On Thu, Feb 21, 2013 at 10:09 AM, Kornel Benko <kor...@lyx.org> wrote: > Am Donnerstag, 21. Februar 2013 um 09:45:16, schrieb Scott Kostyshak > <skost...@lyx.org> > >> > >> In the case of knitr, it's probably better to do something like > >> checking the exit code of the following command: > >> R -e "library(knitr)" > >> Ideally, we would also check > >> R -e "packageVersion(\"knitr\")" > >> To make sure it is recent enough. > > > > Good idea. > > > >> >> Another possibility is that it's up to the users which tests they > >> > > >> >> should run. For example, I could have local .ctestignore where I list > >> > > >> >> the > >> > > >> >> tests for which I don't meet the dependencies. ctest -R would not run > >> > > >> >> the tests listed in .ctestignore. > >> > > >> > > >> > > >> > Where should the ".ctestignore" be? Why not have the info directly in > >> > CMakeLists.txt? > >> > >> This was for if it was up to the users to decide which tests they > >> ignored. I would make .ctestignore and you would have your own one. > >> Ours would be different because we have different dependencies > >> installed. Of course, if CMakeLists.txt can do everything > >> automatically as we're discussing above, that would be nice. > > > > Ctest does not care, so we have to do it at config time to get > > it into the ctest files. It looks more natural to me, to do it the cmake > way. > > > > To ignore a test like knitr I would call > > #ctest -R export/examples -E knitr -E sweave > > > > As long as cmake can get things automatically, it should do so.
OK. It might be nice for ctest to remind us which tests were not added. But it doesn't seem like this is easy with ctest (you either add a test or don't add a test right?). Using message() doesn't seem right either though because the output should be with the other tests. I'm just worried that we will forget about some tests just because we don't have the dependencies whereas if we are reminded that we are ignoring tests we are more likely to remember to find someone who can test them before a release, for example. Maybe this shouldn't be a concern. Scott > > > >> Scott > > > > Kornel
