Re: [Pharo-project] what expectedfailures expected to do?
On Apr 20, 2010, at 11:20 PM, Adrian Lienhard wrote: Yes, if a test that is expected to fail does not fail, this is treated as a failure. I think that makes sense. well it depends about the scenario. you put on expectedfailures something that gets in your way now, so after if it works even better. of course you should get notified that the test is green while expected it to failed. Now it leads to a UI problem where you have a failure that passes so when you click on it nothing happens: no debugger. And you can wonder why the hell do I have a failure when my tests pass. So I think that this implementation of expectedFailures is a hack. Adrian On Apr 20, 2010, at 21:57 , Stéphane Ducasse wrote: Hi I tagged some tests as expected failures and I got a strange behavior. On the the tests which was passing was listed under the failures. When I renamed the method without updating the expected failures my bar was green. So expected failures really expect that the tests failed? We cannto have green tests in there? Stef ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Re: [Pharo-project] what expectedfailures expected to do?
Yea, I agree, the GUI is suboptimal. I still think, though, that treating this case as a failure is correct. For instance, consider the case where you had added a workaround to a known bug and when the bug is fixed you need to remove the workaround again. Maybe it even leads to a wrong behavior now that the bug is gone. In this case you really want to know that the test does not fail anymore. In any case, I think that tagging methods as expected failures should be done with pragmas and not with #expectedFailures. Like this it would also be much easier to understand what's going on when you have a failure in this test although all assertions pass. Adrian On Apr 21, 2010, at 08:22 , Stéphane Ducasse wrote: On Apr 20, 2010, at 11:20 PM, Adrian Lienhard wrote: Yes, if a test that is expected to fail does not fail, this is treated as a failure. I think that makes sense. well it depends about the scenario. you put on expectedfailures something that gets in your way now, so after if it works even better. of course you should get notified that the test is green while expected it to failed. Now it leads to a UI problem where you have a failure that passes so when you click on it nothing happens: no debugger. And you can wonder why the hell do I have a failure when my tests pass. So I think that this implementation of expectedFailures is a hack. Adrian On Apr 20, 2010, at 21:57 , Stéphane Ducasse wrote: Hi I tagged some tests as expected failures and I got a strange behavior. On the the tests which was passing was listed under the failures. When I renamed the method without updating the expected failures my bar was green. So expected failures really expect that the tests failed? We cannto have green tests in there? Stef ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Re: [Pharo-project] what expectedfailures expected to do?
On Apr 21, 2010, at 9:51 AM, Adrian Lienhard wrote: Yea, I agree, the GUI is suboptimal. I still think, though, that treating this case as a failure is correct. For instance, consider the case where you had added a workaround to a known bug and when the bug is fixed you need to remove the workaround again. Maybe it even leads to a wrong behavior now that the bug is gone. In this case you really want to know that the test does not fail anymore. yes Now I have the impression that expectedFailures should be like passes, failed, errors: a state of the tests. Stef In any case, I think that tagging methods as expected failures should be done with pragmas and not with #expectedFailures. Like this it would also be much easier to understand what's going on when you have a failure in this test although all assertions pass. Adrian On Apr 21, 2010, at 08:22 , Stéphane Ducasse wrote: On Apr 20, 2010, at 11:20 PM, Adrian Lienhard wrote: Yes, if a test that is expected to fail does not fail, this is treated as a failure. I think that makes sense. well it depends about the scenario. you put on expectedfailures something that gets in your way now, so after if it works even better. of course you should get notified that the test is green while expected it to failed. Now it leads to a UI problem where you have a failure that passes so when you click on it nothing happens: no debugger. And you can wonder why the hell do I have a failure when my tests pass. So I think that this implementation of expectedFailures is a hack. Adrian On Apr 20, 2010, at 21:57 , Stéphane Ducasse wrote: Hi I tagged some tests as expected failures and I got a strange behavior. On the the tests which was passing was listed under the failures. When I renamed the method without updating the expected failures my bar was green. So expected failures really expect that the tests failed? We cannto have green tests in there? Stef ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Re: [Pharo-project] what expectedfailures expected to do?
