(. OMConfigure does _not_ seem to be
      a workable solution ATM. The problem is that the .omo file whould normally go
      into ~/.omake/cache which may be NFS-shared between many machines with
      different version of TeTeX and other similar programs. So, unless we decide to
      use something like ~/.omake/`hostname`-cache in place of ~/.omake/cache, this
      is not the way to go.
      

Changes by: ( at unknown.email)
Date: 2005-05-12 23:11:06 -0700 (Thu, 12 May 2005)
Revision: 1008
Log message:

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

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-12 23:14:05 -0700 (Thu, 12 May 2005)
Revision: 1009
Log message:

      Adding the OMConfigure solution back on a branch, to make it easier to
      resurrect it if we ever decide to do so.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-12 23:43:11 -0700 (Thu, 12 May 2005)
Revision: 1010
Log message:

      Fixed the scanner rule for LaTeX so that things still work (hopefully) whin
      TeTeX v 1.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-12 23:54:27 -0700 (Thu, 12 May 2005)
Revision: 1011
Log message:

      Documentation update.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-13 11:58:22 -0700 (Fri, 13 May 2005)
Revision: 1012
Log message:

      Added the hostname to .omo files and resurrected Tetex_config.
      
      Lets not squash all the configs into one huge file; it is less
      modular that way.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-13 12:18:18 -0700 (Fri, 13 May 2005)
Revision: 1013
Log message:

      Cache the object files, so that import/extends load the .omo only
      once even if used multiple times.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-13 13:19:36 -0700 (Fri, 13 May 2005)
Revision: 1015
Log message:

      Added a --config option to consider cached .om* files as out-of-date.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-13 14:56:17 -0700 (Fri, 13 May 2005)
Revision: 1016
Log message:

      Allow :effects: in .SCANNER rules.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-13 18:07:45 -0700 (Fri, 13 May 2005)
Revision: 1017
Log message:

      TETEX2_ENABLED requires both options to be present, not just one of them.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-13 18:09:25 -0700 (Fri, 13 May 2005)
Revision: 1018
Log message:

      Instead of "<dotomake>/cache", use "<dotomake>/<hostname>-cache".
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-13 18:31:45 -0700 (Fri, 13 May 2005)
Revision: 1019
Log message:

      Temporatily removing the new Tetex_config stuff (and the hostname-suffixing of
      the .omo files), until we figure come up with a more fundamental solution to
      the configure-once problem.
      
      I will create a "configure" branch shortly and I will re-add the files there.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-13 18:38:51 -0700 (Fri, 13 May 2005)
Revision: 1020
Log message:

      Renamed "--config" into "--flush-includes" for now. Also, reordered the
      "ignore some of the cached stuff" options, trying to make the order more
      meaningful. Finally, marked the "-u" option as being enabled by default.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-13 18:50:02 -0700 (Fri, 13 May 2005)
Revision: 1021
Log message:

      When executing an "import", if the object file is not found, then give a
      proper error message (as opposed to "Fatal error: exception Not_found").
      
      P.S. FOr some reason, the location information is incorrect in the error
      message.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-13 18:55:40 -0700 (Fri, 13 May 2005)
Revision: 1022
Log message:

      Slight change: use cache/<hostname>/ instead of <hostname>-cache/
      

Changes by: ( at unknown.email)
Date: 2005-05-13 18:55:40 -0700 (Fri, 13 May 2005)
Revision: 1023
Log message:

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

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-13 19:16:43 -0700 (Fri, 13 May 2005)
Revision: 1024
Log message:

      Re-adding the Tetex_config on a branch.
      
