Merged revisions 2136-2200,2202-2290,2292-2301 via svnmerge from

http://scons.tigris.org/svn/scons/branches/core

........
  r2145 | stevenknight | 2007-07-17 09:15:12 -0500 (Tue, 17 Jul 2007) | 3 lines

  Don't put null strings (from variable expansion) in a path list.
  (They get turned into the current directory on later expansion.)
........
  r2146 | stevenknight | 2007-07-17 10:47:39 -0500 (Tue, 17 Jul 2007) | 3 lines

  Add support for optional arguments on command-line long options
  by specifying nargs='?'.
........
  r2149 | stevenknight | 2007-07-17 15:22:24 -0500 (Tue, 17 Jul 2007) | 2 lines

  Remove left-over Optik mentions.
........
  r2150 | stevenknight | 2007-07-17 15:39:34 -0500 (Tue, 17 Jul 2007) | 4 lines

  Add a $SWIGPATH variable for finding SWIG dependencies, with
  $SWIGINC{PREFIX,SUFFIX} for adding them to the command line.
........
  r2154 | stevenknight | 2007-07-18 20:05:31 -0500 (Wed, 18 Jul 2007) | 2 lines

  Fix variable misspellings in the doc added for $SWIGOUTPUT.
........
  r2155 | stevenknight | 2007-07-18 20:07:28 -0500 (Wed, 18 Jul 2007) | 2 lines

  Add the Python eggs info file to the RPM packaging build.
........
  r2156 | stevenknight | 2007-07-18 20:15:08 -0500 (Wed, 18 Jul 2007) | 2 lines

  Convert documentation from DocBook SGML to XML.
........
  r2158 | stevenknight | 2007-07-19 17:16:19 -0500 (Thu, 19 Jul 2007) | 3 lines

  Conditionally add the .egg-info the RPM file list only if the distutils
  in the version of Python that rpmbuild will execute knows about them.
........
  r2161 | stevenknight | 2007-07-19 19:12:29 -0500 (Thu, 19 Jul 2007) | 5 lines

  Capture a test case (contributed by Tilo Prutz) where instantiation of
  a private class causes javac to generate an additional anonymous inner
  class file.  (No solution yet, but there's no sense throwing away the
  preparatory work.)
........
  r2162 | stevenknight | 2007-07-20 11:29:56 -0500 (Fri, 20 Jul 2007) | 3 lines

  Support passing a list of .java files as source to the Java() builder.
  (Leanid Nazdrynau)
........
  r2163 | garyo | 2007-07-20 12:00:35 -0500 (Fri, 20 Jul 2007) | 1 line

  Fixed cut-n-paste error in Touch factory method doc in users guide.
........
  r2167 | stevenknight | 2007-07-21 22:59:40 -0500 (Sat, 21 Jul 2007) | 2 lines

  Don't execute the SWIGOUTDIR test if swig isn't installed.
........
  r2168 | stevenknight | 2007-07-21 23:14:17 -0500 (Sat, 21 Jul 2007) | 2 lines

  Fix the test's ability to run under a path name containing spaces.
........
  r2171 | stevenknight | 2007-07-24 15:54:41 -0500 (Tue, 24 Jul 2007) | 2 lines

  Handle white space in key file names in the packaging build.
........
  r2172 | stevenknight | 2007-07-24 21:41:15 -0500 (Tue, 24 Jul 2007) | 2 lines

  More efficient copying of construction environments.
........
  r2173 | stevenknight | 2007-07-25 10:56:02 -0500 (Wed, 25 Jul 2007) | 2 lines

  Update the SCons build for Subversion and general clean-up.
........
  r2174 | stevenknight | 2007-07-25 11:35:16 -0500 (Wed, 25 Jul 2007) | 3 lines

  Suppress the [brackets] around a node in the --tree=prune output if
  the node is a source.
........
  r2175 | stevenknight | 2007-07-25 12:52:18 -0500 (Wed, 25 Jul 2007) | 3 lines

  Commonize the skip_test() method and make its behavior configurable
  via a TESTCOMMON_PASS_SKIPS environment variable.
........
  r2178 | stevenknight | 2007-07-25 21:43:47 -0500 (Wed, 25 Jul 2007) | 3 lines

  Add $JAVACLASSPATH and $JAVASOURCEPATH construction variables.  (Leanid
  Nazdrynau)
........
  r2182 | stevenknight | 2007-07-30 12:10:20 -0500 (Mon, 30 Jul 2007) | 3 lines

  Refactor Builder suffix-adjusting into its own method, so we can
  (potentially) re-use it for Builders with attached source Builders.
........
  r2183 | stevenknight | 2007-07-30 14:51:53 -0500 (Mon, 30 Jul 2007) | 2 lines

  More efficient source-builder suffix matching.
........
  r2184 | stevenknight | 2007-07-30 16:01:42 -0500 (Mon, 30 Jul 2007) | 4 lines

  Encapsulate initialization of the default FS object by an accessor
  function in SCons.Node.FS.  (This also gets rid of an unnecessary
  reference to SCons.Node.FS.default_fs in the LaTeX scanner.)
........
  r2193 | stevenknight | 2007-07-30 18:24:07 -0500 (Mon, 30 Jul 2007) | 3 lines

  Fix interpretation of source arguments that have no suffix when the
  called Builder has both a src_suffix and a src_builder.
........
  r2194 | stevenknight | 2007-07-31 10:25:31 -0500 (Tue, 31 Jul 2007) | 2 lines

  Increase the number of tries for random output from three to ten.
........
  r2195 | stevenknight | 2007-07-31 10:52:28 -0500 (Tue, 31 Jul 2007) | 3 lines