I think Adrian Kuhn did that in his SUnit work. I also remember he also introduced a difference between expectedFailures and expectedErrors. Doru On 21 Apr 2010, at 10:16, Stéphane Ducasse wrote: On Apr 21, 2010, at 9:51 AM, Adrian Lienhard wrote: Yea, I agree, the GUI is suboptimal. I still think, though, that treating this case as a failure is correct. For instance, consider the case where you had added a workaround to a known bug and when the bug is fixed you need to remove the workaround again. Maybe it even leads to a wrong behavior now that the bug is gone. In this case you really want to know that the test does not fail anymore. yes Now I have the impression that expectedFailures should be like passes, failed, errors: a state of the tests. Stef In any case, I think that tagging methods as expected failures should be done with pragmas and not with #expectedFailures. Like this it would also be much easier to understand what's going on when you have a failure in this test although all assertions pass. Adrian On Apr 21, 2010, at 08:22 , Stéphane Ducasse wrote: On Apr 20, 2010, at 11:20 PM, Adrian Lienhard wrote: Yes, if a test that is expected to fail does not fail, this is treated as a failure. I think that makes sense. well it depends about the scenario. you put on expectedfailures something that gets in your way now, so after if it works even better. of course you should get notified that the test is green while expected it to failed. Now it leads to a UI problem where you have a failure that passes so when you click on it nothing happens: no debugger. And you can wonder why the hell do I have a failure when my tests pass. So I think that this implementation of expectedFailures is a hack. Adrian On Apr 20, 2010, at 21:57 , Stéphane Ducasse wrote: Hi I tagged some tests as expected failures and I got a strange behavior. On the the tests which was passing was listed under the failures. When I renamed the method without updating the expected failures my bar was green. So expected failures really expect that the tests failed? We cannto have green tests in there? Stef ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo- project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project -- www.tudorgirba.com Live like you mean it. ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Re: [Pharo-project] what expectedfailures expected to do?
We should have a look at what adrian did now the problem is that understanding a large set of changes is more difficult than a couple of simple ones. If somebody want to help we are open. Stef I think Adrian Kuhn did that in his SUnit work. I also remember he also introduced a difference between expectedFailures and expectedErrors. Doru On 21 Apr 2010, at 10:16, Stéphane Ducasse wrote: On Apr 21, 2010, at 9:51 AM, Adrian Lienhard wrote: Yea, I agree, the GUI is suboptimal. I still think, though, that treating this case as a failure is correct. For instance, consider the case where you had added a workaround to a known bug and when the bug is fixed you need to remove the workaround again. Maybe it even leads to a wrong behavior now that the bug is gone. In this case you really want to know that the test does not fail anymore. yes Now I have the impression that expectedFailures should be like passes, failed, errors: a state of the tests. Stef In any case, I think that tagging methods as expected failures should be done with pragmas and not with #expectedFailures. Like this it would also be much easier to understand what's going on when you have a failure in this test although all assertions pass. Adrian On Apr 21, 2010, at 08:22 , Stéphane Ducasse wrote: On Apr 20, 2010, at 11:20 PM, Adrian Lienhard wrote: Yes, if a test that is expected to fail does not fail, this is treated as a failure. I think that makes sense. well it depends about the scenario. you put on expectedfailures something that gets in your way now, so after if it works even better. of course you should get notified that the test is green while expected it to failed. Now it leads to a UI problem where you have a failure that passes so when you click on it nothing happens: no debugger. And you can wonder why the hell do I have a failure when my tests pass. So I think that this implementation of expectedFailures is a hack. Adrian On Apr 20, 2010, at 21:57 , Stéphane Ducasse wrote: Hi I tagged some tests as expected failures and I got a strange behavior. On the the tests which was passing was listed under the failures. When I renamed the method without updating the expected failures my bar was green. So expected failures really expect that the tests failed? We cannto have green tests in there? Stef ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project -- www.tudorgirba.com Live like you mean it. ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Re: [Pharo-project] what expectedfailures expected to do?
I'm trying to follow along with this discussion, but now I feel really stupid :-( I've been using SUnit since 1998 (when it was still called BeckTestingFramework), and ever since expected failures showed up, I've never understood them. So I wonder if some kind guru-like person could please explain to me what they are useful for? I mean, to my way of thinking, if one writes a test that is expected to fail, then why not invert the test and call it a success instead? for example: self should: [answer = 42] ...as an expected failure could simply be re-written as self should: [answer ~= 42] ...right? No, obviously I've missed something really obvious and important, and that's why I'm asking the question now. Please be gentle ;-) -- Cheers, Peter On Wed, Apr 21, 2010 at 10:30 AM, Stéphane Ducasse stephane.duca...@inria.fr wrote: We should have a look at what adrian did now the problem is that understanding a large set of changes is more difficult than a couple of simple ones. If somebody want to help we are open. Stef I think Adrian Kuhn did that in his SUnit work. I also remember he also introduced a difference between expectedFailures and expectedErrors. Doru On 21 Apr 2010, at 10:16, Stéphane Ducasse wrote: On Apr 21, 2010, at 9:51 AM, Adrian Lienhard wrote: Yea, I agree, the GUI is suboptimal. I still think, though, that treating this case as a failure is correct. For instance, consider the case where you had added a workaround to a known bug and when the bug is fixed you need to remove the workaround again. Maybe it even leads to a wrong behavior now that the bug is gone. In this case you really want to know that the test does not fail anymore. yes Now I have the impression that expectedFailures should be like passes, failed, errors: a state of the tests. Stef In any case, I think that tagging methods as expected failures should be done with pragmas and not with #expectedFailures. Like this it would also be much easier to understand what's going on when you have a failure in this test although all assertions pass. Adrian On Apr 21, 2010, at 08:22 , Stéphane Ducasse wrote: On Apr 20, 2010, at 11:20 PM, Adrian Lienhard wrote: Yes, if a test that is expected to fail does not fail, this is treated as a failure. I think that makes sense. well it depends about the scenario. you put on expectedfailures something that gets in your way now, so after if it works even better. of course you should get notified that the test is green while expected it to failed. Now it leads to a UI problem where you have a failure that passes so when you click on it nothing happens: no debugger. And you can wonder why the hell do I have a failure when my tests pass. So I think that this implementation of expectedFailures is a hack. Adrian On Apr 20, 2010, at 21:57 , Stéphane Ducasse wrote: Hi I tagged some tests as expected failures and I got a strange behavior. On the the tests which was passing was listed under the failures. When I renamed the method without updating the expected failures my bar was green. So expected failures really expect that the tests failed? We cannto have green tests in there? Stef ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project -- www.tudorgirba.com Live like you mean it. ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Re: [Pharo-project] what expectedfailures expected to do?