      Note:
       - This does not include the "hostname in the .omo" change (since it is quite
         likely that the new solution will not be omo-based)
       - Intead of Tetex_config + Tetex_config.src, this uses a single Tetex_config
         (an extra wrapper did not seem to serve any useful purpose).
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-16 23:50:16 -0700 (Mon, 16 May 2005)
Revision: 1025
Log message:

      Added functions:
       - CGeneratedFiles
       - LocalCGeneratedFiles
       - TeXGeneratedFiles
       - LocalTeXGeneratedFiles
      analogous to the corresponding C ones.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-17 01:35:41 -0700 (Tue, 17 May 2005)
Revision: 1026
Log message:

      Use 'stdout = $(stderr)' for output redirection instead of an awk-based hack.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-17 05:04:37 -0700 (Tue, 17 May 2005)
Revision: 1027
Log message:

      Added a configure check for FAMNoExists (which is a gamin-only extension to
      FAM API). Use FAMNoExists on opened FAM connections, when FAMNoExists API is
      present.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-17 05:10:16 -0700 (Tue, 17 May 2005)
Revision: 1028
Log message:

      config.h is a generated file
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-17 05:32:00 -0700 (Tue, 17 May 2005)
Revision: 1029
Log message:

      Bumping the version number
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-17 06:39:47 -0700 (Tue, 17 May 2005)
Revision: 1030
Log message:

      Omake_exec_notify.handle: do not go into Omake_notify.next_event unless
      Omake_notify.pending is true.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-17 07:22:05 -0700 (Tue, 17 May 2005)
Revision: 1031
Log message:

      Remove the .omakedb.lock on omake exit (via the at_exit call).
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-17 14:02:03 -0700 (Tue, 17 May 2005)
Revision: 1032
Log message:

      Bumping the version number (without good reason)
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-18 12:56:18 -0700 (Wed, 18 May 2005)
Revision: 1033
Log message:

      Fixed the deadlock error message - instea of "dependes on %a" we need to print
      "is a dependency of %a".
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-19 00:25:09 -0700 (Thu, 19 May 2005)
Revision: 1034
Log message:

      Be a bit more proactive in detecting notifications.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-19 00:48:35 -0700 (Thu, 19 May 2005)
Revision: 1035
Log message:

      Fixing a nasty typo (we had scanner-ocaml-%.cmo instead of scanner-ocaml-%.ml
      in one place!).
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-19 13:05:14 -0700 (Thu, 19 May 2005)
Revision: 1037
Log message:

      Do not try to kill a job that has already finished.
      This fixes bug #481.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-20 04:18:50 -0700 (Fri, 20 May 2005)
Revision: 1043
Log message:

      Do "start_or_build_commands" for static deps before doing it for scanner deps
      (this results in better error messages - if some of the deps are not
      buildable, an error message complaining about a real file is much nicer than
      an error message about a scanner target).
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-22 11:34:16 -0700 (Sun, 22 May 2005)
Revision: 1045
Log message:

      Memo entries need to be indexed by an explicit target.  This
      addresses bug #387.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-22 13:20:50 -0700 (Sun, 22 May 2005)
Revision: 1046
Log message:

      This is a new version of .omc and .omo files where the files can contain
      multiple versions.  In the .omc, only the source digest and IR version is
      significant.  In the .omo, the hostname is significant too.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-22 13:25:20 -0700 (Sun, 22 May 2005)
Revision: 1047
Log message:

      Put some bounds on the size of hostname strings.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-22 13:57:10 -0700 (Sun, 22 May 2005)
Revision: 1048
Log message:

      Do the right thing with latex .inx/.ind files. Addresses bug #102.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-22 14:40:55 -0700 (Sun, 22 May 2005)
Revision: 1050
Log message:

      Bumping the version number (as I prepare to upgrade all the PRL machines
      with the latest CVS version).
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-22 14:46:34 -0700 (Sun, 22 May 2005)
Revision: 1051
Log message:

      Better makeindex rule.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-22 14:49:37 -0700 (Sun, 22 May 2005)
Revision: 1052
Log message:

      Wrapping the "truncating the file" into an "if !debug_eval".
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-22 15:11:02 -0700 (Sun, 22 May 2005)
Revision: 1053
Log message:

