Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-01 09:55:43 -0800 (Fri, 01 Dec 2006)
Revision: 9794
Log message:

      Windows Installer: add OMake's bin directory to the PATH environment variable on install.
      
      P.S. I've touched the .aip file by a more recent version of the Advanced Installer; a number of changes are due to that.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-01 10:46:19 -0800 (Fri, 01 Dec 2006)
Revision: 9795
Log message:

      Fixed some of the bootstrapping problems on Windows.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-01 11:46:29 -0800 (Fri, 01 Dec 2006)
Revision: 9796
Log message:

      - Fixed the string-escaped function - the function was implemented to simply
        add a $''' ... ''' style wrapper to strings containing special characters;
        however this was incompatible with the dependency files lexer (and the main
        usage of the string-escaped function was to escape file names in custom
        scanners!). I've changed the function to do what its documentation promised
        - to simply prefix all the special symbols with a \.
      
      - Minor documentation fixes.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-01 11:48:44 -0800 (Fri, 01 Dec 2006)
Revision: 9797
Log message:

      Regenerated the documentation.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-01 13:50:28 -0800 (Fri, 01 Dec 2006)
Revision: 9799
Log message:

      - OMake now bootstraps and compiles file on Windows
      
      - Fixed some of the issues with the new C scanner on Windows. In particular, it needs to catch all the error and warning messages and print them to stderr.
      
      - Enabled /W4 /WX on Windows and fixed all the warnings

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-01 14:06:58 -0800 (Fri, 01 Dec 2006)
Revision: 9800
Log message:

      Added all the new files and directories to the Advanced Installer project file (Win32).

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-01 14:14:41 -0800 (Fri, 01 Dec 2006)
Revision: 9801
Log message:

      Minor tweaks

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-01 14:25:13 -0800 (Fri, 01 Dec 2006)
Revision: 9802
Log message:

      More Windows installation tweaks.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-04 10:51:25 -0800 (Mon, 04 Dec 2006)
Revision: 9804
Log message:

      - Fixed a number of problems with "ocamldep -modules" postprocessing.
      - Make the "ocamldep -modules" the default dependency mode for building OMake (but still disabled by default in general).
      - Fixed the static test for ocamldep-omake on Windows

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-04 13:51:52 -0800 (Mon, 04 Dec 2006)
Revision: 9805
Log message:

      Deleting the relicense file (we should always be able to resurrect it off the
      SVN if we ever needed to).
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-04 14:32:31 -0800 (Mon, 04 Dec 2006)
Revision: 9806
Log message:

      Further "ocamldep -modules" fixes.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-04 14:41:09 -0800 (Mon, 04 Dec 2006)
Revision: 9807
Log message:

      Fixed a nasty Windows-specific bug - the Omake_exec_local module used a _global_ buffer for processing the stdout/stderr of all the threads. As a result, when running with -j on Windows, the outputs of different threads would often be intermingled and corrupted! In particular, when running with -j, the scanners that use Shell. functions for postprocessing would result in weirdly corrupted .deps file where part of the correct output is replaced with the output of a different scanner!
      
      Took me quite a while to figure this out, but now the -j seems to be working OK on Windows.
      
      P.S. This bug was also present in 0.9.6.9 and probably in (much?) earlier versions as well.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-05 09:49:15 -0800 (Tue, 05 Dec 2006)
Revision: 9810
Log message:

      When --divert-errors is used, print the diversions inside the "targets were
      not rebuilt because of errors" instead of having two separate lists of things
      that have failed to build.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-05 11:05:43 -0800 (Tue, 05 Dec 2006)
Revision: 9811
Log message:

      Clean up the tmp file.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-05 11:26:46 -0800 (Tue, 05 Dec 2006)
Revision: 9812
Log message:

      Renamed and rearranged the --divert-* options. The current "omake --help"
      output now contains:
      
      --output-normal    *:  Relay the output of the rule commands to the OMake
                             output right away. This is the default when no
                             --output-postpone and no --output-only-errors flags
                             are given.
      --output-postpone  *:  [EXPERIMENTAL] Postpone printing command output until
                             a rule terminates.
      --output-only-errors
                         *:  [EXPERIMENTAL] Same as --output-postpone, but
                             postponed output will only be printed for commands
                             that fail.
      --output-at-end    *:  [EXPERIMENTAL] The output of the failed commands will
                             be printed after OMake have stopped.
      
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-05 15:13:35 -0800 (Tue, 05 Dec 2006)
Revision: 9813
Log message:

      - Fix different weirdness in --output-* diversions by making sure
      things are properly flushed on time.
      
