<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN" "http://www.w3.org/TR/REC-html40/loose.dtd"> <HTML> <HEAD> <META http-equiv="Content-Type" content="text/html; charset=US-ASCII"> <META name="GENERATOR" content="hevea 1.10"> <base target="main"> <script language="JavaScript"> <!-- Begin function loadTop(url) { parent.location.href= url; } // --> </script> <LINK rel="stylesheet" type="text/css" href="cil.css"> <TITLE>The CIL Driver</TITLE> </HEAD> <BODY > <A HREF="attributes.html"><IMG SRC="previous_motif.gif" ALT="Previous"></A> <A HREF="ciltoc.html"><IMG SRC="contents_motif.gif" ALT="Up"></A> <A HREF="ext.html"><IMG SRC="next_motif.gif" ALT="Next"></A> <HR> <H2 CLASS="section"><A NAME="htoc15">7</A>  The CIL Driver</H2><P><A NAME="sec-driver"></A></P><P>We have packaged CIL as an application <TT>cilly</TT> that contains certain example modules, such as <TT>logwrites.ml</TT> (a module that instruments code to print the addresses of memory locations being written). Normally, you write another module like that, add command-line options and an invocation of your module in <TT>src/main.ml</TT>. Once you compile CIL you will obtain the file <TT>obj/cilly.asm.exe</TT>. </P><P>We wrote a driver for this executable that makes it easy to invoke your analysis on existing C code with very little manual intervention. This driver is <TT>bin/cilly</TT> and is quite powerful. Note that the <TT>cilly</TT> script is configured during installation with the path where CIL resides. This means that you can move it to any place you want. </P><P>A simple use of the driver is: </P><PRE CLASS="verbatim">bin/cilly --save-temps -D HAPPY_MOOD -I myincludes hello.c -o hello </PRE><P><FONT COLOR=blue>--save-temps</FONT> tells CIL to save the resulting output files in the current directory. Otherwise, they’ll be put in <TT>/tmp</TT> and deleted automatically. Not that this is the only CIL-specific flag in the list – the other flags use <TT>gcc</TT>’s syntax.</P><P>This performs the following actions: </P><UL CLASS="itemize"><LI CLASS="li-itemize"> preprocessing using the -D and -I arguments with the resulting file left in <TT>hello.i</TT>, </LI><LI CLASS="li-itemize">the invocation of the <TT>cilly.asm</TT> application which parses <TT>hello.i</TT> converts it to CIL and the pretty-prints it to <TT>hello.cil.c</TT> </LI><LI CLASS="li-itemize">another round of preprocessing with the result placed in <TT>hello.cil.i</TT> </LI><LI CLASS="li-itemize">the true compilation with the result in <TT>hello.cil.o</TT> </LI><LI CLASS="li-itemize">a linking phase with the result in <TT>hello</TT> </LI></UL><P>Note that <TT>cilly</TT> behaves like the <TT>gcc</TT> compiler. This makes it easy to use it with existing <TT>Makefiles</TT>: </P><PRE CLASS="verbatim">make CC="bin/cilly" LD="bin/cilly" </PRE><P><TT>cilly</TT> can also behave as the Microsoft Visual C compiler, if the first argument is <TT>--mode=MSVC</TT>: </P><PRE CLASS="verbatim">bin/cilly --mode=MSVC /D HAPPY_MOOD /I myincludes hello.c /Fe hello.exe </PRE><P>(This in turn will pass a <TT>--MSVC</TT> flag to the underlying <TT>cilly.asm</TT> process which will make it understand the Microsoft Visual C extensions)</P><P><TT>cilly</TT> can also behave as the archiver <TT>ar</TT>, if it is passed an argument <TT>--mode=AR</TT>. Note that only the <TT>cr</TT> mode is supported (create a new archive and replace all files in there). Therefore the previous version of the archive is lost. You will also need to remove any other commands that operate on the generated library (e.g. <TT>ranlib</TT>, <TT>lorder</TT>), as the .a file is no longer an actual binary library.