      The "or" and "and" built-in functions need to be lazy (in order to enable
      short-circuit evaluation).
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-22 15:47:45 -0700 (Sun, 22 May 2005)
Revision: 1054
Log message:

      Updating...
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-22 17:12:18 -0700 (Sun, 22 May 2005)
Revision: 1055
Log message:

      For implicit .SCANNER targets, use the environment from the dependent
      rule.  This addresses bug #482.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-23 12:56:51 -0700 (Mon, 23 May 2005)
Revision: 1056
Log message:

      Fixed the problem with output on Win32.
         - Win32 uses threads instead of fork (Unix uses fork
           so it doesn't have this problem).
         - Threads are run with the usual pipe through another
           IO thread that prints the output.
         - So when a thread calls Pervasives.exit directly, omake
           may exit before the IO threads are finished (and
           so the output gets lost).
      
      The solution is to have the omake exit function raise an
      exception instead of calling Pervasives.exit.
      
      I'm a little worried about this scheme for two reasons.
         1. We need to make sure the ExitEception is handled
            properly in all the right exception handlers.
         2. In general, we shouldn't allow omake to exit until
            all threads have finished.  I believe this is done,
            but I am not positive.
      
      Also, I fixed the ./mpopt.bat problem.  The issue a Win32
      "feature".  OMake executes a command like ./x with
         executable = absolute pathname of ./x (using \ as a separator).
         argv[0] = ./x
      
      This is entirely sensible to me, but Win2 doesn't like it.
      So the fix is to replace argv[0] with the absolute pathname.
      
      In any case, we should make the OMakefiles as consistent as
      possible, and remove the path separator for mpopt.bat that
      Yegor introduced (BTW, if you ever really needed it, you should
      use the SLASH variable instead of something new).
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-23 14:20:40 -0700 (Mon, 23 May 2005)
Revision: 1057
Log message:

      Be careful about closing files before ExitException.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-23 16:01:46 -0700 (Mon, 23 May 2005)
Revision: 1058
Log message:

      Rules should not export everything, just the rule value itself.
      This addresses issue #487.
      
      You can always put an explicit "export" at the end of an
      OMakefile if you want to export values back to the parent.
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-23 19:48:35 -0700 (Mon, 23 May 2005)
Revision: 1060
Log message:

      Bumping the version number.
      
      Removing the configure-filled version number from OMakeroot.src, reverting
      back to using the gen-magic (and only adding the version limitations to the
      OMakeroot.install, keeping OMakeroot.src free of it).
      

Changes by: Aleksey Nogin (nogin at cs.caltech.edu)
Date: 2005-05-24 20:50:59 -0700 (Tue, 24 May 2005)
Revision: 1061
Log message:

      Added version.txt to GENMAGIC_DEPS
      

Changes by: ( at unknown.email)
Date: 2005-05-24 20:50:59 -0700 (Tue, 24 May 2005)
Revision: 1062
Log message:

      This commit was manufactured by cvs2svn to create branch 'configure2'.

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-31 16:38:49 -0700 (Tue, 31 May 2005)
Revision: 1064
Log message:

      This is some code I wrote on the plane.
      
      This makes the file database (.omc files) more abstract.
      Also, .omo files no longer exist.  The object, if it is ever
      created, is saved as an entry in the .omc file.
      
      There is a placeholder entry for static values, which are
      intended for configuration.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2005-05-31 19:26:27 -0700 (Tue, 31 May 2005)
Revision: 1065
Log message:

      Added a "static" section.
      
      The idea here is that any object can have static values that are
      persistent across OMake runs.
      
          static. =
              foo = $(println Hello World)
      
      This will print Hello World just once the first time this code
      is evaluated.  Subsequent OMake runs will not re-evaluate foo.
      
      NOTE: this is the basic support, but it isn't working yet.