Loading ...
Sorry, an error occurred while loading the content.

Re: [perlguitest] Re: Win32-GuiTest v1.50.3-ad (Release)

Expand Messages
  • Gabor Szabo
    ... That could be *in addition* to the tests against notepad and co. I belive for real good tests we should also use real and changing targets. Usually I like
    Message 1 of 8 , Feb 4, 2005
    • 0 Attachment
      On Fri, 4 Feb 2005, pkaluski wrote:

      > Hi,
      > I think we should consider creating an application, which would be a
      > target for Win32::GuiTest unit tests. Relying on notepad or other M$
      > stuff is risky, because these applications change too often.
      > It would be easier to maintain tests if we have one application, which
      > hosts most of controls we would like to test.

      That could be *in addition* to the tests against notepad and co.
      I belive for real good tests we should also use real and changing
      targets.
      Usually I like to have two kinds of tests:
      one with data that I control - the application you suggest
      one with data from the wilde - the kind we have now


      >
      > -Piotr
      >
      > PS: By the way - do all tests from the most recent release pass on
      > your PCs? I have 4 of them failing.
      >

      It's hard to tell as my pc is running Linux :)

      ..but when I work on Windows stuff I open one in VMware.
      I just don't have time for that right now. Sorry.

      BTW I started to make a change (or did I just think about doing it ?)
      so that the unit tests will be included in the ppm distribution as well,
      or will be downloadable separately so even people who don't have a
      compiler and use the prebuilt version (either from Dennis or from
      ActiveState) will also be able to run the tests and help improve the
      code.


      Gabor
    • pkaluski
      ... be a ... M$ ... which ... I tend to disagree with you. IMHO, testing against special test application should not be an addition. It should be a core, the
      Message 2 of 8 , Feb 14, 2005
      • 0 Attachment
        --- In perlguitest@yahoogroups.com, Gabor Szabo <gabor@p...> wrote:
        > On Fri, 4 Feb 2005, pkaluski wrote:
        >
        > > Hi,
        > > I think we should consider creating an application, which would
        be a
        > > target for Win32::GuiTest unit tests. Relying on notepad or other
        M$
        > > stuff is risky, because these applications change too often.
        > > It would be easier to maintain tests if we have one application,
        which
        > > hosts most of controls we would like to test.
        >
        > That could be *in addition* to the tests against notepad and co.
        > I belive for real good tests we should also use real and changing
        > targets.
        > Usually I like to have two kinds of tests:
        > one with data that I control - the application you suggest
        > one with data from the wilde - the kind we have now
        >
        >
        I tend to disagree with you.
        IMHO, testing against special test application should not be an
        addition. It should be a core, the main part of the regression test
        suite. Here are my reasons:
        1. Win32::GuiTest is controls oriented, not application oriented.
        What I mean is that is uses the notion of controls rather than
        applications. It operates on controls. Therefore it seams more
        natural to make test controls oriented, rather then application
        oriented.
        2. If our goal is to make a portable, widely used and trusted module,
        we have to create a test suite, which should pass on every system we
        declare.

        I do agree that it would be good to have tests for real applications,
        but I don't think we have resources to maintain test scripts and make
        sure that tests pass on Win2K and on WinWhatSoEver. What if notepad
        in WinWhatSoEver is redesigned and uses different set of controls?
        At current stage, I believe that creating stable, easy to run test
        suite is more important than making sure that GuiTest work for
        particular application.

        -Piotr
      • Gabor Szabo
        ... You are welcome :) I think *both* should be there, that s what I usually do with software I am testing. But in any case, if you have the tuits to develop
        Message 3 of 8 , Feb 14, 2005
        • 0 Attachment
          On Mon, 14 Feb 2005, pkaluski wrote:

          > I tend to disagree with you.
          You are welcome :)

          I think *both* should be there, that's what I usually do
          with software I am testing.

          But in any case, if you have the tuits to develop the
          application to be tested then go ahead and do it and let's
          leave it for Dennis to decide what to include in the distro.

          The more test we have the better.

          Gabor
        • pkaluski
          Well, I did not expect, that I find the example of windows version inconsistency so quickly. Units test made on calculator fail. This is due to the fact, that
          Message 4 of 8 , Feb 15, 2005
          • 0 Attachment
            Well,
            I did not expect, that I find the example of windows version
            inconsistency so quickly.

            Units test made on calculator fail. This is due to the fact, that
            calculator on win2k will write 1969 as "1969. ". On XP it will be
            "1,969. ". There are also differences in formating binary and octal
            numbers.

            -Piotr

            --- In perlguitest@yahoogroups.com, "pkaluski" <pkaluski@p...> wrote:
            >
            > --- In perlguitest@yahoogroups.com, Gabor Szabo <gabor@p...> wrote:
            > > On Fri, 4 Feb 2005, pkaluski wrote:
            > >
            > > > Hi,
            > > > I think we should consider creating an application, which would
            > be a
            > > > target for Win32::GuiTest unit tests. Relying on notepad or
            other
            > M$
            > > > stuff is risky, because these applications change too often.
            > > > It would be easier to maintain tests if we have one
            application,
            > which
            > > > hosts most of controls we would like to test.
            > >
            > > That could be *in addition* to the tests against notepad and co.
            > > I belive for real good tests we should also use real and changing
            > > targets.
            > > Usually I like to have two kinds of tests:
            > > one with data that I control - the application you suggest
            > > one with data from the wilde - the kind we have now
            > >
            > >
            > I tend to disagree with you.
            > IMHO, testing against special test application should not be an
            > addition. It should be a core, the main part of the regression test
            > suite. Here are my reasons:
            > 1. Win32::GuiTest is controls oriented, not application oriented.
            > What I mean is that is uses the notion of controls rather than
            > applications. It operates on controls. Therefore it seams more
            > natural to make test controls oriented, rather then application
            > oriented.
            > 2. If our goal is to make a portable, widely used and trusted
            module,
            > we have to create a test suite, which should pass on every system
            we
            > declare.
            >
            > I do agree that it would be good to have tests for real
            applications,
            > but I don't think we have resources to maintain test scripts and
            make
            > sure that tests pass on Win2K and on WinWhatSoEver. What if notepad
            > in WinWhatSoEver is redesigned and uses different set of controls?
            > At current stage, I believe that creating stable, easy to run test
            > suite is more important than making sure that GuiTest work for
            > particular application.
            >
            > -Piotr
          • Gabor Szabo
            ... Is this difference due to a change in the application itself or is it because of a change in an underlying control that might be used by other applications
            Message 5 of 8 , Feb 18, 2005
            • 0 Attachment
              On Tue, 15 Feb 2005, pkaluski wrote:

              >
              >
              > Well,
              > I did not expect, that I find the example of windows version
              > inconsistency so quickly.
              >
              > Units test made on calculator fail. This is due to the fact, that
              > calculator on win2k will write 1969 as "1969. ". On XP it will be
              > "1,969. ". There are also differences in formating binary and octal
              > numbers.

              Is this difference due to a change in the application itself or is
              it because of a change in an underlying control that might be used
              by other applications as well ?


              I would add a conditional to the test checking the version
              of Windows and based on that decide what is the expected string.

              Gabor
            Your message has been successfully submitted and would be delivered to recipients shortly.