Hi,
when I run unittest class, with debugger the flow goes into testcase
case class as below.
how can i debug only the testscript instead of going into libraries of
extended clases
Testscript
require "test/unit"
class TestSimpleNumber < Test::Unit::TestCase
def test_simple
assert_equal("abc", "abc" )
end
def test_typecheck
assert_equal("abc", "abc" )
end
def test_failure
assert_equal("abc", "abc" )
end
End
Execution:
C:\ruby>ruby -rubygems -rdebug unittest.rb
Debug.rb
Emacs support available.
unittest.rb:1:require "test/unit"
(rdb:1) b 6
Set breakpoint 1 at unittest.rb:6
(rdb:1) cont
Loaded suite unittest
Started
.Breakpoint 1, test_simple at unittest.rb:6
unittest.rb:6: assert_equal("abc", "abc" )
(rdb:1) n
c:/ruby/lib/ruby/1.8/test/unit/testcase.rb:85: begin
(rdb:1) n
c:/ruby/lib/ruby/1.8/test/unit/testcase.rb:86: teardown
(rdb:1) n
c:/ruby/lib/ruby/1.8/test/unit/testcase.rb:94: result.add_run
(rdb:1) n
c:/ruby/lib/ruby/1.8/test/unit/testcase.rb:95: yield(FINISHED,
name)
(rdb:1) n
c:/ruby/lib/ruby/1.8/test/unit/ui/testrunnermediator.rb:47:
notify_listeners(channel, value)
(rdb:1)
Briefly, if you write unit tests, you should not need to debug.
I configure the tests to run after I hit one button. (Report your editor, and someone might know how to do it. Some, like TextMate, make it hard.)
Then I run them over and over again. Change a statement, predict what the tests will do, run the tests, see if your prediction matches. Sometimes, of course, use pp @variable to trace what's going on.
Briefly, if you write unit tests, you should not need to debug.
Oh yes you should. A test never debug; it just point out there are bugs. To
remove them is still your job.
If you use Test-Driven Development, you have the option to use Undo until the bug goes away. Try it; you will see what I mean.
On a big project, you might still sometimes need to debug. I have always used pp for trace statements, out of selected tests. Running them over and over again works so similar to debugging that you won't miss it.
However, someone here will have experience with the Ruby debugger and can answer your actual question. Any port in a storm!
Firstly, a test absolutely does help, even after the fact, as a way to debug.
It provides a concise definition of the bug, and if the test is kept around, it
will show whether the bug remains fixed.
But I think the point was that if your specs are comprehensive enough to begin
with, you will not introduce nearly as many bugs. I'm not arrogant enough to
say "none at all", but I've also never worked on a project which did Behavior-
Driven Design religiously enough for this to work.
···
On Monday 20 April 2009 11:25:16 Urabe Shyouhei wrote:
Phlip wrote:
> Briefly, if you write unit tests, you should not need to debug.
Oh yes you should. A test never debug; it just point out there are bugs.
To remove them is still your job.