  Skip the test gracefully if the zipfile module can't read the file it
  just wrote (which is the case for Python 2.1 on 64-bit systems).
........
  r2196 | stevenknight | 2007-07-31 13:06:21 -0500 (Tue, 31 Jul 2007) | 2 lines

  Move the "import zipfile" so it doesn't fail on Python <= 2.0.
........
  r2197 | stevenknight | 2007-07-31 14:51:50 -0500 (Tue, 31 Jul 2007) | 3 lines

  Commonize initialization of the various Java builders so they can be
  hooked up into a multi-stage Builder chain.  (Leanid Nazdrynau)
........
  r2198 | stevenknight | 2007-07-31 16:15:18 -0500 (Tue, 31 Jul 2007) | 3 lines

  Fix use of ${TARGET.dir} and ${SOURCE.dir} expansions in $FORTRANMODDIR
  $JARCHDIR, $JARFLAGS, $LEXFLAGS, $SWIGFLAGS, $SWIGOUTDIR and $YACCFLAGS.
........
  r2199 | stevenknight | 2007-07-31 16:25:48 -0500 (Tue, 31 Jul 2007) | 2 lines

  Remove left-over Trace() call.
........
  r2202 | stevenknight | 2007-08-01 12:31:48 -0500 (Wed, 01 Aug 2007) | 3 lines

  Bail out via test.skip_test() if wix ("candle") isn't found.
  Put the main body of code flush left instead of under an if: block.
........
  r2203 | stevenknight | 2007-08-01 15:35:55 -0500 (Wed, 01 Aug 2007) | 5 lines

  Fix Tool.packaging.rpm.package() so it doesn't always overwrite
  $RPMFLAGS with -ta.
  Set --buildroot in RPM packaging tests so they don't overwrite
  each other when run simultaneously.
........
  r2204 | stevenknight | 2007-08-01 15:37:36 -0500 (Wed, 01 Aug 2007) | 2 lines

  Fix a nested scope issue with the internal build_sources() function.
........
  r2205 | stevenknight | 2007-08-01 15:46:08 -0500 (Wed, 01 Aug 2007) | 5 lines

  Normalize (X out) the CreationDate field inside embedded, compressed
  PostScript streams within the generated PDF files.  Also normalize
  preceding Length field, since compression length is affected by different
  patterns of input, including the variable CreationDate value.
........
  r2211 | stevenknight | 2007-08-02 08:52:06 -0500 (Thu, 02 Aug 2007) | 2 lines

  Add the new modules from branches/packaging to the SCons packaging build.
........
  r2212 | stevenknight | 2007-08-02 19:59:01 -0500 (Thu, 02 Aug 2007) | 2 lines

  Fix the JAVACLASSPATH test when javah isn't on the default $PATH.
........
  r2214 | stevenknight | 2007-08-03 15:05:21 -0500 (Fri, 03 Aug 2007) | 4 lines

  Hook up the Java builders into a multi-step chain underneath a Java()
  pseudo-builder (wrapper) that examines its arguments and calls the
  appropriate underlying file-or-dir builder.
........
  r2215 | stevenknight | 2007-08-03 15:49:58 -0500 (Fri, 03 Aug 2007) | 2 lines

  Fix for old Python versions:  use apply() instead of *args, **kw.
........
  r2216 | stevenknight | 2007-08-03 16:49:31 -0500 (Fri, 03 Aug 2007) | 2 lines

  Hook up the SWIG builder as a source builder for .java files.
........
  r2217 | stevenknight | 2007-08-03 17:28:19 -0500 (Fri, 03 Aug 2007) | 2 lines

  Don't use .endswith(), which didn't appear until later Python versions.
........
  r2218 | stevenknight | 2007-08-03 17:29:38 -0500 (Fri, 03 Aug 2007) | 2 lines

  Replace tabs with spaces.
........
  r2219 | stevenknight | 2007-08-04 08:06:23 -0500 (Sat, 04 Aug 2007) | 3 lines

  Initialize a loop-invariant lambda for matching .java suffixes outside
  the loop.
........
  r2220 | stevenknight | 2007-08-07 15:06:13 -0500 (Tue, 07 Aug 2007) | 2 lines

  Refactor parallel class-generation loops into one.
........
  r2221 | stevenknight | 2007-08-07 16:04:06 -0500 (Tue, 07 Aug 2007) | 5 lines

  Have the Java multi-step builder test actually check for generated files,
  and fix the generation of .java and .class file names, and interaction
  with the SWIG builder, so that the files are generated in the correct
  place.
........
  r2222 | stevenknight | 2007-08-07 16:45:05 -0500 (Tue, 07 Aug 2007) | 3 lines

  Fix dependencies on SWIG-generated .java files so they don't have to
  be built in multiple passes.
........
  r2226 | stevenknight | 2007-08-07 18:00:22 -0500 (Tue, 07 Aug 2007) | 2 lines

  Fix SWIG when used with BuildDir().
........
  r2227 | stevenknight | 2007-08-07 22:15:55 -0500 (Tue, 07 Aug 2007) | 5 lines

  User's guide updates:
  - Make the multiple files example match its text.
  - Expand a truncated sentence about being able to use Python function actions
    in the Command() Builder.
........
  r2228 | stevenknight | 2007-08-07 23:25:18 -0500 (Tue, 07 Aug 2007) | 3 lines

  Don't generate an error if a #include file matches a same-named
  directory in $CPPPATH (or $FORTRANPATH, etc.).
........
  r2229 | stevenknight | 2007-08-07 23:40:00 -0500 (Tue, 07 Aug 2007) | 2 lines

  Fix a code example.  (Gary Oberbrunner)
........
  r2230 | stevenknight | 2007-08-08 00:05:43 -0500 (Wed, 08 Aug 2007) | 3 lines

