On Mon, Jan 26, 2009 at 02:09:46PM +0000, Simon Marlow wrote: > Thorkil Naur wrote: > > >>Can you check it's the root ghc.mk it's complaining about please, by > >>adding a blank line to the start of the file and checking that the line > >>number changes? > >> > >>(just rerunning > >> make -r --no-print-directory -f ghc.mk phase=0 just-makefiles > >>should be enough to get the error again) > > > >The problem was definitely reported for the root ghc.mk. The make-3.81 > >release notes mentioned correcting a couple of errors in make-3.80 related > >to symbol expansion and continuation lines. I guess requiring make-3.81 is > >one way to get around this problem. In any case, it appears that the > >buildbot slave on this machine has run into a similar problem > >(http://darcs.haskell.org/buildbot/all/builders/tnaur%20x86%20Linux%20head/builds/309/steps/boottestsuite/logs/stdio), > > > >so I'll make the make-3.81 install the preferred one on this machine. > > If we have a dependency on make 3.81, we ought to check that in configure > (and document it). Ian - would you mind doing that? Thanks!
Hmm, we can't do it in configure, because we don't know if you're about to type "make", "gmake" or "/path/to/make". We can put something like this in Makefiles: ifeq "$(MAKE_VERSION)" "3.80" $(error GNU make 3.80 has bugs that make it unable to build GHC. Please upgrade to 3.81 or higher.) endif (Thorkil, do you still have 3.80 installed anywhere? If so, can you check if that works properly please? I'm not sure if MAKE_VERSION will be 3.80 or 3.8. Just put those 3 lines in an empty Makefile, run "make", and see if it fails with that error message) However, we can't put that in any Makefile that exhibits the bug, or you'll just get the "missing `endif'" error instead. Putting it in only the top-level Makefile for GHC and the testsuite might be enough. Alternatively, we could look into whether we can work around the bugs in 3.80. If we do this, we should install it on one of the MSR builders and have a build that explicitly uses it; otherwise, should Thorkil upgrade to 3.81 at some point, we won't spot any regressions. Thanks Ian _______________________________________________ Cvs-ghc mailing list [email protected] http://www.haskell.org/mailman/listinfo/cvs-ghc
