Re: Testcase

[Jacob Bachmeyer]

shruthi k wrote:
> I'm trying to add a test case to an existing testsuite.

This is somewhat covered in section 5.6 "Writing a test case" in the 
manual.  This can also be dependent on the testsuite itself:  some 
testsuites search for certain input files and automatically run all 
files matching some criteria through a test while others require all 
tests be explicitly listed in Tcl code.

Off-list and in the wrong thread, you mentioned adding a "single test 
script", so I will assume that you already know that you will need to 
write some Expect code.  First, choose a tool-and-test-group directory.  
Using current Git DejaGnu as an example 
(<URL:http://git.savannah.gnu.org/cgit/dejagnu.git/tree/testsuite?id=e70fd889a7ae2ae3689c3cd84136dd2ddbd25cc9> 
used in this writing), there are, within the testsuite directory, 
directories for four tools:  launcher, libdejagnu, report-card, and 
runtest.  The libdejagnu directory is ancient and is simply the name of 
its tool, but DejaGnu searches for any directories beginning with the 
tool name, and a null suffix is still accepted for backwards 
compatibility.  Modern practice is to separate the tool name from a test 
group name with a single dot, as seen in the launcher.all, 
report-card.all, and runtest.* directories.  The runtest tests are the 
most complex and are divided into two groups:  main and libs, 
respectively found in the runtest.main and runtest.libs directories.

Your testsuite should have a similar layout, with directories inside 
"testsuite".  Choose one of those that begins with an appropriate prefix 
(the name of the tool your test tests) and create an *.exp file for your 
test script, like "example.exp", but obviously change "example" to 
something meaningful for your test script.  The test script itself 
should somehow automate running the tool and verifying a result, 
eventually calling "pass", "fail", or "unresolved" procedures to report 
a result.  (see section 1.4 "A POSIX compliant test framework" in the 
manual for more information)

> Also,could you please send links/info on how to run an individual test.

This is covered (briefly) near the end of the "Invoking runtest" section 
of the manual.  In short, you can cause DejaGnu to run only tests from 
one or more specific scripts by simply naming them on the command line.  
There is also an option to pass arguments to a test script run in this 
way, but what exactly DejaGnu *does* with that argument and how it is 
passed in to the test script is not yet documented.  (>-<)  (added to 
local TODO list)


-- Jacob