  Capture a test case to make sure AddPostAction() doesn't interfere
  with normal linking.  (Matt Doar, Gary Oberbrunner)
........
  r2233 | stevenknight | 2007-08-08 14:15:44 -0500 (Wed, 08 Aug 2007) | 2 lines

  Fix documentation typo in a construction variable cross-reference.
........
  r2234 | stevenknight | 2007-08-08 17:03:25 -0500 (Wed, 08 Aug 2007) | 2 lines

  Changes to SCons packaging to support checkpoint releases.
........
  r2235 | stevenknight | 2007-08-09 10:10:01 -0500 (Thu, 09 Aug 2007) | 2 lines

  Sidestep false negatives on heavily loaded systems.
........
  r2236 | garyo | 2007-08-09 11:16:26 -0500 (Thu, 09 Aug 2007) | 1 line

  Allow unpackaged files (e.g. *.pyo) to exist in the build dir without being packaged in the RPM.  Without this, on some systems the rpmbuild may error out.
........
  r2237 | stevenknight | 2007-08-09 11:27:56 -0500 (Thu, 09 Aug 2007) | 5 lines

  Fix test/SWIG/build-dir.py so it works on old Python versions without
  distutils.sysconfig.
  Instead of just cutting-and-pasting initialization code from other
  SWIG tests, centralize it in some new TestSCons methods.
........
  r2238 | garyo | 2007-08-09 11:30:58 -0500 (Thu, 09 Aug 2007) | 1 line

  Use docbook 4.3 instead of 4.4 for the XML doctype since some older(?) jade parsers can't handle new 4-byte Unicode chars in the 4.4 version of isogrk4.ent.
........
  r2240 | stevenknight | 2007-08-09 16:35:06 -0500 (Thu, 09 Aug 2007) | 2 lines

  User's Guide updates (post packaging changes).
........
  r2243 | stevenknight | 2007-08-10 10:31:51 -0500 (Fri, 10 Aug 2007) | 3 lines

  Fix the User's Guide build to use openjade, and to accomodate a change
  in the name of the main generated file (book1.html => index.html).
........
  r2245 | stevenknight | 2007-08-10 11:09:16 -0500 (Fri, 10 Aug 2007) | 2 lines

  Update the {CHANGES,RELEASE}.txt datestamp lines.
........
  r2253 | stevenknight | 2007-08-10 16:21:54 -0500 (Fri, 10 Aug 2007) | 2 lines

  Fix the wix Tool module's ability to handle null entries in $PATH.
........
  r2261 | stevenknight | 2007-08-11 23:08:12 -0500 (Sat, 11 Aug 2007) | 3 lines

  Remove unnecessary files (.svnt/*, .{ae,cvs}ignore, www/*) from the
  scons-src packages.
........
  r2262 | stevenknight | 2007-08-11 23:24:49 -0500 (Sat, 11 Aug 2007) | 2 lines

  Add missing __revision__ lines.
........
  r2263 | stevenknight | 2007-08-11 23:33:42 -0500 (Sat, 11 Aug 2007) | 2 lines

  Skip the test if the MANIFEST file hasn't been built.
........
  r2264 | stevenknight | 2007-08-11 23:36:30 -0500 (Sat, 11 Aug 2007) | 2 lines

  Add recent compatibility modules to the relevant exceptions lists.
........
  r2265 | stevenknight | 2007-08-11 23:39:00 -0500 (Sat, 11 Aug 2007) | 3 lines

  Update __VERSION__ strings in the QMTest/*.py modules, so that packaging
  tests (src/test_*.py) will pass after builds of checkpoint releases.
........
  r2266 | stevenknight | 2007-08-12 07:36:19 -0500 (Sun, 12 Aug 2007) | 2 lines

  Add a comment about why we construct the __VERSION__ string at run time.
........
  r2267 | stevenknight | 2007-08-12 07:42:30 -0500 (Sun, 12 Aug 2007) | 2 lines

  Avoid reading the MANIFEST file twice.  (Courtesy review by Greg Noel.)
........
  r2268 | stevenknight | 2007-08-12 08:14:53 -0500 (Sun, 12 Aug 2007) | 3 lines

  Shift Install() and InstallAs() from being documented as functions
  to being documented as Builders.
........
  r2269 | garyo | 2007-08-13 08:49:52 -0500 (Mon, 13 Aug 2007) | 1 line

  Tests: Skip some more Java tests if javac is not installed on the test machine so they don't get marked as failing.
........
  r2270 | garyo | 2007-08-13 11:09:39 -0500 (Mon, 13 Aug 2007) | 1 line

  Fixed typo in test (shows up on non-Linux platforms).
........
  r2271 | garyo | 2007-08-13 14:09:05 -0500 (Mon, 13 Aug 2007) | 4 lines

  Test portability fixes for Darwin/OSX and IRIX.
  This does not make all the tests pass on those OSes,
  but it takes care of some of the more obvious errors that
  I have time for right now.  More to come.
........
  r2272 | stevenknight | 2007-08-13 15:33:29 -0500 (Mon, 13 Aug 2007) | 2 lines

  Tab => space fix.
........
  r2273 | stevenknight | 2007-08-13 15:33:52 -0500 (Mon, 13 Aug 2007) | 2 lines

  Test for swig, too, which is used to build from the .i file.
........
  r2277 | garyo | 2007-08-14 10:40:00 -0500 (Tue, 14 Aug 2007) | 8 lines

  Test portability on IRIX: test/Actions/pre-post creates target file
  before building target, then IRIX CC does not chmod +x afterwards.
  I think this change is safe on all OSes.

  test/AS/ml.py: I think this is only supposed to be run on win32
  (not skipped only on win32); the sense of the skip test was backwards.
........
  r2278 | stevenknight | 2007-08-14 11:04:40 -0500 (Tue, 14 Aug 2007) | 2 lines