      - More consistently redirect the "+/- ..." banner lines to tees when
        appropriate.
      
      This fixes issues 1-3 raised in bug 611.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-05 15:49:47 -0800 (Tue, 05 Dec 2006)
Revision: 9814
Log message:

      Documentation update.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-05 16:08:39 -0800 (Tue, 05 Dec 2006)
Revision: 9815
Log message:

      Updating the changelog for 0.9.8
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-05 16:10:23 -0800 (Tue, 05 Dec 2006)
Revision: 9816
Log message:

      Reordered the changelog items a bit.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-05 16:12:29 -0800 (Tue, 05 Dec 2006)
Revision: 9817
Log message:

      Fixing a typo.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-05 16:45:18 -0800 (Tue, 05 Dec 2006)
Revision: 9818
Log message:

      By default, OMake will now be quiter (as in "omake -S") and print a progress
      bar (unless the stdout is not a TTY).
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-05 18:25:52 -0800 (Tue, 05 Dec 2006)
Revision: 9819
Log message:

      Updating the text of some of the recently added warning messages slightly.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2006-12-05 19:41:00 -0800 (Tue, 05 Dec 2006)
Revision: 9820
Log message:

      Some delayed additions.
      
      Array efficiency:
         It used to be that code like this was quadratic because
         the arrays were flattened during the +=.  Do not do it.
      
         foreach(x, $a)
             y[] += $x
             export
      
      Added a random() function.
      
      Added {decode,encode}-uri functions.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2006-12-05 19:57:15 -0800 (Tue, 05 Dec 2006)
Revision: 9822
Log message:

      We have trouble with functions that use the concat_strings
      function.  The problem is that concat_strings, if given a singleton array,
      exposes it.
      
         concat_strings [v] = v
      
      This is wrong in the following case, because it is an unquote.
      
         concat_strings [ValString x] = ValString x
      
      Perhaps we can fix it generically by defining:
      
         concat_strings [v] = ValQuote v
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-06 09:14:33 -0800 (Wed, 06 Dec 2006)
Revision: 9823
Log message:

      More new functions listed.
      

Changes by: Jason Hickey (jyh at cs.caltech.edu)
Date: 2006-12-06 09:31:51 -0800 (Wed, 06 Dec 2006)
Revision: 9824
Log message:

      Changed array function slightly:
      
         concat_array [ValString s] -> ValData s
         concat_array [ValSequence x] -> ValQuote [ValSequence x]
         concat_array [v] -> v (* otherwise *)
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-06 10:36:02 -0800 (Wed, 06 Dec 2006)
Revision: 9825
Log message:

      Use the LexStringSet module (in place of the StringSet module or a custom
      replacement of it) where appropriate.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-06 11:44:15 -0800 (Wed, 06 Dec 2006)
Revision: 9828
Log message:

      Link in the lm_termsize module (the module is not used just yet).
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-06 12:23:06 -0800 (Wed, 06 Dec 2006)
Revision: 9830
Log message:

      Take the terminal width into account when drawing the progress bar.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-06 14:02:14 -0800 (Wed, 06 Dec 2006)
Revision: 9832
Log message:

      Updated to match the current reality.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-06 15:33:11 -0800 (Wed, 06 Dec 2006)
Revision: 9833
Log message:

      Changed the default value of --output-at-end: it will still be off by default,
      but the -k/-p/-P will now flip this default to true.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-06 16:53:01 -0800 (Wed, 06 Dec 2006)
Revision: 9834
Log message:

      Small improvements in OMake "at end" summary output.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-06 18:59:00 -0800 (Wed, 06 Dec 2006)
Revision: 9835
Log message:

      Minor code optimization: created a "val_of_bool" function.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-06 20:08:17 -0800 (Wed, 06 Dec 2006)
Revision: 9836
Log message:

      Added a --verbose command line option (equivalent to 
      "--no-S --print-status --print-exit VERBOSE=true")
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-06 20:48:06 -0800 (Wed, 06 Dec 2006)
Revision: 9837
Log message:

      (Bug 611, part 4) When creating a new tee for a command, unlink all the tee's
      for the command's effects.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-07 09:03:36 -0800 (Thu, 07 Dec 2006)
Revision: 9838
Log message:

      "-S --progress" is the default now; removing from the bootstrapping Makefiles
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-07 09:05:51 -0800 (Thu, 07 Dec 2006)
Revision: 9839
Log message:

      (Bug 613) Do not try to "unescape" files and directories.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-07 09:22:43 -0800 (Thu, 07 Dec 2006)
Revision: 9840
Log message:

      Added:
      
      OMake 0.9.8 will not be fully backwards-compatible with the earlier releases.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-07 09:57:00 -0800 (Thu, 07 Dec 2006)
Revision: 9841
Log message:

      Minor.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-07 13:13:20 -0800 (Thu, 07 Dec 2006)
Revision: 9842
Log message:

      Omake_build.wait_all function ("*** omake: waiting for all jobs to finish"
      when an OMakefile have changed) was resulting in an infinite loop - fixed.
      
