[BACK]Return to readme.html CVS log [TXT] [DIR] Up to [LON-CAPA] / loncom / build

Diff for /loncom/build/readme.html between versions 1.5 and 1.12

version 1.5, 2001/01/17 11:11:42 version 1.12, 2001/01/17 13:04:59
Line 1 Line 1
   <HTML>
   <HEAD>
   <META NAME="GENERATOR" CONTENT="Scott Harrison and Emacs Version 3.14159265358979">
   <TITLE>LON-CAPA Software Developer Instructions</TITLE>
   </HEAD>
   <BODY>
 <H1>LON-CAPA Software Developer Instructions</H1>  <H1>LON-CAPA Software Developer Instructions</H1>
   <BR><I>Written by Scott Harrison, January 17, 2001</I>
   <BR><I>Last updated, January 17, 2001</I>
 <OL>  <OL>
 <LI><A HREF="#Using_CVS">Using CVS</A>  <LI><A HREF="#Using_CVS">Using CVS</A>
 <UL>  <UL>
Line 45 Line 52
 <FONT COLOR="#008800">  <FONT COLOR="#008800">
 <PRE>  <PRE>
 cd loncom/build  cd loncom/build
 rm -Rf HTML  rm -Rf HTML <I>(or alternatively, "make clean")</I>
 make HTML  make HTML
 cd HTML  cd HTML
 <I>(look at the index.html file with a web browser such as Netscape)</I>  <I>(look at the index.html file with a web browser such as Netscape)</I>
 </PRE>  </PRE>
 </FONT>  </FONT>
   <STRONG>General description of what happens</STRONG>
   <P>
   This is the actual make target code.
   <FONT COLOR="#880000">
   <PRE>
   <!-- LONCAPA MAKETARGET=HTML START -->
   HTML:
           install -d HTML
           cp ../../doc/loncapafiles/*.gif HTML
           perl parse.pl ../../doc/loncapafiles/loncapafiles.html HTML > HTML/index.html
   <!-- LONCAPA MAKETARGET=HTML END -->
   </PRE>
   </FONT>
   What basically happens is that specially marked-up data in the LON-CAPA
   cvs repository file <TT>doc/loncapafiles.html</TT> is parsed into a more
   viewable format by <TT>loncom/build/parse.pl</TT>.  The resulting
   file gives a very well organized view of all the files, directories,
   links, ownerships, permissions, and brief documentation of what each
   file does.
   </P>
 <LI><A NAME="makebuild">  <LI><A NAME="makebuild">
     <H2>Compiling the software (make build)</H2>      <H2>Compiling the software (make build)</H2>
   <STRONG>Commands</STRONG>
   <FONT COLOR="#008800">
   <PRE>
   cd loncom/build
   make build
   </PRE>
   </FONT>
   <STRONG>General description of what happens</STRONG>
   <P>
   This is the actual make target code.
   <FONT COLOR="#880000">
   <PRE>
   <!-- LONCAPA MAKETARGET=HTML START -->
   build:
           perl parse.pl ../../doc/loncapafiles/loncapafiles.html build > Makefile.build
           make -f Makefile.build all
   <!-- LONCAPA MAKETARGET=HTML END -->
   </PRE>
   </FONT>
   <TT>loncom/build/parse.pl</TT> reads in all the build information out
   of <TT>doc/loncapafiles/loncapafiles.html</TT>.  A new Makefile named
   <TT>loncom/build/Makefile.build</TT> is dynamically constructed.
   This dynamically generated Makefile is then run to build/compile
   all the software targets from source.  This currently takes 10 minutes
   (depends on the speed of the machine you compile with).
   </P>
   <STRONG>Example</STRONG>
   <P>
   Here is information for one file <TT>tth.so</TT> provided in
   <TT>doc/loncapafiles/loncapafiles.html</TT>.
   <FONT COLOR="#330066">
   <PRE>
   <BR>&lt;METAGROUP&gt;
   <BR>&lt;LONCAPA TYPE=LOCATION DIST="redhat6.2" SOURCE="loncom/modules/TexConvert/tthperl/tth.so" TARGET="usr/lib/perl5/site_perl/5.005/tth.so" CATEGORY="system file"&gt;
   <BR>&lt;DESCRIPTION&gt;
   <BR>shared library file for dynamic loading and unloading of TeX-to-HTML functionality
   <BR>&lt;/DESCRIPTION&gt;
   <BR>&lt;BUILD&gt;
   <BR>loncom/modules/TexConvert/tthperl/commands
   <BR>&lt;/BUILD&gt;
   <BR>&lt;DEPENDENCIES&gt;
   <BR>../tthdynamic/tthfunc.c
   <BR>../ttmdynamic/ttmfunc.c
   <BR>&lt;/DEPENDENCIES&gt;
   </PRE>
   </FONT>
   <TT>loncom/build/parse.pl</TT> sees the <B>BUILD</B> tags and sets up
   a dynamic file <TT>Makefile.build</TT> to run the command inside the
   <B>BUILD</B> tags (currently, <B>DEPENDENCIES</B> is not used for anything
   besides documentation).
   </P>
   <P>
   Here is an example of a dynamically generated <TT>Makefile.build</TT>
   that builds two LON-CAPA files (one of which is <TT>tth.so</TT>).
   <FONT COLOR="#330066">
   <PRE>
   all: ../homework/caparesponse/capa.so ../modules/TexConvert/tthperl/tth.so 
   