Hi, The idea behind this is that you can use unit tests to document things you have not done yet, or bugs that you know about but you do not want to spend time working on right now. Simply reverting the assertion in your test does make the test runner green, but it fails to document the intention, and a newcomer might get to the false conclusion that the answer is not 42. By marking the test as expected failure, you make the test runner green, but you also explicitly say that the answer should be 42, but the machine is not quite perfect yet. Cheers, Doru On 21 Apr 2010, at 11:06, Peter Hugosson-Miller wrote: I'm trying to follow along with this discussion, but now I feel really stupid :-( I've been using SUnit since 1998 (when it was still called BeckTestingFramework), and ever since expected failures showed up, I've never understood them. So I wonder if some kind guru-like person could please explain to me what they are useful for? I mean, to my way of thinking, if one writes a test that is expected to fail, then why not invert the test and call it a success instead? for example: self should: [answer = 42] ...as an expected failure could simply be re-written as self should: [answer ~= 42] ...right? No, obviously I've missed something really obvious and important, and that's why I'm asking the question now. Please be gentle ;-) -- Cheers, Peter On Wed, Apr 21, 2010 at 10:30 AM, Stéphane Ducasse stephane.duca...@inria.fr wrote: We should have a look at what adrian did now the problem is that understanding a large set of changes is more difficult than a couple of simple ones. If somebody want to help we are open. Stef I think Adrian Kuhn did that in his SUnit work. I also remember he also introduced a difference between expectedFailures and expectedErrors. Doru On 21 Apr 2010, at 10:16, Stéphane Ducasse wrote: On Apr 21, 2010, at 9:51 AM, Adrian Lienhard wrote: Yea, I agree, the GUI is suboptimal. I still think, though, that treating this case as a failure is correct. For instance, consider the case where you had added a workaround to a known bug and when the bug is fixed you need to remove the workaround again. Maybe it even leads to a wrong behavior now that the bug is gone. In this case you really want to know that the test does not fail anymore. yes Now I have the impression that expectedFailures should be like passes, failed, errors: a state of the tests. Stef In any case, I think that tagging methods as expected failures should be done with pragmas and not with #expectedFailures. Like this it would also be much easier to understand what's going on when you have a failure in this test although all assertions pass. Adrian On Apr 21, 2010, at 08:22 , Stéphane Ducasse wrote: On Apr 20, 2010, at 11:20 PM, Adrian Lienhard wrote: Yes, if a test that is expected to fail does not fail, this is treated as a failure. I think that makes sense. well it depends about the scenario. you put on expectedfailures something that gets in your way now, so after if it works even better. of course you should get notified that the test is green while expected it to failed. Now it leads to a UI problem where you have a failure that passes so when you click on it nothing happens: no debugger. And you can wonder why the hell do I have a failure when my tests pass. So I think that this implementation of expectedFailures is a hack. Adrian On Apr 20, 2010, at 21:57 , Stéphane Ducasse wrote: Hi I tagged some tests as expected failures and I got a strange behavior. On the the tests which was passing was listed under the failures. When I renamed the method without updating the expected failures my bar was green. So expected failures really expect that the tests failed? We cannto have green tests in there? Stef ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo- project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project -- www.tudorgirba.com Live like you mean it. ___ Pharo-project mailing list
Re: [Pharo-project] what expectedfailures expected to do?