  Add -tt when running tests, to catch inconsistent tab usage.
........
  r2279 | stevenknight | 2007-08-14 14:00:43 -0500 (Tue, 14 Aug 2007) | 2 lines

  Minor refactor of logic in File.retrieve_from_cache().
........
  r2280 | stevenknight | 2007-08-15 01:11:40 -0500 (Wed, 15 Aug 2007) | 2 lines

  Refactor CacheDir support into its own module.
........
  r2281 | stevenknight | 2007-08-15 07:24:51 -0500 (Wed, 15 Aug 2007) | 2 lines

  Move the cachepath() method from FS.File to the CacheDir class.
........
  r2282 | stevenknight | 2007-08-15 08:31:34 -0500 (Wed, 15 Aug 2007) | 2 lines

  Python 1.5.2 fix in the new Null class.
........
  r2283 | stevenknight | 2007-08-15 10:45:53 -0500 (Wed, 15 Aug 2007) | 5 lines

  Refactor CacheDir unit tests to:
  - restore functionality that was dropped in the transition;
  - commonize creation of test Nodes and other (mock) objects
  - separate CacheDir tests from tests of CacheDir through Node.FS.File.
........
  r2284 | stevenknight | 2007-08-15 11:46:38 -0500 (Wed, 15 Aug 2007) | 3 lines

  Replace the Executor.Null.NullEnvironment object with a real Null object,
  so it will absorb the CacheDir method calls as well.
........
  r2285 | stevenknight | 2007-08-15 11:52:57 -0500 (Wed, 15 Aug 2007) | 5 lines

  Add a get_CacheDir() method to a construction environment, which will
  be used to fetch per-environment CacheDir specifications.  (Right now
  all calls to it still just return the one attached to underlying default
  FS object.)
........
  r2286 | stevenknight | 2007-08-15 15:15:46 -0500 (Wed, 15 Aug 2007) | 2 lines

  Support per-construction-environment configuration of CacheDir().
........
  r2287 | stevenknight | 2007-08-15 15:33:04 -0500 (Wed, 15 Aug 2007) | 2 lines

  Move the tests of CacheDir()-related command-line options into test/CacheDir.
........
  r2293 | stevenknight | 2007-08-16 11:14:49 -0500 (Thu, 16 Aug 2007) | 3 lines

  Add the Package() builder description to the documentation build,
  fixing the XML so that it will build.
........
  r2294 | stevenknight | 2007-08-16 12:51:19 -0500 (Thu, 16 Aug 2007) | 3 lines

  Reorganize packaging documentation:  alphabetize the variable definitions
  (and function names), document Tag() as a function, not a builder.
........
  r2296 | stevenknight | 2007-08-16 12:55:01 -0500 (Thu, 16 Aug 2007) | 2 lines

  Add a build command.
........
  r2300 | stevenknight | 2007-08-16 16:49:13 -0500 (Thu, 16 Aug 2007) | 2 lines

  First cut at documenting packaging variables.
........
  r2301 | stevenknight | 2007-08-16 16:51:21 -0500 (Thu, 16 Aug 2007) | 3 lines

  Construct the .src.rpm and .arch.rpm file names independnetly, not
  by trying to massage one into the other.
........

