> This is a patch for net/core/Makefile to link only those files in
> net/core/core.o (not being a kernel guru, I really don't know if
> it's ok, but at least it compiles and has worked a couple of days)
OOPS. Another thing that I broke when I did the Makefiles. Thanks for the
patch.
> After that I had some more problems... I couldn't compile anything
> with CONFIG_MODVERSIONS (I don't like to recompile modules every time,
I think the following patch will fix this problem. My long term fix is to
change the makefile target "fastdep" back into "dep". The only reason I
changed it in the first place is so I could have things compile with both
targets in the Makefiles as I was converting them. Anyway, let me know if
the following patch does not fix MODVERSIONS:
diff -N -ru linux.orig/kernel/Makefile linux/kernel/Makefile
--- linux.orig/kernel/Makefile Wed Aug 16 07:57:02 1995
+++ linux/kernel/Makefile Fri Aug 18 14:04:32 1995
@@ -20,6 +20,8 @@
ifdef CONFIG_MODVERSIONS
$(O_TARGET): $(SYMTAB_OBJS:.o=.ver)
+dep: $(SYMTAB_OBJS:.o=.ver)
+fastdep: $(SYMTAB_OBJS:.o=.ver)
endif
include $(TOPDIR)/Rules.make
> having a 386Dx/25 with only 4Mb... BTW it would be nice to select
> which modules to compile).
Thats the next thing on my list.
> Also, the Makefiles have some problems with modules in the fs
> directory... the "Hack for fs subdirectories" in Rules.make works
> iff I have defined CONFIG_BINFMT_ELF (can't chdir to binfmt_elf.o ;-)
> Here's a patch that solves this:
Thanks again. I keep thinking about building a script that compiles kernels
with random configurations to see if I can find these problems. I obviously
have not done that yet :-)
Jim