</P><P>Furthermore, <TT>cilly</TT> allows you to pass some arguments on to the underlying <TT>cilly.asm</TT> process. As a general rule all arguments that start with <TT>--</TT> and that <TT>cilly</TT> itself does not process, are passed on. For example, </P><PRE CLASS="verbatim">bin/cilly --dologwrites -D HAPPY_MOOD -I myincludes hello.c -o hello.exe </PRE><P>will produce a file <TT>hello.cil.c</TT> that prints all the memory addresses written by the application. </P><P>The most powerful feature of <TT>cilly</TT> is that it can collect all the sources in your project, merge them into one file and then apply CIL. This makes it a breeze to do whole-program analysis and transformation. All you have to do is to pass the <TT>--merge</TT> flag to <TT>cilly</TT>: </P><PRE CLASS="verbatim">make CC="bin/cilly --save-temps --dologwrites --merge" </PRE><P>You can even leave some files untouched: </P><PRE CLASS="verbatim">make CC="bin/cilly --save-temps --dologwrites --merge --leavealone=foo --leavealone=bar" </PRE><P>This will merge all the files except those with the basename <TT>foo</TT> and <TT>bar</TT>. Those files will be compiled as usual and then linked in at the very end. </P><P>The sequence of actions performed by <TT>cilly</TT> depends on whether merging is turned on or not: </P><UL CLASS="itemize"><LI CLASS="li-itemize"> If merging is off <OL CLASS="enumerate" type=1><LI CLASS="li-enumerate"> For every file <TT>file.c</TT> to compile <OL CLASS="enumerate" type=a><LI CLASS="li-enumerate"> Preprocess the file with the given arguments to produce <TT>file.i</TT> </LI><LI CLASS="li-enumerate">Invoke <TT>cilly.asm</TT> to produce a <TT>file.cil.c</TT> </LI><LI CLASS="li-enumerate">Preprocess to <TT>file.cil.i</TT> </LI><LI CLASS="li-enumerate">Invoke the underlying compiler to produce <TT>file.cil.o</TT> </LI></OL> </LI><LI CLASS="li-enumerate">Link the resulting objects </LI></OL> </LI><LI CLASS="li-itemize">If merging is on <OL CLASS="enumerate" type=1><LI CLASS="li-enumerate"> For every file <TT>file.c</TT> to compile <OL CLASS="enumerate" type=a><LI CLASS="li-enumerate"> Preprocess the file with the given arguments to produce <TT>file.i</TT> </LI><LI CLASS="li-enumerate">Save the preprocessed source as <TT>file.o</TT> </LI></OL> </LI><LI CLASS="li-enumerate">When linking executable <TT>hello.exe</TT>, look at every object file that must be linked and see if it actually contains preprocessed source. Pass all those files to a special merging application (described in Section <A HREF="merger.html#sec-merger">13</A>) to produce <TT>hello.exe_comb.c</TT> </LI><LI CLASS="li-enumerate">Invoke <TT>cilly.asm</TT> to produce a <TT>hello.exe_comb.cil.c</TT> </LI><LI CLASS="li-enumerate">Preprocess to <TT>hello.exe_comb.cil.i</TT> </LI><LI CLASS="li-enumerate">Invoke the underlying compiler to produce <TT>hello.exe_comb.cil.o</TT> </LI><LI CLASS="li-enumerate">Invoke the actual linker to produce <TT>hello.exe</TT> </LI></OL> </LI></UL><P>Note that files that you specify with <TT>--leavealone</TT> are not merged and never presented to CIL. They are compiled as usual and then are linked in at the end. </P><P>And a final feature of <TT>cilly</TT> is that it can substitute copies of the system’s include files:</P><PRE CLASS="verbatim">make CC="bin/cilly --includedir=myinclude" </PRE><P>This will force the preprocessor to use the file <TT>myinclude/xxx/stdio.h</TT> (if it exists) whenever it encounters <TT>#include <stdio.h></TT>. The <TT>xxx</TT> is a string that identifies the compiler version you are using. This modified include files should be produced with the patcher script (see Section <A HREF="patcher.html#sec-patcher">14</A>).