      P.S. This demonstrates why using "let _ = ... in" is a bad idea and should be
      avoided when "let () = ... in" or "ignore(...); " suffice.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-07 14:27:30 -0800 (Thu, 07 Dec 2006)
Revision: 9843
Log message:

      Print progress bar for targets that are pulled out of the cache and found
      up-to-date (since this alone might take a while in a large project),
      not only for those that are actually built.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-07 15:03:59 -0800 (Thu, 07 Dec 2006)
Revision: 9844
Log message:

      Updated the installer a bit.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-07 16:28:39 -0800 (Thu, 07 Dec 2006)
Revision: 9845
Log message:

      Include more information in "These file are targeted separately" warnings.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-07 18:14:49 -0800 (Thu, 07 Dec 2006)
Revision: 9846
Log message:

      (Bug 614, part 1) - fixed the documentation of find-build-targets adn added a
      warning that it might not work.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-08 09:42:50 -0800 (Fri, 08 Dec 2006)
Revision: 9848
Log message:

      Minor tweak.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-08 09:54:24 -0800 (Fri, 08 Dec 2006)
Revision: 9849
Log message:

      Call Pervasives.flush_all before forking. This is needed as the out_channel
      buffers will get duplicated on fork, so anything that is in them may end up
      being outputted several times if we do not flush!
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-08 10:21:06 -0800 (Fri, 08 Dec 2006)
Revision: 9850
Log message:

      When -S is enabled, the scanner commands that produce output (stderr only, as
      stdout is redirected!) should still get an appropriate "banner". We used to
      ignore both stdout and stderr for the purposes of activating the "Lazy Banner"
      on scanner commands, which was wrong.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-08 10:33:39 -0800 (Fri, 08 Dec 2006)
Revision: 9851
Log message:

      In -S mode, the "*** process creation failed" error messages should be
      preceeded by the command "banner".
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-08 10:58:17 -0800 (Fri, 08 Dec 2006)
Revision: 9852
Log message:

      (Bug 588) Printing debug messages to stderr right after the stderr is closed
      is not a very good idea.
      

Changes by: Yegor N. Bryukhov (ybryukhov at gmail.com)
Date: 2006-12-08 11:21:40 -0800 (Fri, 08 Dec 2006)
Revision: 9853
Log message:

      1. Visual C++ 2005 considers strcpy, sprintf, etc deprecated.
      Disabled warning 4996.
      
      2. readline had one function with no return type defined, VC2005 didn't like it, changed to void
      
      I think it would be better to have warning disabling statement kept in one place so that it would be easier to
      change/disable it.
      
      I hope, libmojave part goes to the right place

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-08 11:50:56 -0800 (Fri, 08 Dec 2006)
Revision: 9854
Log message:

      Slightly better interaction between the "saved .omakedb message" and the
      progress bar.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-08 13:52:04 -0800 (Fri, 08 Dec 2006)
Revision: 9855
Log message:

      Fixing a minor issue with the Windows formatting of the "erase the progress bar" string.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-08 14:52:01 -0800 (Fri, 08 Dec 2006)
Revision: 9856
Log message:

      When the output is postponed, do not erase the progress bar when outputting to the diversion file, but do erase it before printing the diversion file in the end.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-08 15:07:48 -0800 (Fri, 08 Dec 2006)
Revision: 9857
Log message:

      On Windows, we spawn two separate threads for each job - one watching stdout and one watching stderr. The two threads might get activated in parallel, so we need to have two separate buffers to make sure the stdout and stderr do not get intermingled (which, in particular, might lead to corrupted scanner output).