Thanks for answering, Doru! So if I've understood you correctly, expected failures are useful when one wants to give some code to someone else, with all the tests running green, but at the same time let that person know that some specific tests are really still red, but known about. In other words, are they simply a way of documenting work in progress, and not for production code? -- Cheers, Peter On Wed, Apr 21, 2010 at 11:12 AM, Tudor Girba tudor.gi...@gmail.com wrote: Hi, The idea behind this is that you can use unit tests to document things you have not done yet, or bugs that you know about but you do not want to spend time working on right now. Simply reverting the assertion in your test does make the test runner green, but it fails to document the intention, and a newcomer might get to the false conclusion that the answer is not 42. By marking the test as expected failure, you make the test runner green, but you also explicitly say that the answer should be 42, but the machine is not quite perfect yet. Cheers, Doru On 21 Apr 2010, at 11:06, Peter Hugosson-Miller wrote: I'm trying to follow along with this discussion, but now I feel really stupid :-( I've been using SUnit since 1998 (when it was still called BeckTestingFramework), and ever since expected failures showed up, I've never understood them. So I wonder if some kind guru-like person could please explain to me what they are useful for? I mean, to my way of thinking, if one writes a test that is expected to fail, then why not invert the test and call it a success instead? for example: self should: [answer = 42] ...as an expected failure could simply be re-written as self should: [answer ~= 42] ...right? No, obviously I've missed something really obvious and important, and that's why I'm asking the question now. Please be gentle ;-) -- Cheers, Peter On Wed, Apr 21, 2010 at 10:30 AM, Stéphane Ducasse stephane.duca...@inria.fr wrote: We should have a look at what adrian did now the problem is that understanding a large set of changes is more difficult than a couple of simple ones. If somebody want to help we are open. Stef I think Adrian Kuhn did that in his SUnit work. I also remember he also introduced a difference between expectedFailures and expectedErrors. Doru On 21 Apr 2010, at 10:16, Stéphane Ducasse wrote: On Apr 21, 2010, at 9:51 AM, Adrian Lienhard wrote: Yea, I agree, the GUI is suboptimal. I still think, though, that treating this case as a failure is correct. For instance, consider the case where you had added a workaround to a known bug and when the bug is fixed you need to remove the workaround again. Maybe it even leads to a wrong behavior now that the bug is gone. In this case you really want to know that the test does not fail anymore. yes Now I have the impression that expectedFailures should be like passes, failed, errors: a state of the tests. Stef In any case, I think that tagging methods as expected failures should be done with pragmas and not with #expectedFailures. Like this it would also be much easier to understand what's going on when you have a failure in this test although all assertions pass. Adrian On Apr 21, 2010, at 08:22 , Stéphane Ducasse wrote: On Apr 20, 2010, at 11:20 PM, Adrian Lienhard wrote: Yes, if a test that is expected to fail does not fail, this is treated as a failure. I think that makes sense. well it depends about the scenario. you put on expectedfailures something that gets in your way now, so after if it works even better. of course you should get notified that the test is green while expected it to failed. Now it leads to a UI problem where you have a failure that passes so when you click on it nothing happens: no debugger. And you can wonder why the hell do I have a failure when my tests pass. So I think that this implementation of expectedFailures is a hack. Adrian On Apr 20, 2010, at 21:57 , Stéphane Ducasse wrote: Hi I tagged some tests as expected failures and I got a strange behavior. On the the tests which was passing was listed under the failures. When I renamed the method without updating the expected failures my bar was green. So expected failures really expect that the tests failed? We cannto have green tests in there? Stef ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Re: [Pharo-project] what expectedfailures expected to do?
Basically yes. It's not just for giving it to someone else, but also for yourself. The idea is to quickly notice when tests change color (from green to red), so instead of remembering how many failed before and comparing with the current number of failures, you simply mark them as expected failures and your bar becomes green. Doru On 21 Apr 2010, at 11:28, Peter Hugosson-Miller wrote: Thanks for answering, Doru! So if I've understood you correctly, expected failures are useful when one wants to give some code to someone else, with all the tests running green, but at the same time let that person know that some specific tests are really still red, but known about. In other words, are they simply a way of documenting work in progress, and not for production code? -- Cheers, Peter On Wed, Apr 21, 2010 at 11:12 AM, Tudor Girba tudor.gi...