</P><H3 CLASS="subsection"><A NAME="toc8"></A><A NAME="htoc16">7.1</A>  <TT>cilly</TT> Options</H3><P>Among the options for the <TT>cilly</TT> you can put anything that can normally go in the command line of the compiler that <TT>cilly</TT> is impersonating. <TT>cilly</TT> will do its best to pass those options along to the appropriate subprocess. In addition, the following options are supported (a complete and up-to-date list can always be obtained by running <TT>cilly --help</TT>):</P><UL CLASS="itemize"><LI CLASS="li-itemize"> <TT>--gcc=command</TT> Tell <TT>cilly</TT> to use <TT>command</TT> to invoke gcc, e.g. <TT>--gcc=arm-elf-gcc</TT> to use a cross-compiler. See also the <TT>--envmachine</TT> option below that tells CIL to assume a different machine model. </LI><LI CLASS="li-itemize"><TT>--mode=mode</TT> This must be the first argument if present. It makes <TT>cilly</TT> behave as a given compiled. The following modes are recognized: <UL CLASS="itemize"><LI CLASS="li-itemize"> GNUCC - the GNU C Compiler. This is the default. </LI><LI CLASS="li-itemize">MSVC - the Microsoft Visual C compiler. Of course, you should pass only MSVC valid options in this case. </LI><LI CLASS="li-itemize">AR - the archiver <TT>ar</TT>. Only the mode <TT>cr</TT> is supported and the original version of the archive is lost. </LI></UL> </LI><LI CLASS="li-itemize"><TT>--help</TT> Prints a list of the options supported. </LI><LI CLASS="li-itemize"><TT>--verbose</TT> Prints lots of messages about what is going on. </LI><LI CLASS="li-itemize"><TT>--stages</TT> Less than <TT>--verbose</TT> but lets you see what <TT>cilly</TT> is doing. </LI><LI CLASS="li-itemize"><TT>--merge</TT> This tells <TT>cilly</TT> to first attempt to collect into one source file all of the sources that make your application, and then to apply <TT>cilly.asm</TT> on the resulting source. The sequence of actions in this case is described above and the merger itself is described in Section <A HREF="merger.html#sec-merger">13</A>.</LI><LI CLASS="li-itemize"><TT>--leavealone=xxx</TT>. Do not merge and do not present to CIL the files whose basename is "xxx". These files are compiled as usual and linked in at the end. </LI><LI CLASS="li-itemize"><TT>--includedir=xxx</TT>. Override the include files with those in the given directory. The given directory is the same name that was given an an argument to the patcher (see Section <A HREF="patcher.html#sec-patcher">14</A>). In particular this means that that directory contains subdirectories named based on the current compiler version. The patcher creates those directories. </LI><LI CLASS="li-itemize"><TT>--usecabs</TT>. Do not CIL, but instead just parse the source and print its AST out. This should looked like the preprocessed file. This is useful when you suspect that the conversion to CIL phase changes the meaning of the program. </LI><LI CLASS="li-itemize"><TT>--save-temps=xxx</TT>. Temporary files are preserved in the xxx directory. For example, the output of CIL will be put in a file named <TT>*.cil.c</TT>. </LI><LI CLASS="li-itemize"><TT>--save-temps</TT>. Temporay files are preserved in the current directory. </LI></UL><H3 CLASS="subsection"><A NAME="toc9"></A><A NAME="htoc17">7.2</A>  <TT>cilly.asm</TT> Options</H3><P> <A NAME="sec-cilly-asm-options"></A></P><P>All of the options that start with <TT>--</TT> and are not understood by <TT>cilly</TT> are passed on to <TT>cilly.asm</TT>. <TT>cilly</TT> also passes along to <TT>cilly.asm</TT> flags such as <TT>--MSVC</TT> that both need to know about. The following options are supported. Many of these flags also have corresponding “<TT>--no</TT>*” versions if you need to go back to the default, as in “<TT>--nowarnall</TT>”.