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-08 15:54:10 -0800 (Fri, 08 Dec 2006)
Revision: 9858
Log message:

      Include the "scanner produced ill-formed output" in the diverted output
      (instead of always directing it to stderr).
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-08 16:47:20 -0800 (Fri, 08 Dec 2006)
Revision: 9859
Log message:

      Removed couple of unused "open" statements.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-08 16:51:40 -0800 (Fri, 08 Dec 2006)
Revision: 9860
Log message:

      In my rev. 9856 commit (where I reorginized the progress bar erasure code), I
      forgot to add the progress bar erasure in one place - fixing.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-08 17:04:52 -0800 (Fri, 08 Dec 2006)
Revision: 9861
Log message:

      If the progress bar war erased to make place for some output, print it back on
      the next iteration of the mail build loop.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-08 17:11:24 -0800 (Fri, 08 Dec 2006)
Revision: 9862
Log message:

      LazyError functions must use fprintf, not eprintf.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-08 17:26:45 -0800 (Fri, 08 Dec 2006)
Revision: 9863
Log message:

      Added a hack to keep retrying unlinking those diversion files that we failed to unlink. This is needed on Windows, where some unrelated threads might still have a handle to the diversion file.
      
      The correct approach would probably be to set the close_on_exec flag for the diversion fd _after_ all the threads related to a job have been spawned, but I am too lazy for that.

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-09 09:51:59 -0800 (Sat, 09 Dec 2006)
Revision: 9865
Log message:

      Regenerated the Makefiles.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-10 14:17:25 -0800 (Sun, 10 Dec 2006)
Revision: 9866
Log message:

      The $0, $1, $2, ... variables introduced by the awk, match, and similar
      functions should be ValData, not ValString.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-11 09:57:32 -0800 (Mon, 11 Dec 2006)
Revision: 9867
Log message:

      Preparing to release 0.9.8!
      
      - Finished filling in the changelogs.
      - Bumped the version number to 0.9.8-1 in the approproate places.
      - Regenerated the documentation (had to fix a couple of minor problems with it).
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-11 10:06:12 -0800 (Mon, 11 Dec 2006)
Revision: 9868
Log message:

      Fixing the "publish" script.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-11 10:18:44 -0800 (Mon, 11 Dec 2006)
Revision: 9869
Log message:

      Some .spec file fixes.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-11 10:23:59 -0800 (Mon, 11 Dec 2006)
Revision: 9870
Log message:

      Adding %{?dist} to the release.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-11 11:03:26 -0800 (Mon, 11 Dec 2006)
Revision: 9871
Log message:

      Deleted the "#!/usr/local/bin/ocamlrun" from the ocamldep-omake to avoid
      problems with various packaging tools (like RPM) that take this to be a
      dependency.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-11 11:20:44 -0800 (Mon, 11 Dec 2006)
Revision: 9874
Log message:

      Minor documentation update.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-11 12:18:06 -0800 (Mon, 11 Dec 2006)
Revision: 9875
Log message:

      Tagging the released version.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-11 20:43:47 -0800 (Mon, 11 Dec 2006)
Revision: 9876
Log message:

      We will have two OS X packages - PPC and x86.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-11 21:04:06 -0800 (Mon, 11 Dec 2006)
Revision: 9877
Log message:

      Spelling.
      

Changes by: Nathaniel Gray (n8gray at caltech.edu)
Date: 2006-12-12 00:16:35 -0800 (Tue, 12 Dec 2006)
Revision: 9878
Log message:

      Changes I made for packaging 0.9.8-1.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-13 13:33:00 -0800 (Wed, 13 Dec 2006)
Revision: 9879
Log message:

      - (Bug 617) Disabling the OCAMLDEP_MODULES_ENABLED for self-build for now.
      
      - Making "omake clean" a bit more clean.
      
      - Bumping the version to 0.9.8-2 - I will create new a source tarballs, but no
        need to create new binaries as the OCAMLDEP_MODULES_ENABLED problem is only
        affecting the bootstrapping.
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-13 13:56:54 -0800 (Wed, 13 Dec 2006)
Revision: 9880
Log message:

      Tagging 0.9.8-2
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-18 09:43:23 -0800 (Mon, 18 Dec 2006)
Revision: 9888
Log message:

      - If the lockf on .omakedb.lock fails with Unix.ENOLCK, we should not abort.
        Thanks to Phil Endecott for reporting this!
      
      - Bumping the version number to 0.9.8.0+
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-21 09:11:14 -0800 (Thu, 21 Dec 2006)
Revision: 9890
Log message:

      It's --no--progress, not --no-progress (thanks to malc for noticing the
      problem).
      

Changes by: Aleksey Nogin (nogin at metaprl.org)
Date: 2006-12-21 09:45:05 -0800 (Thu, 21 Dec 2006)
Revision: 9891
Log message:

      Rebuilt the documentation.