Megatest

</div>
<div class="sect1">
<h2 id="_debugging">Debugging</h2>
<div class="sectionbody">
<div class="sect2">
<h3 id="_well_written_tests">Well Written Tests</h3>
<div class="sect3">
<h4 id="_test_design_and_surfacing_errors">Test Design and Surfacing Errors</h4>
<div class="paragraph"><p>Design your tests to surface errors. Ensure that all logs are
processed by logpro (or a custom log processing tool) and can be
reached by a mouse click or two from the test control panel.</p></div>
<div class="paragraph"><p>To illustrate, here is a set of scripts with nested calls where script1.sh calls script2.sh which calls script3.sh which finally calls the Cadence EDA tool virtuoso:</p></div>
<div class="literalblock">
<div class="title">script1.sh</div>
<div class="content monospaced">
<pre>#!/bin/bash
code ...
script2.sh some parameters &gt; script2.log
more code ...</pre>
</div></div>
<div class="literalblock">
<div class="title">script2.sh</div>
<div class="content monospaced">
<pre>#!/bin/bash
code ...
script3.sh some more parameters &gt; script3.log
more code ...</pre>
</div></div>
<div class="literalblock">
<div class="title">script3.sh</div>
<div class="content monospaced">
<pre>#!/bin/bash
code ...
virtuoso params and switches ...
more code ...</pre>
</div></div>
<div class="paragraph"><p>The log files script2.log, script3.log and the log output from
virtuoso are not accessible from the test control panel. It would be
much better for future users of your automation to use steps more
fully. One easy option would be to post process the logs in downstream
additional steps:</p></div>
<div class="literalblock">
<div class="title">testconfig</div>
<div class="content monospaced">
<pre>[ezsteps]
step1 script1.sh
step2 cat script2.log
step3 cat script3.log

[logpro]
step1 ;; some logpro rules
  (expect:required in "LogFileBody" &gt; 0 "Expect this output" #/something expected/)
step2 ;; some logpro rules for script2.sh
step3 ;; some logpro rules for script3.sh

[scripts]
script1.sh #!/bin/bash
 code ...

...</pre>
</div></div>
<div class="paragraph"><p>With the above testconfig the logs for every critical part of the
automation are fully surfaced and rules can be created to flag errors,
warnings, aborts and to ignore false errors. A user of your automation
will be able to see the important error with two mouse clicks from the
runs view.</p></div>
<div class="paragraph"><p>An even better would be to eliminate the nesting if possible. As a
general statement with layers - less is usually more. By flattening
the automation into a sequence of steps you can use the test control
panel to re-run a step with a single click or from the test xterm run
only the errant step from the command line.</p></div>
<div class="paragraph"><p>The message here is make debugging and maintenace easy for future
users (and yourself) by keeping clicks-to-error in mind.</p></div>
</div>
</div>
<div class="sect2">
<h3 id="_examining_the_test_logs_and_environment">Examining The Test Logs and Environment</h3>
<div class="sect3">
<h4 id="_test_control_panel_xterm">Test Control Panel - xterm</h4>
<div class="paragraph"><p>From the dashboard click on a test PASS/FAIL button. This brings up a
test control panel. Aproximately near the center left of the window
there is a button "Start Xterm". Push this to get an xterm with the
full context and environment loaded for that test. You can run scripts
or ezsteps by copying from the testconfig (hint, load up the
testconfig in a separate text editor window).</p></div>
<div class="paragraph"><p>With more recent versions of Megatest you can step through your test
from the test control panel. Click on the cell labeled "rerun this
step" to only rerun the step or click on "restart from here" to rerun
that step and downstream steps.</p></div>
<div class="paragraph"><p>NOTE 1: visual feedback can take some time, give it a few seconds and
you will see the step change color to blue as it starts running.</p></div>
<div class="paragraph"><p>NOTE 2: steping through only works if you are using ezsteps.</p></div>
</div>
</div>
<div class="sect2">
<h3 id="_a_word_on_bisecting">A word on Bisecting</h3>
<div class="paragraph"><p>Bisecting is a debug strategy intended to speed up finding the root
cause of some bug.</p></div>
<div class="imageblock graphviz">
<div class="content">
<img src="bisecting.png" alt="bisecting.png">
</div>
<div class="title">Figure 1. A complex process with a problem found in stage "E"</div>
</div>
<div class="paragraph"><p>It is common to start debugging where the problem was observed and
then work back. However by inspecting the output at stage "C" in the
example above you would potentially save a lot of debug effort, this
is similar to the feature in source control tools like git and fossil
called biseceting.</p></div>
</div>
<div class="sect2">
<h3 id="_tough_bugs">Tough Bugs</h3>
<div class="paragraph"><p>Most bugs in Megatest based automation will be in the scripts called
in your test steps and if you utilize the good design practice
described above should be fairly easy for you to reproduce, isolate
and find.</p></div>
<div class="paragraph"><p>Some bugs however will come from subtle and hard to detect
interactions between Megatest and your OS and Unix environment. This
includes things like constructed variables that are legal in one
context (e.g. tcsh) but illegal in another context (e.g. bash),
variables that come from your login scripts and access and permissions
issues (e.g. a script that silently fails due to no access to needed
data). Other bugs might be due to Megatest itself.</p></div>
<div class="paragraph"><p>To isolate bugs like this you may need to look at the log files at
various stages in the execution process of your run and tests.</p></div>
<div class="imageblock graphviz">
<div class="content">
<img src="megatest-test-stages.png" alt="megatest-test-stages.png">
</div>
<div class="title">Figure 2. A simplified diagram of the stages Megatest goes through to run a test.</div>
</div>
<table class="tableblock frame-topbot grid-all"