@gmail.com wrote: Hi, The idea behind this is that you can use unit tests to document things you have not done yet, or bugs that you know about but you do not want to spend time working on right now. Simply reverting the assertion in your test does make the test runner green, but it fails to document the intention, and a newcomer might get to the false conclusion that the answer is not 42. By marking the test as expected failure, you make the test runner green, but you also explicitly say that the answer should be 42, but the machine is not quite perfect yet. Cheers, Doru On 21 Apr 2010, at 11:06, Peter Hugosson-Miller wrote: I'm trying to follow along with this discussion, but now I feel really stupid :-( I've been using SUnit since 1998 (when it was still called BeckTestingFramework), and ever since expected failures showed up, I've never understood them. So I wonder if some kind guru-like person could please explain to me what they are useful for? I mean, to my way of thinking, if one writes a test that is expected to fail, then why not invert the test and call it a success instead? for example: self should: [answer = 42] ...as an expected failure could simply be re-written as self should: [answer ~= 42] ...right? No, obviously I've missed something really obvious and important, and that's why I'm asking the question now. Please be gentle ;-) -- Cheers, Peter On Wed, Apr 21, 2010 at 10:30 AM, Stéphane Ducasse stephane.duca...@inria.fr wrote: We should have a look at what adrian did now the problem is that understanding a large set of changes is more difficult than a couple of simple ones. If somebody want to help we are open. Stef I think Adrian Kuhn did that in his SUnit work. I also remember he also introduced a difference between expectedFailures and expectedErrors. Doru On 21 Apr 2010, at 10:16, Stéphane Ducasse wrote: On Apr 21, 2010, at 9:51 AM, Adrian Lienhard wrote: Yea, I agree, the GUI is suboptimal. I still think, though, that treating this case as a failure is correct. For instance, consider the case where you had added a workaround to a known bug and when the bug is fixed you need to remove the workaround again. Maybe it even leads to a wrong behavior now that the bug is gone. In this case you really want to know that the test does not fail anymore. yes Now I have the impression that expectedFailures should be like passes, failed, errors: a state of the tests. Stef In any case, I think that tagging methods as expected failures should be done with pragmas and not with #expectedFailures. Like this it would also be much easier to understand what's going on when you have a failure in this test although all assertions pass. Adrian On Apr 21, 2010, at 08:22 , Stéphane Ducasse wrote: On Apr 20, 2010, at 11:20 PM, Adrian Lienhard wrote: Yes, if a test that is expected to fail does not fail, this is treated as a failure. I think that makes sense. well it depends about the scenario. you put on expectedfailures something that gets in your way now, so after if it works even better. of course you should get notified that the test is green while expected it to failed. Now it leads to a UI problem where you have a failure that passes so when you click on it nothing happens: no debugger. And you can wonder why the hell do I have a failure when my tests pass. So I think that this implementation of expectedFailures is a hack. Adrian On Apr 20, 2010, at 21:57 , Stéphane Ducasse wrote: Hi I tagged some tests as expected failures and I got a strange behavior. On the the tests which was passing was listed under the failures. When I renamed the method without updating the expected failures my bar was green. So expected failures really expect that the tests failed? We cannto have green tests in there? Stef ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr
Re: [Pharo-project] what expectedfailures expected to do?
On Wed, Apr 21, 2010 at 11:34 AM, Tudor Girba tudor.gi...@gmail.com wrote: Basically yes. It's not just for giving it to someone else, but also for yourself. The idea is to quickly notice when tests change color (from green to red), so instead of remembering how many failed before and comparing with the current number of failures, you simply mark them as expected failures and your bar becomes green. Thanks, I think I get it now. So really, I shouldn't expect to see any expected failures in Pharo 1.0, but since Pharo 1.1 is still in alpha, it would be OK to have some there until it's ready for release. -- Cheers, Peter Doru On 21 Apr 2010, at 11:28, Peter Hugosson-Miller wrote: Thanks for answering, Doru! So if I've understood you correctly, expected failures are useful when one wants to give some code to someone else, with all the tests running green, but at the same time let that person know that some specific tests are really still red, but known about. In other words, are they simply a way of documenting work in progress, and not for production code? -- Cheers, Peter On Wed, Apr 21, 2010 at 11:12 AM, Tudor Girba tudor.gi...@gmail.com wrote: Hi, The idea behind this is that you can use unit tests to document things you have not done yet, or bugs that you know about but you do not want to spend time working on right now. Simply reverting the assertion in your test does make the test runner green, but it fails to document the intention, and a newcomer might get to the false conclusion that the answer is not 42. By marking the test as expected failure, you make the test runner green, but you also explicitly say that the answer should be 42, but the machine is not quite perfect yet. Cheers, Doru On 21 Apr 2010, at 11:06, Peter Hugosson-Miller wrote: I'm trying to follow along with this discussion, but now I feel really stupid :-( I've been using SUnit since 1998 (when it was still called BeckTestingFramework), and ever since expected failures showed up, I've never understood them. So I wonder if some kind guru-like person could please explain to me what they are useful for? I mean, to my way of thinking, if one writes a test that is expected to fail, then why not invert the test and call it a success instead? for example: self should: [answer = 42] ...as an expected failure could simply be re-written as self should: [answer ~= 42] ...right? No, obviously I've missed something really obvious and important, and that's why I'm asking the question now. Please be gentle ;-) -- Cheers, Peter On Wed, Apr 21, 2010 at 10:30 AM, Stéphane Ducasse stephane.duca...