8:23 a.m.
Modules always fail to build for me on FreeBSD, turns out that the
Makefile needed a little tweaking (gnu make works fine). Not tested
anywhere other than FreeBSD.
2003-01-30 Robert Pluim <rpluim(a)bigfoot.com>
* Makefile.common: FreeBSD make compatability fixes, expand
SRC_SRCS immediately, use MODCC rather than CC.
PS Please CC me on any replies, I'm no longer on xemacs-beta
Robert
Index: Makefile.common
===================================================================
RCS file: /pack/xemacscvs/XEmacs/xemacs/modules/common/Makefile.common,v
retrieving revision 1.2
diff -u -r1.2 Makefile.common
--- Makefile.common 2002/11/18 06:52:30 1.2
+++ Makefile.common 2003/01/30 08:06:59
＠＠ -27,7 +27,7 ＠＠
## Note: This will be appended to the individual module Makefiles by configure.
-SRC_SRCS=$(SRCS:%=$(srcdir)/%)
+SRC_SRCS:=$(SRCS:%=$(srcdir)/%)
OBJS=$(SRCS:.c=.o)
SHELL=/bin/sh
＠＠ -47,7 +47,7 ＠＠
srcdir=＠srcdir＠
VPATH=＠srcdir＠
-CC=＠MOD_CC＠
+MODCC=＠MOD_CC＠
MODARCHDIR=＠MODARCHDIR＠
MAKE_DOCFILE=＠MAKE_DOCFILE＠
MODCFLAGS=＠MODCFLAGS＠
＠＠ -59,13 +59,13 ＠＠
all: $(OBJECT_TO_BUILD)
.c.o:
- $(CC) $(MODCFLAGS) -c $<
+ $(MODCC) $(MODCFLAGS) -c $<
$(MODNAME).ell: $(OBJS) $(MODNAME)_i.o
- $(CC) --mode=link --mode=verbose --mod-output=$＠ $^ $(LDFLAGS)
+ $(MODCC) --mode=link --mode=verbose --mod-output=$＠ $^ $(LDFLAGS)
$(MODNAME)_i.c: $(SRCS)
- ELLMAKEDOC=$(MAKE_DOCFILE) $(CC) --mode=init --mod-output=$＠ \
+ ELLMAKEDOC=$(MAKE_DOCFILE) $(ELLCC) --mode=init --mod-output=$＠ \
--mod-name=$(MODNAME) --mod-version=$(MODVER) \
--mod-title=$(MODTITLE) $(SRC_SRCS)
--
4:13 p.m.
Hi Robert,
Thank you for your submission.
Robert Pluim <rpluim(a)bigfoot.com> wrote:
Modules always fail to build for me on FreeBSD, turns out that the
Makefile needed a little tweaking (gnu make works fine). Not tested
anywhere other than FreeBSD.
The SRC_SRCS change seems a little strange to me, but I can well believe
that some make might need to do that. However, the other changes look
like noops to me. Or are you saying that the *name* CC is hardwired
into FreeBSD make, and that we just need to choose something, anything,
different from CC to make it work? Also, why the one use of ELLCC
instead of MODCC?
Thanks for your help with this.
--
Jerry James
http://www.ittc.ku.edu/~james/
9:06 a.m.
Jerry James writes:
Hi Robert,
Thank you for your submission.
Robert Pluim <rpluim(a)bigfoot.com> wrote:
> Modules always fail to build for me on FreeBSD, turns out that the
> Makefile needed a little tweaking (gnu make works fine). Not tested
> anywhere other than FreeBSD.
The SRC_SRCS change seems a little strange to me, but I can well believe
that some make might need to do that.
If SRC_SRCS is assigned using '=', then the ${srcdir} doesn't get
expanded when SRC_SRCS is used in the actual rule later on. With
':=' it is expanded at assignment and everything works fine.
However, the other changes look
like noops to me. Or are you saying that the *name* CC is hardwired
into FreeBSD make, and that we just need to choose something, anything,
different from CC to make it work?
make is invoked recursively by the top-level makefile something like
'cd modules/foo CC=gcc', which overrides the subdir makefile's setting
of CC. Hmm, I've just noticed that the top-level GNUmakefile
_doesn't_ do that, it just recurses into the sub-directory. I have no
idea why. [1]
Also, why the one use of ELLCC
instead of MODCC?
