hi all, have you seen this? http://clam.iua.upf.edu/testfarm/ardour2/ any developer wants to do a (void) commit to see if testfarm awakes? parumi: scary :) sampo_v2, is testfarm no bigbrother :) parumi: ;) parumi: does it start the build with scons? sampo_v2, yes. look at the execution log. i thing its all quite simple parumi: then the commit shouldn't be void as scons won't do anything sampo_v2, the testfarm client is periodically checking for new commits. with "void" i mean a trivial change in a comment... parumi: scons doesn't do anything if the code hasn't really changed parumi: let's test it sampo_v2, now i get you! anyway you'll get a new box. a quick one parumi: how long does it take for the farm to build? sampo_v2, you mean to awake, now it's set to 5 minutes (too long). then the server is publishing data each 2 minutes. parumi: i mean, you said it was fast: how fast does it take to build 2.0 ? sampo_v2, considering that the build will take nothing we will probably not see the box in yellow (in-progress), but a forth green box. lets wait parumi: i'll do a very significant change (from scons point of view). parumi: commit done ... now sampo_v2, to build? well it depends on the change because i'm not doing a scons -c. but hey, i placed a diagram of the building time! parumi: what is on the y axis? seconds? sampo_v2, the first (clean) build was 2500 secons. let's see now parumi: that is not fast at all sampo_v2, clic,clac,clic... sampo_v2, do you want less delay? i can set shorter polling. but my experience is that usually testfarm is on the background sampo_v2, awaken now parumi: depends on what is the farm for. if the farm is for testing whether commits compile, 5 minutes is good sampo_v2, well we use it for very long tests, usually >30min, and if green, it do packaging and uploads to the web. parumi: have you talked to las about this? sampo_v2, you can see here the kind of scripts we run http://clam.iua.upf.edu/testfarm/ sampo_v2, no. and he is on holidays i think sampo_v2, the power of this is when it is multiplatform, of course parumi: he is yes, i was wondering whether you talked to him before parumi: of course sampo_v2, do you have already some similar scripts? i remember las saying that svn is guaranteed that builds, or something parumi: we don't sampo_v2, good, now you do :-) parumi: but las is right. commits rarely break compilation parumi: what box is that running on? sampo_v2, yes that's why i started a thread on regression testing in the ml. sampo_v2, P4 500Mb Linux ubuntu 2.6.15-25-386 sampo_v2, i really don't remember other details is a very cheap box ok any osx ardour-dev here to explain how you compile both intel and ppc archs? parumi: essej and taybin are the people you are looking for sampo_v2, i know :) sampo_v2, look it finished with 1000 sec. half a clean build parumi: hi taybin, here you are! parumi: the secret is to have both a intel and ppc box taybin, shit! parumi: you can cross compile, but it is such a pain in the ass none of us have successfully done it parumi: because you have to cross compile *every single dependent library* taybin, yes. but i'll have to do it for clam apps. i don't have a ppc taybin: doesn't darwinports (or anyone else for that matter) provide cross compilation libraries? sampo_v2: no, no they don't taybin: weird parumi: but you should read this: http://developer.apple.com/technotes/tn2005/tn2137.html parumi: that'll get you most of the way there parumi: it's not super difficult, but it can be tricky is it possible to have parallel darwinports installs? parumi: the hard part is telling darwinports to cross compile everything imagine having a osx testfarm client that uploads a dmg on each commit. it would be great parumi: you have to edit every package's configure.args file taybin, i guess you can specify crosscompile options there. but install prefix also? parumi: maybe. that might be a global darwinports settings parumi: is there anything you would want to reconfigure in your build farm before i revert the previous commit (with another commit)? parumi: the secret is to sneak "-arch x86 -arch ppc" into the CFLAGS envvar parumi: and also another setting to LDFLAGS sampo_v2, not really parumi: it might be easier to not use darwinports at all taybin, thats useful. but will they install in separate dirs automatically? parumi: will what? parumi: change reverted parumi: the cross compiling creates a single file with both platform binaries inside it taybin, aha! parumi: read the link I sent you taybin, (reading) why you say it might be easier without darwinports? parumi: it depends on the number of dependencies clam has parumi: if it's just a few, you can just run the ./configure for the dependencies manually instead of trying to get darwinports to do what you want taybin, ok. i'll investigate further. time to dinner. thanks for your help!