@inria.fr wrote: We should have a look at what adrian did now the problem is that understanding a large set of changes is more difficult than a couple of simple ones. If somebody want to help we are open. Stef I think Adrian Kuhn did that in his SUnit work. I also remember he also introduced a difference between expectedFailures and expectedErrors. Doru On 21 Apr 2010, at 10:16, Stéphane Ducasse wrote: On Apr 21, 2010, at 9:51 AM, Adrian Lienhard wrote: Yea, I agree, the GUI is suboptimal. I still think, though, that treating this case as a failure is correct. For instance, consider the case where you had added a workaround to a known bug and when the bug is fixed you need to remove the workaround again. Maybe it even leads to a wrong behavior now that the bug is gone. In this case you really want to know that the test does not fail anymore. yes Now I have the impression that expectedFailures should be like passes, failed, errors: a state of the tests. Stef In any case, I think that tagging methods as expected failures should be done with pragmas and not with #expectedFailures. Like this it would also be much easier to understand what's going on when you have a failure in this test although all assertions pass. Adrian On Apr 21, 2010, at 08:22 , Stéphane Ducasse wrote: On Apr 20, 2010, at 11:20 PM, Adrian Lienhard wrote: Yes, if a test that is expected to fail does not fail, this is treated as a failure. I think that makes sense. well it depends about the scenario. you put on expectedfailures something that gets in your way now, so after if it works even better. of course you should get notified that the test is green while expected it to failed. Now it leads to a UI problem where you have a failure that passes so when you click on it nothing happens: no debugger. And you can wonder why the hell do I have a failure when my tests pass. So I think that this implementation of expectedFailures is a hack. Adrian On Apr 20, 2010, at 21:57 , Stéphane Ducasse wrote: Hi I tagged some tests as expected failures and I got a strange behavior. On the the tests which was passing was listed under the failures. When I renamed the method without updating the expected
Re: [Pharo-project] what expectedfailures expected to do?
On Apr 21, 2010, at 11:06 AM, Peter Hugosson-Miller wrote: I'm trying to follow along with this discussion, but now I feel really stupid :-( I've been using SUnit since 1998 (when it was still called BeckTestingFramework), testingFramework on VW :) and ever since expected failures showed up, I've never understood them. So I wonder if some kind guru-like person could please explain to me what they are useful for? I mean, to my way of thinking, if one writes a test that is expected to fail, then why not invert the test and call it a success instead? for example: self should: [answer = 42] self assert: answer = 42 because [] are for self should: [] raise: Error and friend ...as an expected failure could simply be re-written as self should: [answer ~= 42] ...right? No, obviously I've missed something really obvious and important, and that's why I'm asking the question now. Please be gentle ;-) -- Cheers, Peter On Wed, Apr 21, 2010 at 10:30 AM, Stéphane Ducasse stephane.duca...@inria.fr wrote: We should have a look at what adrian did now the problem is that understanding a large set of changes is more difficult than a couple of simple ones. If somebody want to help we are open. Stef I think Adrian Kuhn did that in his SUnit work. I also remember he also introduced a difference between expectedFailures and expectedErrors. Doru On 21 Apr 2010, at 10:16, Stéphane Ducasse wrote: On Apr 21, 2010, at 9:51 AM, Adrian Lienhard wrote: Yea, I agree, the GUI is suboptimal. I still think, though, that treating this case as a failure is correct. For instance, consider the case where you had added a workaround to a known bug and when the bug is fixed you need to remove the workaround again. Maybe it even leads to a wrong behavior now that the bug is gone. In this case you really want to know that the test does not fail anymore. yes Now I have the impression that expectedFailures should be like passes, failed, errors: a state of the tests. Stef In any case, I think that tagging methods as expected failures should be done with pragmas and not with #expectedFailures. Like this it would also be much easier to understand what's going on when you have a failure in this test although all assertions pass. Adrian On Apr 21, 2010, at 08:22 , Stéphane Ducasse wrote: On Apr 20, 2010, at 11:20 PM, Adrian Lienhard wrote: Yes, if a test that is expected to fail does not fail, this is treated as a failure. I think that makes sense. well it depends about the scenario. you put on expectedfailures something that gets in your way now, so after if it works even better. of course you should get notified that the test is green while expected it to failed. Now it leads to a UI problem where you have a failure that passes so when you click on it nothing happens: no debugger. And you can wonder why the hell do I have a failure when my tests pass. So I think that this implementation of expectedFailures is a hack. Adrian On Apr 20, 2010, at 21:57 , Stéphane Ducasse wrote: Hi I tagged some tests as expected failures and I got a strange behavior. On the the tests which was passing was listed under the failures. When I renamed the method without updating the expected failures my bar was green. So expected failures really expect that the tests failed? We cannto have green tests in there? Stef ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project -- www.tudorgirba.com Live like you mean it. ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list
Re: [Pharo-project] what expectedfailures expected to do?
