doclib: Add extensive documentation about the documentation build process

doclib/SConscript

 # -*- python -*-
 
+# Some internal information on how the documentation is generated. All
+# this is quite a mess, the complete documentation generation setup is
+# in dire need of a complete redesign, but ...
+# 
+#
+# The documentation generation process is tightly integrated with the
+# scons build framework:
+# 
+# * SCons analyzes the Doxyfile's to find all the documentation
+#   dependencies. This happens in the doxygen builder in
+#   senfscons/Doxygen.py.
+#
+# * possibly the doclib/doxy-header.html and/or
+#   doclib/doxy-footer.html files are regenerated
+# 
+# * If any documentation is out-of-date with respect to it's source
+#   files, the documentation is regenerated.
+# 
+# * When building the complete documentation ('all_docs'), the
+#   cross-reference page will be rebuilt if necessary (The global list
+#   of 'Open Issues').
+#
+# * To fix some link errors, the additional 'linlint' and 'fixlinks'
+#   targets are used
+#
+# 
+# 1. Scanning the Doxyfile's
+# 
+# The doxygen builder scans all documentation source files which have
+# the text 'doxyfile' in any case within them. It understands @INCLUDE
+# directives and will find all the dependencies of the documentation:
+# 
+# * All the source files as selected by INPUT, INPUT_PATTERN,
+#   RECURSIVE and so on.
+# 
+# * Any referenced tag-files
+# 
+# * Documentation header and/or footer
+# 
+# * The INPUT_FILTER program
+# 
+# * Any included doxygen configuration files
+# 
+#
+# 2. Regenerating header and/or footer
+#
+# If needed, the doxy-header.html and/or doxy-footer.html file will be
+# regenerated. The dependencies are *not* complete, just adding a new
+# subdirectory sadly does not automatically update the header (which
+# contains the menu)
+#
+# The header and/or footer are written are generated from templates
+# using a simple python based templating system called yaptu which is
+# included in doclib/.
+#
+# 
+# 3. Calling doxygen
+# 
+# The doxygen call itself is quite complex since there is some pre-
+# and post-processing going on. We can separate this step into tree
+# parts
+#
+# * Building prerequisites (e.g. images)
+#
+# * The processing done by the Doxygen builder
+#
+# * Additional processing added by the SENFSCons.Doxygen helper.
+#
+#
+# 3.1. Building prerequisites
+#
+# The prerequisites are images referenced by the documentation. These
+# images are mostly generated using the Dia2Png builder.
+#
+#
+# 3.2. The main doxygen build (Doxygen builder)
+#
+# * doxygen proper is called
+# 
+# * doxygen is configured in Doxyfile.global to call
+#   'doclib/filter.pl' on each source file. This filter will strip
+#   excess whitespace from the beginning of lines in '\code' and
+#   '<pre>' blocks. Additionally it will expand all tabs, tab width is
+#   8 spaces (there should be no tabs in the source but ...)
+# 
+# * doxygen is configured in Doxyfile.global to call 'doclib/dot' to
+#   generate the 'dot' images.
+#
+# * 'doclib/dot' calls 'doclib/dot-munge.pl' on the .dot
+#    files. dot-munge.pl changes the font and font-size and adds
+#    line-breaks to long labels
+#
+# * 'doclib/dot' calls the real dot binary. If the resulting image is
+#   more than 800 pixels wide, dot is called again, this time using
+#   the oposite rang direction (top-bottom vs. left-right). Then the
+#   image with the smaller width is selected and returned.
+#
+# * after doxygen is finished, the list of referenced tag-files is
+#   checked. For each tag file the directory is found, where the
+#   documentation is generated (by scanning the Doxyfile which is
+#   repsonsible for building the tag file). For every tag file, the
+#   correct 'insstalldox' command is generated
+#
+# * The stamp files are created
+#
+#
+# 3.3. Postprocessing
+#
+# The following steps are mostly added to work around some stupid
+# doxygen problems
+#
+# * If a tag file is generated, 'senfscons/tagmunge.xsl' is called on the
+#   tag file. This XSLT stylesheet removes all namespace components
+#   from the tag file. Without this task, doxygen will completely barf
+#   when two different documentation parts have members in the same
+#   namespace.
+#
+# * All html files are processed by 'doclib/html-munge.xsl'. However,
+#   since the documentation generated by doxygen is completely broken
+#   we need to preprocess the html files with a simple 'sed' script
+#   and 'tidy' before 'xsltproc' even accepts the html code.
+#
+# * We use the generated xml output of doxygen to generate an XML
+#   fragment for the global cross reference. This fragment is
+#   generated by 'senfscons/xrefxtract.xslt'
+#
+#
+# 4. Building the global cross-reference
+#
+# The global cross reference is built from the cross-refernce
+# fragments generated for each of the documentation parts and from the
+# doxy-header.html/doxy-footer.html files. This conversion is
+# controlled by the 'senfscons/xrefhtml.xslt' stylesheet.
+#
+#
+# 5. Fixing broken links
+#
+# After the documentation has been generated, additional calls first
+# to the 'linklint' and then to the 'fixlinks' target will try to fix
+# broken links generated by doxygen. First, 'linklint' will call the
+# linklint tool to check for broken links in the documentation.
+#
+# 'fixlinks' is then called which calls 'doclib/fixlinks.py' which
+# scans *all* html files, builds an index of all (unique) anchors and
+# then fixes the url part of all links with correct anchor but bad
+# file name.
+#
+
+
 Import('env')
 import SENFSCons
 

senf.dict

@@ -73,6 +73,8 @@ dl
 DNS
 dontinclude
 dox
+doxyfile
+doxygen
 DSMCCSection
 dt
 ElementParser
@@ -279,6 +281,7 @@ PriorityJoin
 protocolbundle
 protocolbundles
 protocolType
+py
 QueueingDiscipline
 queueSize
 RateFilter
@@ -306,6 +309,7 @@ ScopeId
 screenshot
 sec
 senf
+senfscons
 ServerSocketHandle
 setBegin
 setEnd