Oops, they should all be MODCC, patch redone and attached.
ChangeLog:
2003-01-30 Robert Pluim <rpluim(a)bigfoot.com>
* Makefile.common: FreeBSD make compatability fixes, expand
SRC_SRCS immediately, use MODCC rather than CC.
Index: Makefile.common
===================================================================
RCS file: /pack/xemacscvs/XEmacs/xemacs/modules/common/Makefile.common,v
retrieving revision 1.2
diff -u -r1.2 Makefile.common
--- Makefile.common 2002/11/18 06:52:30 1.2
+++ Makefile.common 2003/01/31 08:43:50
＠＠ -27,7 +27,7 ＠＠
## Note: This will be appended to the individual module Makefiles by configure.
-SRC_SRCS=$(SRCS:%=$(srcdir)/%)
+SRC_SRCS:=$(SRCS:%=$(srcdir)/%)
OBJS=$(SRCS:.c=.o)
SHELL=/bin/sh
＠＠ -47,7 +47,7 ＠＠
srcdir=＠srcdir＠
VPATH=＠srcdir＠
-CC=＠MOD_CC＠
+MODCC=＠MOD_CC＠
MODARCHDIR=＠MODARCHDIR＠
MAKE_DOCFILE=＠MAKE_DOCFILE＠
MODCFLAGS=＠MODCFLAGS＠
＠＠ -59,13 +59,13 ＠＠
all: $(OBJECT_TO_BUILD)
.c.o:
- $(CC) $(MODCFLAGS) -c $<
+ $(MODCC) $(MODCFLAGS) -c $<
$(MODNAME).ell: $(OBJS) $(MODNAME)_i.o
- $(CC) --mode=link --mode=verbose --mod-output=$＠ $^ $(LDFLAGS)
+ $(MODCC) --mode=link --mode=verbose --mod-output=$＠ $^ $(LDFLAGS)
$(MODNAME)_i.c: $(SRCS)
- ELLMAKEDOC=$(MAKE_DOCFILE) $(CC) --mode=init --mod-output=$＠ \
+ ELLMAKEDOC=$(MAKE_DOCFILE) $(MODCC) --mode=init --mod-output=$＠ \
--mod-name=$(MODNAME) --mod-version=$(MODVER) \
--mod-title=$(MODTITLE) $(SRC_SRCS)
Footnotes:
[1] Why do we try to stay compatible with GNU make _and_ generic make?
Why don't we just pick one and stick with it?
--
Mobile: +33 6 19 49 54 48
Tel: +33 4 92 94 05 94
11:31 p.m.
Robert Pluim <rpluim(a)bigfoot.com> wrote:
If SRC_SRCS is assigned using '=', then the ${srcdir}
doesn't get
expanded when SRC_SRCS is used in the actual rule later on. With
':=' it is expanded at assignment and everything works fine.
Okay, I understand.
make is invoked recursively by the top-level makefile something like
'cd modules/foo CC=gcc', which overrides the subdir makefile's setting
of CC. Hmm, I've just noticed that the top-level GNUmakefile
_doesn't_ do that, it just recurses into the sub-directory. I have no
idea why. [1]
I'm on a Linux system, and therefore using GNU make, which is why I
haven't noticed this behavior before. I'll have to remember to test
stuff out on a Solaris machine with the native make in the future.
It would feel cleaner to me in some ways to not have the top-level
Makefile override the value of CC, but I don't actually know how to do
that without jeopardizing the build in src and lwlib, so I'm happy to go
with your approach. Would you mind sending your patch one more time,
but to xemacs-patches(a)xemacs.org this time? (Incidentally, do you have
commit privileges?)
Thanks,
--
Jerry James
http://www.ittc.ku.edu/~james/
5:23 p.m.
Jerry James writes:
It would feel cleaner to me in some ways to not have the top-level
Makefile override the value of CC, but I don't actually know how to do
that without jeopardizing the build in src and lwlib, so I'm happy to go
with your approach. Would you mind sending your patch one more time,
but to xemacs-patches(a)xemacs.org this time? (Incidentally, do you have
commit privileges?)
I'll resend the patch (and no, I don't have commit privileges, nor
would I want them).
Robert
--
7755
days inactive
7757
days old
4 comments
2 participants
participants (2)
-
Jerry James -
Robert Pluim