</P><P>       <B>General Options:</B> </P><UL CLASS="itemize"><LI CLASS="li-itemize"> <TT>--version</TT> output version information and exit </LI><LI CLASS="li-itemize"><TT>--verbose</TT> Print lots of random stuff. This is passed on from cilly </LI><LI CLASS="li-itemize"><TT>--warnall</TT> Show all warnings. </LI><LI CLASS="li-itemize"><TT>--debug=xxx</TT> turns on debugging flag xxx </LI><LI CLASS="li-itemize"><TT>--nodebug=xxx</TT> turns off debugging flag xxx </LI><LI CLASS="li-itemize"><TT>--flush</TT> Flush the output streams often (aids debugging). </LI><LI CLASS="li-itemize"><TT>--check</TT> Run a consistency check over the CIL after every operation. </LI><LI CLASS="li-itemize"><TT>--strictcheck</TT> Same as <TT>--check</TT>, but it treats consistency problems as errors instead of warnings. </LI><LI CLASS="li-itemize"><TT>--nocheck</TT> turns off consistency checking of CIL. </LI><LI CLASS="li-itemize"><TT>--noPrintLn</TT> Don’t output #line directives in the output. </LI><LI CLASS="li-itemize"><TT>--commPrintLn</TT> Print #line directives in the output, but put them in comments. </LI><LI CLASS="li-itemize"><TT>--commPrintLnSparse</TT> Like <TT>--commPrintLn</TT> but print only new line numbers. </LI><LI CLASS="li-itemize"><TT>--log=xxx</TT> Set the name of the log file. By default stderr is used </LI><LI CLASS="li-itemize"><TT>--MSVC</TT> Enable MSVC compatibility. Default is GNU. </LI><LI CLASS="li-itemize"><TT>--ignore-merge-conflicts</TT> ignore merging conflicts. </LI><LI CLASS="li-itemize"><TT>--extrafiles=filename</TT>: the name of a file that contains a list of additional files to process, separated by whitespace. </LI><LI CLASS="li-itemize"><TT>--stats</TT> Print statistics about the running time of the parser, conversion to CIL, etc. Also prints memory-usage statistics. You can time parts of your own code as well. Calling (<TT>Stats.time ‘‘label’’ func arg</TT>) will evaluate <TT>(func arg)</TT> and remember how long this takes. If you call <TT>Stats.time</TT> repeatedly with the same label, CIL will report the aggregate time.<P>If available, CIL uses the x86 performance counters for these stats. This is very precise, but results in “wall-clock time.” To report only user-mode time, find the call to <TT>Stats.reset</TT> in <TT>main.ml</TT>, and change it to <TT>Stats.reset Stats.SoftwareTimer</TT>.</P></LI><LI CLASS="li-itemize"><TT>--envmachine</TT>. Use machine model specified in <TT>CIL_MACHINE</TT> environment variable, rather than the one compiled into CIL. Note that you should not pass gcc’s 32/64-bit <TT>-m32</TT> and <TT>-m64</TT> options to <TT>cilly</TT> if you use <TT>--envmachine</TT> (they use <TT>--envmachine</TT> under the hood). See Section <A HREF="#sec-cilmachine">7.4</A> for a description of <TT>CIL_MACHINE</TT>’s format.<P><B>Lowering Options</B> </P></LI><LI CLASS="li-itemize"><TT>--noLowerConstants</TT> do not lower constant expressions. </LI><LI CLASS="li-itemize"><TT>--noInsertImplicitCasts</TT> do not insert implicit casts. </LI><LI CLASS="li-itemize"><TT>--forceRLArgEval</TT> Forces right to left evaluation of function arguments. </LI><LI CLASS="li-itemize"><TT>--disallowDuplication</TT> Prevent small chunks of code from being duplicated. </LI><LI CLASS="li-itemize"><TT>--keepunused</TT> Do not remove the unused variables and types. </LI><LI CLASS="li-itemize"><TT>--rmUnusedInlines</TT> Delete any unused inline functions. This is the default in MSVC mode.<P><B>Output Options:</B> </P></LI><LI CLASS="li-itemize"><TT>--printCilAsIs</TT> Do not try to simplify the CIL when printing. Without this flag, CIL will attempt to produce prettier output by e.g. changing <TT>while(1)</TT> into more meaningful loops. </LI><LI CLASS="li-itemize"><TT>--noWrap</TT> do not wrap long lines when printing </LI><LI CLASS="li-itemize"><TT>--out=xxx</TT> the name of the output CIL file. <TT>cilly</TT> sets this for you. </LI><LI CLASS="li-itemize"><TT>--mergedout=xxx</TT> specify the name of the merged file </LI><LI CLASS="li-itemize"><TT>--cabsonly=xxx</TT> CABS output file name <P><B>Selected features.