1 files changed, 917 insertions, 0 deletions

diff --git a/doc/user/depends.xml b/doc/user/depends.xml
new file mode 100644
index 0000000..9e055ee
--- /dev/null
+++ b/doc/user/depends.xml
@@ -0,0 +1,917 @@
+<!--
+
+  __COPYRIGHT__
+
+  Permission is hereby granted, free of charge, to any person obtaining
+  a copy of this software and associated documentation files (the
+  "Software"), to deal in the Software without restriction, including
+  without limitation the rights to use, copy, modify, merge, publish,
+  distribute, sublicense, and/or sell copies of the Software, and to
+  permit persons to whom the Software is furnished to do so, subject to
+  the following conditions:
+
+  The above copyright notice and this permission notice shall be included
+  in all copies or substantial portions of the Software.
+
+  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY
+  KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE
+  WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
+  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+-->
+
+  <para>
+
+  So far we've seen how &SCons; handles one-time builds.
+  But the real point of a build tool like &SCons;
+  is to rebuild only the necessary things
+  when source files change--or, put another way,
+  &SCons; should <emphasis>not</emphasis>
+  waste time rebuilding things that have already been built.
+  You can see this at work simply be re-invoking &SCons;
+  after building our simple &hello; example:
+
+  </para>
+
+  
+
+  <screen>
+     % <userinput>scons -Q</userinput>
+     cc -o hello.o -c hello.c
+     cc -o hello hello.o
+     % <userinput>scons -Q</userinput>
+     scons: `.' is up to date.
+  </screen>
+
+  <para>
+
+  The second time it is executed,
+  &SCons; realizes that the &hello; program
+  is up-to-date with respect to the current &hello_c; source file,
+  and avoids rebuilding it.
+  You can see this more clearly by naming
+  the &hello; program explicitly on the command line:
+
+  </para>
+
+  <screen>
+     % <userinput>scons -Q hello</userinput>
+     cc -o hello.o -c hello.c
+     cc -o hello hello.o
+     % <userinput>scons -Q hello</userinput>
+     scons: `hello' is up to date.
+  </screen>
+
+  <para>
+
+  Note that &SCons; reports <literal>"...is up to date"</literal>
+  only for target files named explicitly on the command line,
+  to avoid cluttering the output.
+
+  </para>
+
+  <section>
+  <title>Deciding When a Source File Has Changed:  the &SourceSignatures; Function</title>
+
+    <para>
+
+    The other side of avoiding unnecessary rebuilds
+    is the fundamental build tool behavior
+    of <emphasis>rebuilding</emphasis>
+    things when a source file changes,
+    so that the built software is up to date.
+    &SCons; keeps track of this through a
+    &signature; for each source file,
+    and allows you to configure
+    whether you want to use the source
+    file contents or the modification time (timestamp)
+    as the signature.
+
+    </para>
+
+    <section>
+    <title>MD5 Source File Signatures</title>
+
+      <para>
+
+      By default,
+      &SCons; keeps track of whether a source file has changed
+      based on the file's contents,
+      not the modification time.
+      This means that you may be surprised by the
+      default &SCons; behavior if you are used to the
+      &Make; convention of forcing
+      a rebuild by updating the file's modification time
+      (using the &touch; command, for example):
+
+      </para>
+
+      <screen>
+         % <userinput>scons -Q hello</userinput>
+         cc -o hello.o -c hello.c
+         cc -o hello hello.o
+         % <userinput>touch hello.c</userinput>
+         % <userinput>scons -Q hello</userinput>
+         scons: `hello' is up to date.
+      </screen>
+
+      <para>
+
+      Even though the file's modification time has changed,
+      &SCons; realizes that the contents of the
+      &hello_c; file have <emphasis>not</emphasis> changed,
+      and therefore that the &hello; program
+      need not be rebuilt.
+      This avoids unnecessary rebuilds when,
+      for example, someone rewrites the
+      contents of a file without making a change.
+      But if the contents of the file really do change,
+      then &SCons; detects the change
+      and rebuilds the program as required:
+
+      </para>
+
+      <screen>
+         % <userinput>scons -Q hello</userinput>
+         cc -o hello.o -c hello.c
+         cc -o hello hello.o
+         % <userinput>edit hello.c</userinput>
+             [CHANGE THE CONTENTS OF hello.c]
+         % <userinput>scons -Q hello</userinput>
+         cc -o hello.o -c hello.c
+         cc -o hello hello.o
+      </screen>
+
+      <para>
+
+      Note that you can, if you wish,
+      specify this default behavior
+      (MD5 signatures) explicitly
+      using the &SourceSignatures; function as follows:
+
+      </para>
+
+      <programlisting>
+        Program('hello.c')
+        SourceSignatures('MD5')
+      </programlisting>
+
+    </section>
+
+    <section>
+    <title>Source File Time Stamps</title>
+
+      <para>
+
+      If you prefer, you can
+      configure &SCons; to use the modification time
+      of source files,
+      not the file contents,
+      when deciding if something needs to be rebuilt.
+      To do this, call the &SourceSignatures;
+      function as follows:
+
+      </para>
+
+      <programlisting>
+        Program('hello.c')
+        SourceSignatures('timestamp')
+      </programlisting>
+
+      <para>
+
+      This makes &SCons; act like &Make;
+      when a file's modification time is updated
+      (using the &touch; command, for example):
+
+      </para>
+
+      <screen>
+         % <userinput>scons -Q hello</userinput>
+         cc -o hello.o -c hello.c
+         cc -o hello hello.o
+         % <userinput>touch hello.c</userinput>
+         % <userinput>scons -Q hello</userinput>
+         cc -o hello.o -c hello.c
+         cc -o hello hello.o
+      </screen>
+
+    </section>
+
+  </section>
+
+  <section>
+  <title>Deciding When a Target File Has Changed:  the &TargetSignatures; Function</title>
+
+    <para>
+
+    As you've just seen,
+    &SCons; uses signatures to decide whether a
+    target file is up to date or must be rebuilt.
+    When a target file depends on another target file,
+    &SCons; allows you to configure separately
+    how the signatures of "intermediate" target files
+    are used when deciding if a dependent target file
