On 26/05/14 20:43, Andrew Wagner wrote: > I don't follow the question... why is execution tied to linking in your > setup? Also, if each test is a separate executable, it sounds like your > build is already extremely modular... > > If your tests are generating a report file or something, you could add the > running of each test as a rule that depends on the test executable (along > with any other input data). If you are using some magic inside your > application to run tests at dynamic link time, perhaps this would be > cleaner... > > If you are trying to temporarily switch off building and running of the > tests, you can use a configuration variable and ifdefs in your Tupfiles to > add/remove them from your build...
They're having the tests run by Tup as part of the build process. It makes sense to me, since it means that tests that ran successfully and are unaffected by the new modifications won't be re-run (although there are downsides to this, such as the loss of that warm fuzzy feeling you get from seeing 20 green "test passed" messages!). Execution is tied to linking in that tup obviously re-runs the executable if it has been modified (ie. re-linked).
signature.asc
Description: OpenPGP digital signature