</B> See Section <A HREF="ext.html#sec-Extension">8</A> for more information. </P></LI><LI CLASS="li-itemize"><TT>--dologcalls</TT>. Insert code in the processed source to print the name of functions as are called. Implemented in <TT>src/ext/logcalls.ml</TT>. </LI><LI CLASS="li-itemize"><TT>--dologwrites</TT>. Insert code in the processed source to print the address of all memory writes. Implemented in <TT>src/ext/logwrites.ml</TT>. </LI><LI CLASS="li-itemize"><TT>--dooneRet</TT>. Make each function have at most one ’return’. Implemented in <TT>src/ext/oneret.ml</TT>. </LI><LI CLASS="li-itemize"><TT>--dostackGuard</TT>. Instrument function calls and returns to maintain a separate stack for return addresses. Implemeted in <TT>src/ext/heapify.ml</TT>. </LI><LI CLASS="li-itemize"><TT>--domakeCFG</TT>. Make the program look more like a CFG. Implemented in <TT>src/cil.ml</TT>. </LI><LI CLASS="li-itemize"><TT>--dopartial</TT>. Do interprocedural partial evaluation and constant folding. Implemented in <TT>src/ext/partial.ml</TT>. </LI><LI CLASS="li-itemize"><TT>--dosimpleMem</TT>. Simplify all memory expressions. Implemented in <TT>src/ext/simplemem.ml</TT>. <P>For an up-to-date list of available options, run <TT>cilly.asm --help</TT>. </P></LI></UL><H3 CLASS="subsection"><A NAME="toc10"></A><A NAME="htoc18">7.3</A>  Internal Options</H3><P> <A NAME="sec-cilly-internal-options"></A></P><P>All of the <TT>cilly.asm</TT> options described above can be set programmatically – see <TT>src/ciloptions.ml</TT> or the individual extensions to see how. Some options should be set before parsing to be effective. </P><P>Additionally, a few CIL options have no command-line flag and can only be set programmatically. These options may be useful for certain analyses:</P><UL CLASS="itemize"><LI CLASS="li-itemize"> <TT>Cabs2cil.doCollapseCallCast</TT>:This is false by default. Set to true to replicate the behavior of CIL 1.3.5 and earlier.<P>When false, all casts in the program are made explicit using the <TT>CastE</TT> expression. Accordingly, the destination of a Call instruction will always have the same type as the function’s return type.</P><P>If true, the destination type of a Call may differ from the return type, so there is an implicit cast. This is useful for analyses involving <TT>malloc</TT>. Without this option, CIL converts “<TT>T* x = malloc(n);</TT>” into “<TT>void* tmp = malloc(n); T* x = (T*)tmp;</TT>”. If you don’t need all casts to be made explicit, you can set <TT>Cabs2cil.doCollapseCallCast</TT> to true so that CIL won’t insert a temporary and you can more easily determine the allocation type from calls to <TT>malloc</TT>.</P></LI></UL><H3 CLASS="subsection"><A NAME="toc11"></A><A NAME="htoc19">7.4</A>  Specifying a machine model</H3><P> <A NAME="sec-cilmachine"></A></P><P>The <TT>--envmachine</TT> option tells CIL to get its machine model from the <TT>CIL_MACHINE</TT> environment variable, rather than use the model compiled in to CIL itself. This is necessary when using CIL as part of a cross-compilation setup, and to handle gcc’s <TT>-m32</TT> and <TT>-m64</TT> which select between a 32-bit and 64-bit machine model.