+    must be rebuilt.
+
+    </para>
+
+    <section>
+    <title>Build Signatures</title>
+
+      <para>
+
+      Modifying a source file
+      will cause not only its direct target file to be rebuilt,
+      but also the target file(s)
+      that depend on that direct target file.
+      In our example,
+      changing the contents of the &hello_c; file causes
+      the &hello_o; file to be rebuilt,
+      which in turn causes the
+      &hello; program to be rebuilt:
+
+      </para>
+
+      <screen>
+         % <userinput>scons -Q hello</userinput>
+         cc -o hello.o -c hello.c
+         cc -o hello hello.o
+         % <userinput>edit hello.c</userinput>
+             [CHANGE THE CONTENTS OF hello.c]
+         % <userinput>scons -Q hello</userinput>
+         cc -o hello.o -c hello.c
+         cc -o hello hello.o
+      </screen>
+
+      <para>
+
+      What's not obvious, though,
+      is that &SCons; internally handles the signature of
+      the target file(s)
+      (&hello_o; in the above example)
+      differently from the signature of the source file
+      (&hello_c;).
+      By default,
+      &SCons; tracks whether a target file must be rebuilt
+      by using a &buildsignature;
+      that consists of the combined
+      signatures of all the files
+      that go into making the target file.
+      This is efficient because
+      the accumulated signatures
+      actually give &SCons; all of the
+      information it needs
+      to decide if the target file is out of date.
+
+      </para>
+
+      <para>
+
+      If you wish, you can
+      specify this default behavior
+      (build signatures) explicitly
+      using the &TargetSignatures; function:
+
+      </para>
+
+      <programlisting>
+        Program('hello.c')
+        TargetSignatures('build')
+      </programlisting>
+
+    </section>
+
+    <section>
+    <title>File Contents</title>
+
+      <para>
+
+      Sometimes a source file can be changed
+      in such a way that the contents of the
+      rebuilt target file(s)
+      will be exactly the same as the last time
+      the file was built.
+      If so, then any other target files
+      that depend on such a built-but-not-changed target
+      file actually need not be rebuilt.
+      You can make &SCons;
+      realize that it does not need to rebuild
+      a dependent target file in this situation
+      using the &TargetSignatures; function as follows:
+
+      </para>
+
+      <programlisting>
+        Program('hello.c')
+        TargetSignatures('content')
+      </programlisting>
+
+      <para>
+
+      So if, for example,
+      a user were to only change a comment in a C file,
+      then the rebuilt &hello_o; file
+      would be exactly the same as the one previously built
+      (assuming the compiler doesn't put any build-specific
+      information in the object file).
+      &SCons; would then realize that it would not
+      need to rebuild the &hello; program as follows:
+
+      </para>
+
+      <screen>
+         % <userinput>scons -Q hello</userinput>
+         cc -o hello.o -c hello.c
+         cc -o hello hello.o
+         % <userinput>edit hello.c</userinput>
+           [CHANGE A COMMENT IN hello.c]
+         % <userinput>scons -Q hello</userinput>
+         cc -o hello.o -c hello.c
+         scons: `hello' is up to date.
+      </screen>
+
+      <para>
+
+      In essence, &SCons; has
+      "short-circuited" any dependent builds
+      when it realizes that a target file
+      has been rebuilt to exactly the same file as the last build.
+      So configured,
+      &SCons; does take some extra processing time
+      to scan the contents of the target (&hello_o;) file,
+      but this may save time
+      if the rebuild that was avoided
+      would have been very time-consuming and expensive.
+
+      </para>
+
+    </section>
+
+  </section>
+
+  <section>
+  <title>Implicit Dependencies:  The &cv-CPPPATH; Construction Variable</title>
+
+    <para>
+
+    Now suppose that our "Hello, World!" program
+    actually has a <literal>#include</literal> line
+    to include the &hello_h; file in the compilation:
+
+    </para>
+
+    <programlisting>
+       #include &lt;hello.h&gt;
+       int
+       main()
+       {
+           printf("Hello, %s!\n", string);
+       }
+    </programlisting>
+
+    <para>
+
+    And, for completeness, the &hello_h; file looks like this:
+
+    </para>
+
+    
+    <programlisting>
+       #define string    "world"
+      </programlisting>
+
+    <para>
+
+    In this case, we want &SCons; to recognize that,
+    if the contents of the &hello_h; file change,
+    the &hello; program must be recompiled.
+    To do this, we need to modify the
+    &SConstruct; file like so:
+
+    </para>
+
+    
+    <programlisting>
+       Program('hello.c', CPPPATH = '.')
+      </programlisting>
+
+    <para>
+
+    The &cv-link-CPPPATH; value
+    tells &SCons; to look in the current directory
+    (<literal>'.'</literal>)
+    for any files included by C source files
+    (<filename>.c</filename> or <filename>.h</filename> files).
+    With this assignment in the &SConstruct; file:
+
+    </para>
+
+    <screen>
+       % <userinput>scons -Q hello</userinput>
+       cc -o hello.o -c -I. hello.c
+       cc -o hello hello.o
+       % <userinput>scons -Q hello</userinput>
+       scons: `hello' is up to date.
+       % <userinput>edit hello.h</userinput>
+           [CHANGE THE CONTENTS OF hello.h]
+       % <userinput>scons -Q hello</userinput>
+       cc -o hello.o -c -I. hello.c
+       cc -o hello hello.o
+    </screen>
+
+    <para>
+
+    First, notice that &SCons;
+    added the <literal>-I.</literal> argument
+    from the &cv-CPPPATH; variable
+    so that the compilation would find the
+    &hello_h; file in the local directory.
+
+    </para>
+
+    <para>
+
+    Second, realize that &SCons; knows that the &hello;
+    program must be rebuilt
+    because it scans the contents of
+    the &hello_c; file
+    for the <literal>#include</literal> lines that indicate
+    another file is being included in the compilation.
+    &SCons; records these as
+    <emphasis>implicit dependencies</emphasis>
+    of the target file,
+    Consequently,
+    when the &hello_h; file changes,
+    &SCons; realizes that the &hello_c; file includes it,
+    and rebuilds the resulting &hello; program
+    that depends on both the &hello_c; and &hello_h; files.
+
+    </para>
+