On Apr 21, 2010, at 11:47 AM, Peter Hugosson-Miller wrote: On Wed, Apr 21, 2010 at 11:34 AM, Tudor Girba tudor.gi...@gmail.com wrote: Basically yes. It's not just for giving it to someone else, but also for yourself. The idea is to quickly notice when tests change color (from green to red), so instead of remembering how many failed before and comparing with the current number of failures, you simply mark them as expected failures and your bar becomes green. Thanks, I think I get it now. So really, I shouldn't expect to see any expected failures in Pharo 1.0, but since Pharo 1.1 is still in alpha, it would be OK to have some there until it's ready for release. in a perfect world now in pharo we have tests that are red and we could not fixed easily. -- Cheers, Peter Doru On 21 Apr 2010, at 11:28, Peter Hugosson-Miller wrote: Thanks for answering, Doru! So if I've understood you correctly, expected failures are useful when one wants to give some code to someone else, with all the tests running green, but at the same time let that person know that some specific tests are really still red, but known about. In other words, are they simply a way of documenting work in progress, and not for production code? -- Cheers, Peter On Wed, Apr 21, 2010 at 11:12 AM, Tudor Girba tudor.gi...@gmail.com wrote: Hi, The idea behind this is that you can use unit tests to document things you have not done yet, or bugs that you know about but you do not want to spend time working on right now. Simply reverting the assertion in your test does make the test runner green, but it fails to document the intention, and a newcomer might get to the false conclusion that the answer is not 42. By marking the test as expected failure, you make the test runner green, but you also explicitly say that the answer should be 42, but the machine is not quite perfect yet. Cheers, Doru On 21 Apr 2010, at 11:06, Peter Hugosson-Miller wrote: I'm trying to follow along with this discussion, but now I feel really stupid :-( I've been using SUnit since 1998 (when it was still called BeckTestingFramework), and ever since expected failures showed up, I've never understood them. So I wonder if some kind guru-like person could please explain to me what they are useful for? I mean, to my way of thinking, if one writes a test that is expected to fail, then why not invert the test and call it a success instead? for example: self should: [answer = 42] ...as an expected failure could simply be re-written as self should: [answer ~= 42] ...right? No, obviously I've missed something really obvious and important, and that's why I'm asking the question now. Please be gentle ;-) -- Cheers, Peter On Wed, Apr 21, 2010 at 10:30 AM, Stéphane Ducasse stephane.duca...@inria.fr wrote: We should have a look at what adrian did now the problem is that understanding a large set of changes is more difficult than a couple of simple ones. If somebody want to help we are open. Stef I think Adrian Kuhn did that in his SUnit work. I also remember he also introduced a difference between expectedFailures and expectedErrors. Doru On 21 Apr 2010, at 10:16, Stéphane Ducasse wrote: On Apr 21, 2010, at 9:51 AM, Adrian Lienhard wrote: Yea, I agree, the GUI is suboptimal. I still think, though, that treating this case as a failure is correct. For instance, consider the case where you had added a workaround to a known bug and when the bug is fixed you need to remove the workaround again. Maybe it even leads to a wrong behavior now that the bug is gone. In this case you really want to know that the test does not fail anymore. yes Now I have the impression that expectedFailures should be like passes, failed, errors: a state of the tests. Stef In any case, I think that tagging methods as expected failures should be done with pragmas and not with #expectedFailures. Like this it would also be much easier to understand what's going on when you have a failure in this test although all assertions pass. Adrian On Apr 21, 2010, at 08:22 , Stéphane Ducasse wrote: On Apr 20, 2010, at 11:20 PM, Adrian Lienhard wrote: Yes, if a test that is expected to fail does not fail, this is treated as a failure. I think that makes sense. well it depends about the scenario. you put on expectedfailures something that gets in your way now, so after if it works even better. of course you should get notified that the test is green while expected it to failed. Now it leads to a UI problem where you have a failure that passes so when you click on it nothing happens: no debugger. And you can wonder why the hell do I have a failure when my tests pass. So I think that this implementation of expectedFailures is a hack. Adrian On Apr 20, 2010, at 21:57 ,
Re: [Pharo-project] what expectedfailures expected to do?
