On Wed, Sep 16, 2009 at 12:00:12AM +0200, Matthias Kilian wrote: > > > This unbreaks make install again. > > Why are some parts of the build contantly broken?
I'd been meaning to send a mail about this, as 5 or 10 people have run into problems in the last couple of weeks. I suspect the approaching 6.12 release is why so many people are trying it at the moment, rather than it just being more broken than normal. The short answer is: Because those parts of the build aren't important for people working on GHC. Fixing compiler bugs, adding new features, etc, doesn't require you to be able to install the compiler, make a binary dist, make a source tarball, etc. We could add this sort of thing to validate; in particular, I think "make binary-dist && make -C bindisttest" would be the most sensible thing to add. Currently validate takes me 27 mins (which is significantly higher than it used to be, due to growth in a number of parts of the build, and in particular building the dynamic libs). Making a binary dist takes a little over 2 mins, and installing (or bindisttest) takes another minute. In terms of disk space, an unbuilt tree is about 450M, and a built tree 1.7G. "make install" needs another 550M, and bindisttest needs another 1.1G. So adding "make binary-dist && make -C bindisttest" would be about a 10% time increase, and 65% space increase. One issue this doesn't address is that many install problems are platform specific, so the person making the patch may still not discover them. So, does anyone have any thoughts on whether we should add this to validate or not? Thanks Ian _______________________________________________ Cvs-ghc mailing list [email protected] http://www.haskell.org/mailman/listinfo/cvs-ghc