<tr>
<td class="tableblock halign-left valign-top" ><p class="tableblock">H,H1,H2: step exectution</p></td>
<td class="tableblock halign-left valign-top" ><p class="tableblock monospaced">look at &lt;stepname&gt;.log, &lt;stepname&gt;.html and your own internal logs</p></td>
<td class="tableblock halign-left valign-top" ><p class="tableblock monospaced">Do you have sufficiently tight logpro rules? You must always have a "required" rule!</p></td>
</tr>
</tbody>
</table>
<div class="sect3">
<h4 id="_bisecting_megatest_csh_sh">Bisecting megatest.csh/sh</h4>
<div class="paragraph"><p>Sometimes finding the environment variable that is causing the problem
can be very difficult. Bisection can be applied.</p></div>
<div class="paragraph"><p>Edit the megatest.csh or megatest.sh file and comment out 50% per
round, source in fresh xterm and run the test.</p></div>
<div class="paragraph"><p>This idea can also be applied to your .cshrc, .bashrc, .aliases and
other similar files.</p></div>
</div>
<div class="sect3">

<h4 id="_csh_and_f">csh and -f</h4>
<div class="paragraph"><p>A common issue when tcsh or csh shells are used for scripting is to
forget or choose to not use -f in your #! line.</p></div>
<div class="literalblock">

<div class="title">Not good</div>
<div class="content monospaced">




<pre>#!/bin/tcsh
...</pre>
</div></div>
<div class="literalblock">


<div class="title">Good</div>
<div class="content monospaced">
<pre>#!/bin/tcsh -f
...</pre>
</div></div>
</div>
<div class="sect3">
<h4 id="_config_file_processing">Config File Processing</h4>
<div class="paragraph"><p>As described above it is often helpful to know the content of
variables in various contexts as Megatest works through the actions
needed to run your tests. A handy technique is to force the startup of
an xterm in the context being examined.</p></div>

<pre>[setup]
ezpropvars sh</pre>
</div></div>
<div class="listingblock">
<div class="title">Write your ezsteps. The loadenv.csh step will use /bin/csh as its shell, other steps will use bash.</div>
<div class="content monospaced">
<pre>[ezsteps]
# if your upstream file is csh you can force csh like this
loadenv.csh source $REF/ourenviron.csh
# if your upstream is bash
loadenv     source $REF/ourenviron.sh

compile make
install make install</pre>
</div></div>
<div class="paragraph"><p>Bash and csh are supported. You can override the shell binary location
from the default /bin/bash and /bin/csh if needed.</p></div>
<div class="listingblock">
<div class="title">Turn on auto propagate for csh</div>