On Wed, Apr 21, 2010 at 1:37 PM, Stéphane Ducasse stephane.duca...@inria.fr wrote: On Apr 21, 2010, at 11:47 AM, Peter Hugosson-Miller wrote: On Wed, Apr 21, 2010 at 11:34 AM, Tudor Girba tudor.gi...@gmail.com wrote: Basically yes. It's not just for giving it to someone else, but also for yourself. The idea is to quickly notice when tests change color (from green to red), so instead of remembering how many failed before and comparing with the current number of failures, you simply mark them as expected failures and your bar becomes green. Thanks, I think I get it now. So really, I shouldn't expect to see any expected failures in Pharo 1.0, but since Pharo 1.1 is still in alpha, it would be OK to have some there until it's ready for release. in a perfect world now in pharo we have tests that are red and we could not fixed easily. Ah, OK. I guess it's really a question of perception. The world in which I work is a bit different from the wonderful world of Pharo (no sarcasm intended!), in that we can't ship *anything* unless *all* our tests are green. The concept of expected failures just has no place, because as long as there are failures we still have work to do. I can understand Doru's description of an expected failure as a kind of bookmark, to be added and discarded during the day, but I get a bit nervous that the sight of too much green might make me miss the fact that I'm not finished yet :-p -- Cheers, Peter Doru On 21 Apr 2010, at 11:28, Peter Hugosson-Miller wrote: Thanks for answering, Doru! So if I've understood you correctly, expected failures are useful when one wants to give some code to someone else, with all the tests running green, but at the same time let that person know that some specific tests are really still red, but known about. In other words, are they simply a way of documenting work in progress, and not for production code? -- Cheers, Peter On Wed, Apr 21, 2010 at 11:12 AM, Tudor Girba tudor.gi...@gmail.com wrote: Hi, The idea behind this is that you can use unit tests to document things you have not done yet, or bugs that you know about but you do not want to spend time working on right now. Simply reverting the assertion in your test does make the test runner green, but it fails to document the intention, and a newcomer might get to the false conclusion that the answer is not 42. By marking the test as expected failure, you make the test runner green, but you also explicitly say that the answer should be 42, but the machine is not quite perfect yet. Cheers, Doru On 21 Apr 2010, at 11:06, Peter Hugosson-Miller wrote: I'm trying to follow along with this discussion, but now I feel really stupid :-( I've been using SUnit since 1998 (when it was still called BeckTestingFramework), and ever since expected failures showed up, I've never understood them. So I wonder if some kind guru-like person could please explain to me what they are useful for? I mean, to my way of thinking, if one writes a test that is expected to fail, then why not invert the test and call it a success instead? for example: self should: [answer = 42] ...as an expected failure could simply be re-written as self should: [answer ~= 42] ...right? No, obviously I've missed something really obvious and important, and that's why I'm asking the question now. Please be gentle ;-) -- Cheers, Peter On Wed, Apr 21, 2010 at 10:30 AM, Stéphane Ducasse stephane.duca...@inria.fr wrote: We should have a look at what adrian did now the problem is that understanding a large set of changes is more difficult than a couple of simple ones. If somebody want to help we are open. Stef I think Adrian Kuhn did that in his SUnit work. I also remember he also introduced a difference between expectedFailures and expectedErrors. Doru On 21 Apr 2010, at 10:16, Stéphane Ducasse wrote: On Apr 21, 2010, at 9:51 AM, Adrian Lienhard wrote: Yea, I agree, the GUI is suboptimal. I still think, though, that treating this case as a failure is correct. For instance, consider the case where you had added a workaround to a known bug and when the bug is fixed you need to remove the workaround again. Maybe it even leads to a wrong behavior now that the bug is gone. In this case you really want to know that the test does not fail anymore. yes Now I have the impression that expectedFailures should be like passes, failed, errors: a state of the tests. Stef In any case, I think that tagging methods as expected failures should be done with pragmas and not with #expectedFailures. Like this it would also be much easier to understand what's going on when you have a failure in this test although all assertions pass. Adrian On Apr 21, 2010, at 08:22 , Stéphane Ducasse wrote: On Apr 20, 2010, at 11:20 PM,
Re: [Pharo-project] what expectedfailures expected to do?
Ah, OK. I guess it's really a question of perception. not perception. Just money. Give me three engineers full times during two years and you will not recognize pharo :) The world in which I work is a bit different from the wonderful world of Pharo (no sarcasm intended!), in that we can't ship *anything* unless *all* our tests are green. The concept of expected failures just has no place, because as long as there are failures we still have work to do. Oh this is easy we can remove tests and your customers will be happy :) Now some tests represents the semantics of blocclosure that no smalltalk truly implement, so this would be a pity to throw them away. I can understand Doru's description of an expected failure as a kind of bookmark, to be added and discarded during the day, but I get a bit nervous that the sight of too much green might make me miss the fact that I'm not finished yet :-p Exact this is why removing expected failures at the start of each cycle is the thing to do ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
[Pharo-project] what expectedfailures expected to do?
Hi I tagged some tests as expected failures and I got a strange behavior. On the the tests which was passing was listed under the failures. When I renamed the method without updating the expected failures my bar was green. So expected failures really expect that the tests failed? We cannto have green tests in there? Stef ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Re: [Pharo-project] what expectedfailures expected to do?
Yes, if a test that is expected to fail does not fail, this is treated as a failure. I think that makes sense. Adrian On Apr 20, 2010, at 21:57 , Stéphane Ducasse wrote: Hi I tagged some tests as expected failures and I got a strange behavior. On the the tests which was passing was listed under the failures. When I renamed the method without updating the expected failures my bar was green. So expected failures really expect that the tests failed? We cannto have green tests in there? Stef ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project ___ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