   ../homework/caparesponse/capa.so:  ../homework/caparesponse/caparesponse.c ../ho
   mework/caparesponse/caparesponse.pm alwaysrun
           cd ../homework/caparesponse/; sh ./commands
   
   ../modules/TexConvert/tthperl/tth.so:  ../modules/TexConvert/tthperl/../tthdynam
   ic/tthfunc.c ../modules/TexConvert/tthperl/../ttmdynamic/ttmfunc.c
           cd ../modules/TexConvert/tthperl/; sh ./commands
   
   alwaysrun:
   </PRE>
   </FONT>
   </P>
 <LI><A NAME="loncapafiles">  <LI><A NAME="loncapafiles">
     <H2>Adding/removing files from the LON-CAPA installation (doc/loncapafiles/loncapafiles.html)</H2>      <H2>Adding/removing files from the LON-CAPA installation (doc/loncapafiles/loncapafiles.html)</H2>
   <STRONG>To add and remove (and alter)</STRONG>
   <P>
   All that you have to do to alter the behavior of the installation is
   edit a single file (<TT>doc/loncapafiles/loncapafiles.html</TT>).
   Adding, removing, and altering files requires proper attention
   to the syntax of file format of course.
   </P>
   <STRONG>File Format</STRONG>
   <P>
   The preceding <A HREF=#"makebuild">"make build"</A> documentation
   gives an example <B>METAGROUP</B> entry describing one particular file.
   All data within <TT>loncapafiles.html</TT> is specified according
   to markup tags.  The format and syntax of <TT>loncapafiles.html</TT>
   is currently best described by the HTML documentation code at the beginning of
   loncapafiles.html (as well as, by example, seeing how various
   information is coded).  All in all, the syntax is quite simple.
   </P>
   <STRONG>Philosophy and notes (the thing nobody reads)</STRONG>
   <P>
   Packaging the software from CVS onto a machine file system requires many
   things:
   <UL>
   <LI>documenting every component of the software,
   <LI>handling CVS <U>source</U> to file system <U>target</U> information
   <LI>handling (according to a hierarchical scheme of grouping) file
   ownership and permissions,
   <LI>handling (according to a hierarchical scheme of grouping) directory
   ownership and permissions,
   <LI>handling symbolic links
   <LI>providing for multiple options of installation targets
   (RedHat versus Debian for instance),
   <LI>providing for different file ownerships and permissions to apply
   to the same file,
   <LI>allowing system software documentation to be automatically generated
   (see information on <A HREF="#makeHTML">"make html"</A>),
   <LI>providing information in an easily adjustable form as new demands
   are made on the software packaging system,
   <LI>providing software package information (for RPM),
   <LI>having information in a format that allows for troubleshooting
   the current status of the machine file system,
   <LI>allow for changes to the structure of the CVS repository,
   <LI>and something that is simple enough for any one to immediately work with,
   without having to learn specifics (or hidden traps) of complicated Makefile's
   or a new macro language (m4?).
   </UL>
   </P>
   <P>
   I looked into, and tried, different ways of accomplishing the above
   including automake and recursive make.  The automake system seemed quite
   complicated (and needlessly so in terms of this project since, by and large,
   it works to coordinate many different types of build/compilation parameters
   whereas we are more concerned with installation parameters).  Recursive make
   has significant deficiencies in the sense that not all the information
   is kept in one place, and there are significant levels of dependency
   between all the things that must be done to keep software packaging
   up to date.  A particularly convincing article I found when looking into
   much of this was
   <A HREF="http://www.pcug.org.au/~millerp/rmch/recu-make-cons-harm.html">
   "Recursive Make Considered Harmful" by Peter Miller</A>.  Complicating
   matters was, at the time, it was unclear as to what categories
