Re: superfluous GNUism in the makefile
Ivan Shmakov <[email protected]>
| Newsgroups | gmane.lisp.scheme.scheme48 |
|---|---|
| Message-ID | <[email protected]> |
>>>>> Taylor R Campbell <[email protected]> writes: >> It's fine to remove $< from the c/srfi-27.o rule, but doing so >> didn't add portability to the Makefile.in as a whole, while trying >> to achieve such portability will ruin the --srcdir= support. Which >> is, in my opinion, to high price to pay. > Wait -- is that even supposed to work now? Yes, if we define ``now'' as ``the tip of the development tree''. > I tried it in five different versions -- 0.36, 0.46, 0.57, 1.3, and > 1.8 -- and it failed in every one. In 1.8 and 1.3, it failed in the > rule for scheme48vm, not having created all the necessary object > files to link scheme48vm; in 0.57 and 0.46, it failed trying to > create c/unix/misc.o, perhaps because c/unix in the build directory > didn't exist. 0.36 came closest to working: it needed -I. in CFLAGS > and ./$(VM) rather than $(VM) to execute the new virtual machine, > both of which changes have been made in the makefile in later > versions. I vaguely remember making some fixes to the build system (like 2f466a265f68?) around the 1.8 release. > If it has never worked, how much of a price is this? > What I ran, by the way, was, in an empty directory, > /path/to/scheme48-X.Y/configure --srcdir=/path/to/scheme48-X.Y > make I've just tested it, both with GNU Make and NetBSD Make (pmake; see my other posting), and it works for me. -- FSF associate member #7257