Re: dmd testsuite naming scheme
On Tue, 03 Jan 2012 19:57:14 +0100, Alex Rønne Petersen xtzgzo...@gmail.com wrote: On 03-01-2012 16:44, Martin Nowak wrote: On Tue, 03 Jan 2012 15:39:34 +0100, Alex Rønne Petersen xtzgzo...@gmail.com wrote: On 03-01-2012 13:36, Trass3r wrote: I honestly wouldn't know where to add or search for a test case. It doesn't really matter where they go. A collection of test cases with a theme to them might go in a named file, a random one might be appended to any of the test* files. Won't this potentially lead to test duplication? Considering that the testsuite already takes quite some time to run this isn't a desirable trend imho. Test duplication isn't necessarily a bad thing. In my experience, it often happens that a tiny difference between two seemingly equal tests can be all that matters. On the other hand, grouping tests into files based on language features might be a good idea. If anything, to be able to navigate the test suite. - Alex There is some opportunity in creating systematic feature tests backed with coverage analysis. There are still too many uncovered areas. This not only helps to find remaining bugs but gives a specification like overview of a feature state. I still say D needs a formal specification more than a test suite as some kind of excuse for a specification. (And no, I don't consider d-p-l.org a spec; a guide at best.) - Alex I'd still like to see that the website, language specification and specification tests become the same.
Re: dmd testsuite naming scheme
Definitely! +1 On Tue, Jan 3, 2012 at 11:18 PM, Andrei Alexandrescu seewebsiteforem...@erdani.org wrote: On 1/3/12 12:57 PM, Alex Rønne Petersen wrote: On 03-01-2012 16:44, Martin Nowak wrote: On Tue, 03 Jan 2012 15:39:34 +0100, Alex Rønne Petersen xtzgzo...@gmail.com wrote: On 03-01-2012 13:36, Trass3r wrote: I honestly wouldn't know where to add or search for a test case. It doesn't really matter where they go. A collection of test cases with a theme to them might go in a named file, a random one might be appended to any of the test* files. Won't this potentially lead to test duplication? Considering that the testsuite already takes quite some time to run this isn't a desirable trend imho. Test duplication isn't necessarily a bad thing. In my experience, it often happens that a tiny difference between two seemingly equal tests can be all that matters. On the other hand, grouping tests into files based on language features might be a good idea. If anything, to be able to navigate the test suite. - Alex There is some opportunity in creating systematic feature tests backed with coverage analysis. There are still too many uncovered areas. This not only helps to find remaining bugs but gives a specification like overview of a feature state. I still say D needs a formal specification more than a test suite as some kind of excuse for a specification. (And no, I don't consider d-p-l.org a spec; a guide at best.) - Alex Agreed. Andrei -- Bye, Gor Gyolchanyan.
Re: dmd testsuite naming scheme
I honestly wouldn't know where to add or search for a test case. It doesn't really matter where they go. A collection of test cases with a theme to them might go in a named file, a random one might be appended to any of the test* files. Won't this potentially lead to test duplication? Considering that the testsuite already takes quite some time to run this isn't a desirable trend imho.
Re: dmd testsuite naming scheme
On 03-01-2012 13:36, Trass3r wrote: I honestly wouldn't know where to add or search for a test case. It doesn't really matter where they go. A collection of test cases with a theme to them might go in a named file, a random one might be appended to any of the test* files. Won't this potentially lead to test duplication? Considering that the testsuite already takes quite some time to run this isn't a desirable trend imho. Test duplication isn't necessarily a bad thing. In my experience, it often happens that a tiny difference between two seemingly equal tests can be all that matters. On the other hand, grouping tests into files based on language features might be a good idea. If anything, to be able to navigate the test suite. - Alex
Re: dmd testsuite naming scheme
On Tue, 03 Jan 2012 15:39:34 +0100, Alex Rønne Petersen xtzgzo...@gmail.com wrote: On 03-01-2012 13:36, Trass3r wrote: I honestly wouldn't know where to add or search for a test case. It doesn't really matter where they go. A collection of test cases with a theme to them might go in a named file, a random one might be appended to any of the test* files. Won't this potentially lead to test duplication? Considering that the testsuite already takes quite some time to run this isn't a desirable trend imho. Test duplication isn't necessarily a bad thing. In my experience, it often happens that a tiny difference between two seemingly equal tests can be all that matters. On the other hand, grouping tests into files based on language features might be a good idea. If anything, to be able to navigate the test suite. - Alex There is some opportunity in creating systematic feature tests backed with coverage analysis. There are still too many uncovered areas. This not only helps to find remaining bugs but gives a specification like overview of a feature state.
Re: dmd testsuite naming scheme
On 03-01-2012 16:44, Martin Nowak wrote: On Tue, 03 Jan 2012 15:39:34 +0100, Alex Rønne Petersen xtzgzo...@gmail.com wrote: On 03-01-2012 13:36, Trass3r wrote: I honestly wouldn't know where to add or search for a test case. It doesn't really matter where they go. A collection of test cases with a theme to them might go in a named file, a random one might be appended to any of the test* files. Won't this potentially lead to test duplication? Considering that the testsuite already takes quite some time to run this isn't a desirable trend imho. Test duplication isn't necessarily a bad thing. In my experience, it often happens that a tiny difference between two seemingly equal tests can be all that matters. On the other hand, grouping tests into files based on language features might be a good idea. If anything, to be able to navigate the test suite. - Alex There is some opportunity in creating systematic feature tests backed with coverage analysis. There are still too many uncovered areas. This not only helps to find remaining bugs but gives a specification like overview of a feature state. I still say D needs a formal specification more than a test suite as some kind of excuse for a specification. (And no, I don't consider d-p-l.org a spec; a guide at best.) - Alex
Re: dmd testsuite naming scheme
On 1/3/12 12:57 PM, Alex Rønne Petersen wrote: On 03-01-2012 16:44, Martin Nowak wrote: On Tue, 03 Jan 2012 15:39:34 +0100, Alex Rønne Petersen xtzgzo...@gmail.com wrote: On 03-01-2012 13:36, Trass3r wrote: I honestly wouldn't know where to add or search for a test case. It doesn't really matter where they go. A collection of test cases with a theme to them might go in a named file, a random one might be appended to any of the test* files. Won't this potentially lead to test duplication? Considering that the testsuite already takes quite some time to run this isn't a desirable trend imho. Test duplication isn't necessarily a bad thing. In my experience, it often happens that a tiny difference between two seemingly equal tests can be all that matters. On the other hand, grouping tests into files based on language features might be a good idea. If anything, to be able to navigate the test suite. - Alex There is some opportunity in creating systematic feature tests backed with coverage analysis. There are still too many uncovered areas. This not only helps to find remaining bugs but gives a specification like overview of a feature state. I still say D needs a formal specification more than a test suite as some kind of excuse for a specification. (And no, I don't consider d-p-l.org a spec; a guide at best.) - Alex Agreed. Andrei
Re: dmd testsuite naming scheme
On 1/2/2012 9:58 AM, Trass3r wrote: Is there any pattern in the testsuite organization? No. There are loads of test[0-9]+. files etc. And folders are only used to group compilable/runnable... I honestly wouldn't know where to add or search for a test case. It doesn't really matter where they go. A collection of test cases with a theme to them might go in a named file, a random one might be appended to any of the test* files.