+    <para>
+
+    Like the &cv-link-LIBPATH; variable,
+    the &cv-CPPPATH; variable
+    may be a list of directories,
+    or a string separated by
+    the system-specific path separate character
+    (':' on POSIX/Linux, ';' on Windows).
+    Either way, &SCons; creates the
+    right command-line options
+    so that the following example:
+
+    </para>
+
+    <programlisting>
+       Program('hello.c', CPPPATH = ['include', '/home/project/inc'])
+    </programlisting>
+
+    <para>
+
+    Will look like this on POSIX or Linux:
+
+    </para>
+
+    <screen>
+       % <userinput>scons -Q hello</userinput>
+       cc -o hello.o -c -Iinclude -I/home/project/inc hello.c
+       cc -o hello hello.o
+    </screen>
+
+    <para>
+
+    And like this on Windows:
+
+    </para>
+
+    <screen>
+       C:\><userinput>scons -Q hello.exe</userinput>
+       cl /nologo /Iinclude /I\home\project\inc /c hello.c /Fohello.obj
+       link /nologo /OUT:hello.exe hello.obj
+    </screen>
+
+  </section>
+
+  <section>
+  <title>Caching Implicit Dependencies</title>
+
+    <para>
+
+    Scanning each file for <literal>#include</literal> lines
+    does take some extra processing time.
+    When you're doing a full build of a large system,
+    the scanning time is usually a very small percentage
+    of the overall time spent on the build.
+    You're most likely to notice the scanning time,
+    however, when you <emphasis>rebuild</emphasis>
+    all or part of a large system:
+    &SCons; will likely take some extra time to "think about"
+    what must be built before it issues the
+    first build command
+    (or decides that everything is up to date
+    and nothing must be rebuilt).
+
+ <!--
+ Isn't this expensive? The answer is, it depends. If you do a full build of a
+ large system, the scanning time is insignificant. If you do a rebuild of a
+ large system, then Cons will spend a fair amount of time thinking about it
+ before it decides that nothing has to be done (although not necessarily more
+ time than make!). The good news is that Cons makes it very easy to
+ intelligently subset your build, when you are working on localized changes.
+ -->
+
+    </para>
+
+    <para>
+
+    In practice, having &SCons; scan files saves time
+    relative to the amount of potential time
+    lost to tracking down subtle problems
+    introduced by incorrect dependencies.
+    Nevertheless, the "waiting time"
+    while &SCons; scans files can annoy
+    individual developers waiting for their builds to finish.
+    Consequently, &SCons; lets you cache
+    the implicit dependencies
+    that its scanners find,
+    for use by later builds.
+    You can do this by specifying the
+    &implicit-cache; option on the command line:
+
+    </para>
+
+    <screen>
+       % <userinput>scons -Q --implicit-cache hello</userinput>
+       cc -o hello.o -c hello.c
+       cc -o hello hello.o
+       % <userinput>scons -Q hello</userinput>
+       scons: `hello' is up to date.
+    </screen>
+
+    <para>
+
+    If you don't want to specify &implicit-cache;
+    on the command line each time,
+    you can make it the default behavior for your build
+    by setting the &implicit_cache; option
+    in an &SConscript; file:
+
+    </para>
+
+    <programlisting>
+       SetOption('implicit_cache', 1)
+    </programlisting>
+
+    <para>
+
+    &SCons; does not cache implicit dependencies like this by default
+    because the &implicit-cache; causes &SCons; to simply use the implicit
+    dependencies stored during the last run, without any checking
+    for whether or not those dependencies are still correct.
+    Specifically, this means &implicit-cache; instructs &SCons;
+    to <emphasis>not</emphasis> rebuild "correctly" in the
+    following cases:
+
+
+    </para>
+
+    <itemizedlist>
+
+      <listitem>
+        <para>
+
+        When &implicit-cache; is used, &SCons; will ignore any changes that
+        may have been made to search paths
+        (like &cv-CPPPATH; or &cv-LIBPATH;,).
+        This can lead to &SCons; not rebuilding a file if a change to
+        &cv-CPPPATH; would normally cause a different, same-named file from
+        a different directory to be used.
+
+        </para>
+      </listitem>
+
+      <listitem>
+        <para>
+
+        When &implicit-cache; is used, &SCons; will not detect if a
+        same-named file has been added to a directory that is earlier in
+        the search path than the directory in which the file was found
+        last time.
+
+        </para>
+      </listitem>
+
+    </itemizedlist>
+
+    <section>
+    <title>The &implicit-deps-changed; Option</title>
+
+      <para>
+
+      When using cached implicit dependencies,
+      sometimes you want to "start fresh"
+      and have &SCons; re-scan the files
+      for which it previously cached the dependencies.
+      For example,
+      if you have recently installed a new version of
+      external code that you use for compilation,
+      the external header files will have changed
+      and the previously-cached implicit dependencies
+      will be out of date.
+      You can update them by
+      running &SCons; with the &implicit-deps-changed; option:
+
+      </para>
+
+      <screen>
+         % <userinput>scons -Q --implicit-deps-changed hello</userinput>
+         cc -o hello.o -c hello.c
+         cc -o hello hello.o
+         % <userinput>scons -Q hello</userinput>
+         scons: `hello' is up to date.
+      </screen>
+
+      <para>
+
+      In this case, &SCons; will re-scan all of the implicit dependencies
+      and cache updated copies of the information.
+
+      </para>
+
+    </section>
+
+    <section>
+    <title>The &implicit-deps-unchanged; Option</title>
+
+      <para>
+
+      By default when caching dependencies,
+      &SCons; notices when a file has been modified
+      and re-scans the file for any updated
+      implicit dependency information.
+      Sometimes, however, you may want
+      to force &SCons; to use the cached implicit dependencies,
+      even if the source files changed.
+      This can speed up a build for example,
+      when you have changed your source files
+      but know that you haven't changed
+      any <literal>#include</literal> lines.
+      In this case,
+      you can use the &implicit-deps-unchanged; option:
+
+      </para>
+
+      <screen>
+         % <userinput>scons -Q --implicit-deps-unchanged hello</userinput>
+         cc -o hello.o -c hello.c
