<div dir="ltr"><div dir="ltr"><br><div><br></div><div><br></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Thu, Jan 24, 2019 at 3:04 AM Jiri Gaisler <<a href="mailto:jiri@gaisler.se">jiri@gaisler.se</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<div bgcolor="#FFFFFF">
<p><br>
</p>
<div class="gmail-m_-1994342416576470996moz-cite-prefix">On 1/24/19 9:33 AM, Jiri Gaisler wrote:<br>
</div>
<blockquote type="cite">
<br>
<div class="gmail-m_-1994342416576470996moz-cite-prefix">On 1/24/19 12:45 AM, Joel Sherrill
wrote:<br>
</div>
<blockquote type="cite">
<div dir="ltr">
<div dir="ltr"><br>
</div>
<br>
<div class="gmail_quote">
<div dir="ltr" class="gmail-m_-1994342416576470996gmail_attr">On Wed, Jan 23, 2019 at
4:45 PM Chris Johns <<a href="mailto:chrisj@rtems.org" target="_blank">chrisj@rtems.org</a>> wrote:<br>
</div>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On 24/1/19 9:12 am,
Jiri Gaisler wrote:<br>
> After some trial and error, I got covoar to run with:<br>
<br>
Great.<br>
<br>
> However, the coverage is always 0% in the summary
report:<br>
<br>
Does adding -v to the command line provide any more
detail? Repeating -v<br>
increases the level but things can become rather verbose
and if you trip full<br>
DWARF tracing there is a lot of output.<br>
</blockquote>
<div><br>
</div>
<div>I would be suspicious of the code that is invoked
around</div>
<div>CoverageReaderTSIM.cc:83 as a starting point. For some
method you</div>
<div>know you should be analyzed, there should be
aCoverageMap. Since you</div>
<div>are looking at only one executable, you could hard-code
a test with</div>
<div>address range of a single method under test to see what
happens.</div>
<br>
</div>
</div>
</blockquote>
<br>
<p>Well, the problem is that covoar in rtems-tools is hard-coded
to use the QEMU format, and the -f switch to change format is
not used. I have attached a patch to fix that. After that,
covoar works fine with the coverage data from sis:</p>
<p>cat coverage/summary.txt <br>
Bytes Analyzed : 20034<br>
Bytes Not Executed : 11734<br>
Percentage Executed : 41.43<br>
Percentage Not Executed : 58.57<br>
Uncovered ranges found : 242<br>
Total branches found : 443<br>
Uncovered branches found : 167<br>
68 branches always taken<br>
99 branches never taken</p>
<p>If you approve to the patch, then I will check it in (now that
I know how ... :-))<br>
</p>
</blockquote>
<p>Here is a better patch that maintains the old behavior of QEMU as
default, while honoring the -f switch if provided.</p></div></blockquote><div><br></div><div>Please push this. Although I would be perfectly happy returning to using sparc by default for coverage. </div><div>We developed covoar using tsim output. The qemu input requires the AdaCore Couverture variant of </div><div>qemu to get traces. That was the only option to do coverage with open, free tools. With sis including</div><div>coverage information, it is likely easier to debug covoar issues using it. </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div bgcolor="#FFFFFF">
<p>Jiri.<br>
</p>
</div>
</blockquote></div></div>