   of software files we had, and whether or not the directory structure
   of CVS would remain constant.  With an ever-developing directory structure
   to CVS, I preferred to organize the information on a per-file basis
   as opposed to a per-directory basis (although there is a successful
   implementation of a standard big Makefile in <TT>loncom/Makefile</TT>).
   Additionally, a standard big Makefile assumes certain "normalcy" to
   the directory structure of different potential operating system directories
   (RedHat vs. Debian).
   </P>
   <P>
   If you take time to look at loncapafiles.html
   (and perhaps run the <A HREF="#makeHTML">make HTML</A> command)
   you will find that the organizing information according to the markup
   syntax in <TT>loncapafiles.html</TT> is simple.  Simple is good.
   </P>
   <P>
   <TT>loncom/build/parse.pl</TT> is the script (invoked automatically
   by the various targets in <TT>loncom/build/Makefile</TT>) that reads
   <TT>doc/loncapafiles/loncapafiles.html</TT>.  <TT>parse.pl</TT>
   is capable of reading and returning different types of information
   from <TT>loncapafiles.html</TT> depending on how <TT>parse.pl</TT>
   is invoked.  <TT>parse.pl</TT> has yet to have introduced new sources
   of error, and has been tested in quite a number of ways.  As with
   any parser however, I remain paranoid.
   </P>
   <P>
   My regrets with the current system is that <TT>parse.pl</TT> is
   slow (can take 1 minute to run) and includes a few tidbits of code,
   specific to the make process, that probably should be in
   <TT>loncom/build/Makefile</TT>.  Additionally, <TT>loncapafiles.html</TT>
   should have a DTD and all those other good SGML-ish things (and parsing
   should be done with a real SGML-derived parser).
   </P>
   <P>
   On the plus side, the <TT>parse.pl</TT>-<TT>loncapafiles.html</TT> 
   combination has been working very efficiently and error-free.
   </P>
 <LI><A NAME="configversusnonconfig">  <LI><A NAME="configversusnonconfig">
     <H2>Configurable files versus non-configurable files</H2>      <H2>Configurable files versus non-configurable files</H2>
   <STRONG>Machine-specific information is the difference</STRONG>
   <P>
   The current list of configurable files for the LON-CAPA system is
   /etc/httpd/access.conf.
   All of these configurable files contain machine-specific information.
   For instance, the LON-CAPA system relies on unique host IDs such
   as msua3, s1, s2, msul1, and 103a1 (specified as a "PerlSetVar lonHostID"
   field within /etc/httpd/access.conf).
   Non-configurable files simply do NOT have machine-specific information.
   <STRONG>The impact on updating software</STRONG>
   <P>
   What this means in terms of software updating is that
   <UL>
   <LI>non-configurable files can be simply overwritten with newer versions
   (without "anything" else to worry about),
   <LI>and configurable files must follow these steps to be safely overwritten
   <OL>
   <LI>have their machine specific information saved,
   <LI>be overwritten, and then
   <LI>have their machine specific information restored.
   </UL>
   </P>
 <LI><A NAME="makeinstall">  <LI><A NAME="makeinstall">
     <H2>Updating the non-configurable files on your machine (make install)</H2>      <H2>Updating the non-configurable files on your machine (make install)</H2>
   
   
 <LI><A NAME="makeconfiginstall">  <LI><A NAME="makeconfiginstall">
     <H2>Updating the configurable files on your machine (make configinstall)</H2>      <H2>Updating the configurable files on your machine (make configinstall)</H2>
   
   
 <LI><A NAME="makeRPM">  <LI><A NAME="makeRPM">
     <H2>Building RPMs (make RPM)</H2>      <H2>Building RPMs (make RPM)</H2>
 </OL>  </OL>
   </BODY>
   </HTML>
   
   
   
   
Removed from v.1.5  
changed lines
  Added in v.1.12

FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>