+         cc -o hello hello.o
+         % <userinput>scons -Q hello</userinput>
+         scons: `hello' is up to date.
+      </screen>
+
+      <para>
+
+      In this case,
+      &SCons; will assume that the cached implicit
+      dependencies are correct and
+      will not bother to re-scan changed files.
+      For typical builds after small,
+      incremental changes to source files,
+      the savings may not be very big,
+      but sometimes every bit of
+      improved performance counts.
+
+      </para>
+
+    </section>
+
+    <!--
+
+    <section>
+    <title>XXX max drift</title>
+
+      XXX SetOption('max_drift')
+
+    </section>
+
+    -->
+
+  </section>
+
+  <section>
+  <title>Ignoring Dependencies:  the &Ignore; Method</title>
+
+    <para>
+
+    Sometimes it makes sense
+    to not rebuild a program,
+    even if a dependency file changes.
+    In this case,
+    you would tell &SCons; specifically
+    to ignore a dependency as follows:
+
+    </para>
+
+    <programlisting>
+      hello = Program('hello.c')
+      Ignore(hello, 'hello.h')
+    </programlisting>
+
+    <!-- XXX mention that you can use arrays for target and source? -->
+
+    <!--
+    <scons_output example="ignore">
+      <scons_output_command>scons -Q hello</scons_output_command>
+      <scons_output_command>scons -Q hello</scons_output_command>
+      <scons_output_command output="    [CHANGE THE CONTENTS OF hello.h]">edit hello.h</scons_output_command>
+      <scons_output_command>scons -Q hello</scons_output_command>
+      XXX THIS EXAMPLE SHOULD BE UP-TO-DATE! XXX
+    </scons_output>
+    -->
+
+    <screen>
+      % <userinput>scons -Q hello</userinput>
+      cc -c -o hello.o hello.c
+      cc -o hello hello.o
+      % <userinput>scons -Q hello</userinput>
+      scons: `hello' is up to date.
+      % <userinput>edit hello.h</userinput>
+        [CHANGE THE CONTENTS OF hello.h]
+      % <userinput>scons -Q hello</userinput>
+      scons: `hello' is up to date.
+    </screen>
+
+    <para>
+
+    Now, the above example is a little contrived,
+    because it's hard to imagine a real-world situation
+    where you wouldn't to rebuild &hello;
+    if the &hello_h; file changed.
+    A more realistic example
+    might be if the &hello;
+    program is being built in a
+    directory that is shared between multiple systems
+    that have different copies of the
+    &stdio_h; include file.
+    In that case,
+    &SCons; would notice the differences between
+    the different systems' copies of &stdio_h;
+    and would rebuild &hello;
+    each time you change systems.
+    You could avoid these rebuilds as follows:
+
+    </para>
+
+    <programlisting>
+       hello = Program('hello.c')
+       Ignore(hello, '/usr/include/stdio.h')
+    </programlisting>
+
+  </section>
+
+  <section>
+  <title>Explicit Dependencies:  the &Depends; Method</title>
+
+    <para>
+
+    On the other hand,
+    sometimes a file depends on another file
+    that is not detected by an &SCons; scanner.
+    For this situation,
+    &SCons; allows you to specific explicitly that one file
+    depends on another file,
+    and must be rebuilt whenever that file changes.
+    This is specified using the &Depends; method:
+
+    </para>
+
+    <programlisting>
+       hello = Program('hello.c')
+       Depends(hello, 'other_file')
+    </programlisting>
+
+    <!-- XXX mention that you can use arrays for target and source? -->
+
+    <screen>
+       % <userinput>scons -Q hello</userinput>
+       cc -c hello.c -o hello.o
+       cc -o hello hello.o
+       % <userinput>scons -Q hello</userinput>
+       scons: `hello' is up to date.
+       % <userinput>edit other_file</userinput>
+           [CHANGE THE CONTENTS OF other_file]
+       % <userinput>scons -Q hello</userinput>
+       cc -c hello.c -o hello.o
+       cc -o hello hello.o
+    </screen>
+
+  </section>
+
+  <section>
+  <title>The &AlwaysBuild; Method</title>
+
+    <para>
+
+    How &SCons; handles dependencies can also be affected
+    by the &AlwaysBuild; method.
+    When a file is passed to the &AlwaysBuild; method,
+    like so:
+
+    </para>
+
+    <programlisting>
+      hello = Program('hello.c')
+      AlwaysBuild(hello)
+    </programlisting>
+
+    <para>
+
+    Then the specified target file (&hello; in our example)
+    will always be considered out-of-date and
+    rebuilt whenever that target file is evaluated
+    while walking the dependency graph:
+
+    </para>
+
+    <screen>
+      % <userinput>scons -Q</userinput>
+      cc -o hello.o -c hello.c
+      cc -o hello hello.o
+      % <userinput>scons -Q</userinput>
+      cc -o hello hello.o
+    </screen>
+
+    <para>
+
+    The &AlwaysBuild; function has a somewhat misleading name,
+    because it does not actually mean the target file will
+    be rebuilt every single time &SCons; is invoked.
+    Instead, it means that the target will, in fact,
+    be rebuilt whenever the target file is encountered
+    while evaluating the targets specified on
+    the command line (and their dependencies).
+    So specifying some other target on the command line,
+    a target that does <emphasis>not</emphasis>
+    itself depend on the &AlwaysBuild; target,
+    will still be rebuilt only if it's out-of-date
+    with respect to its dependencies:
+
+    </para>
+
+    <screen>
+      % <userinput>scons -Q</userinput>
+      cc -o hello.o -c hello.c
+      cc -o hello hello.o
+      % <userinput>scons -Q hello.o</userinput>
+      scons: `hello.o' is up to date.
+    </screen>
+
+    <!--
+
+      XXX AlwaysBuild() and Alias Nodes
+
+      XXX AlwaysBuild() and Dir Nodes
+
+      XXX AlwaysBuild() with no sources
+
+    -->
+
+  </section>
+
+  <!--
+
+  <section>
+  <title>The &Salt; Method</title>
+
+    <para>
+
+    XXX Salt() (are we going to implement this ?)
+
+        original Cons classic POD documentation:
+
+=head2 The C<Salt> method
+
+The C<Salt> method adds a constant value to the signature calculation
+for every derived file.  It is invoked as follows:
+
+  Salt $string;
+
+Changing the Salt value will force a complete rebuild of every derived
+file.  This can be used to force rebuilds in certain desired
+circumstances.  For example,
+
+  Salt `uname -s`;
+
+Would force a complete rebuild of every derived file whenever the
+operating system on which the build is performed (as reported by C<uname
+-s>) changes.
+
+    </para>
+
+  </section>
+
+  -->