</P><P><TT>CIL_MACHINE</TT> is a space-separated list of key=value pairs. Unknown keys are ignored. The following keys must be defined: </P><DIV CLASS="center"> <TABLE BORDER=1 CELLSPACING=0 CELLPADDING=1><TR><TD ALIGN=left NOWRAP> Key</TD><TD ALIGN=left NOWRAP>Value</TD></TR> <TR><TD ALIGN=left NOWRAP> bool</TD><TD ALIGN=left NOWRAP>sizeof(_Bool),alignof(_Bool)</TD></TR> <TR><TD ALIGN=left NOWRAP> short</TD><TD ALIGN=left NOWRAP>sizeof(short),alignof(short)</TD></TR> <TR><TD ALIGN=left NOWRAP> int</TD><TD ALIGN=left NOWRAP>sizeof(int),alignof(int)</TD></TR> <TR><TD ALIGN=left NOWRAP> long</TD><TD ALIGN=left NOWRAP>sizeof(long),alignof(long)</TD></TR> <TR><TD ALIGN=left NOWRAP> long_long</TD><TD ALIGN=left NOWRAP>sizeof(long long),alignof(long long)</TD></TR> <TR><TD ALIGN=left NOWRAP> float</TD><TD ALIGN=left NOWRAP>sizeof(float),alignof(float)</TD></TR> <TR><TD ALIGN=left NOWRAP> double</TD><TD ALIGN=left NOWRAP>sizeof(double),alignof(double)</TD></TR> <TR><TD ALIGN=left NOWRAP> long_double</TD><TD ALIGN=left NOWRAP>sizeof(long double),alignof(long double)</TD></TR> <TR><TD ALIGN=left NOWRAP> pointer</TD><TD ALIGN=left NOWRAP>sizeof(all pointers),alignof(all pointers)</TD></TR> <TR><TD ALIGN=left NOWRAP> enum</TD><TD ALIGN=left NOWRAP>sizeof(all enums),alignof(all enums)</TD></TR> <TR><TD ALIGN=left NOWRAP> fun</TD><TD ALIGN=left NOWRAP>sizeof(all fn ptrs),alignof(all fn ptrs)</TD></TR> <TR><TD ALIGN=left NOWRAP> alignof_string</TD><TD ALIGN=left NOWRAP>alignof(all string constants)</TD></TR> <TR><TD ALIGN=left NOWRAP> max_alignment</TD><TD ALIGN=left NOWRAP>maximum alignment for any type</TD></TR> <TR><TD ALIGN=left NOWRAP> size_t</TD><TD ALIGN=left NOWRAP>the definition of size_t</TD></TR> <TR><TD ALIGN=left NOWRAP> wchar_t</TD><TD ALIGN=left NOWRAP>the definition of wchar_t</TD></TR> <TR><TD ALIGN=left NOWRAP> char_signed</TD><TD ALIGN=left NOWRAP>true if char is signed</TD></TR> <TR><TD ALIGN=left NOWRAP> big_endian</TD><TD ALIGN=left NOWRAP>true for big-endian machine models</TD></TR> <TR><TD ALIGN=left NOWRAP> const_string_literals</TD><TD ALIGN=left NOWRAP>true if string constants are const</TD></TR> <TR><TD ALIGN=left NOWRAP> __thread_is_keyword</TD><TD ALIGN=left NOWRAP>true if <TT>__thread</TT> is a keyword</TD></TR> <TR><TD ALIGN=left NOWRAP> __builtin_va_list</TD><TD ALIGN=left NOWRAP>true if <TT>__builtin_va_list</TT> is a builtin type</TD></TR> <TR><TD ALIGN=left NOWRAP> underscore_name</TD><TD ALIGN=left NOWRAP>true if generated symbols preceded by _</TD></TR> </TABLE> </DIV><P>Some notes: </P><UL CLASS="itemize"><LI CLASS="li-itemize"> Values cannot contain spaces as spaces separate key/value pairs. </LI><LI CLASS="li-itemize">The value of size_t is the text for the type defined as size_t, e.g. <TT>unsigned long</TT>. As spaces are not allowed in values, CIL will replace underscores by spaces: <TT>size_t=unsigned_long</TT>. </LI><LI CLASS="li-itemize">sizeof(t) and alignof(t) are respectively the size and alignment of type t. </LI><LI CLASS="li-itemize">The boolean-valued keys expect <TT>true</TT> for true, and <TT>false</TT> for false. </LI><LI CLASS="li-itemize">The <TT>src/machdep-ml.c</TT> program will print a <TT>CIL_MACHINE</TT> value when run with the <TT>--env</TT> option. </LI></UL><P>As an example, here’s the <TT>CIL_MACHINE</TT> value for 64-bit targets on an x86-based Mac OS X machine: </P><PRE CLASS="verbatim"> short=2,2 int=4,4 long=8,8 long_long=8,8 pointer=8,8 enum=4,4 float=4,4 double=8,8 long_double=16,16 void=1 bool=1,1 fun=1,1 alignof_string=1 max_alignment=16 size_t=unsigned_long wchar_t=int char_signed=true const_string_literals=true big_endian=false __thread_is_keyword=true __builtin_va_list=true underscore_name=true </PRE><HR> <A HREF="attributes.html"><IMG SRC="previous_motif.gif" ALT="Previous"></A> <A HREF="ciltoc.html"><IMG SRC="contents_motif.gif" ALT="Up"></A> <A HREF="ext.html"><IMG SRC="next_motif.gif" ALT="Next"></A> </BODY> </HTML>