From bug-octave-request  Fri May 21 12:27:38 1993
Subject: bug report?
From: Operator <root at ptolemy dot arc dot nasa dot gov>
To: bug-octave
Date: Fri, 21 May 93 10:30:13 PDT

Dear blackhole:

This is in regards to a previous message regarding a apparent linking problem.
The problem is actually in the ./libcruft/Makefile. 

There seems to be a problem with the Makefile in ./libcruft. The make log file
indicates it's archiving FORTRAN source whose behavior appears to be governed
by the following lines in ./libcruft/Makefile 

	CRUFT_FSRC = $(foreach dir, $(SUBDIRS), $(wildcard $(srcdir)/$(dir)/*.f))
	CRUFT_OBJ2 = $(patsubst $(srcdir)/%.f, %.o, $(CRUFT_FSRC))
	CRUFT_OBJ1 = $(subst misc/d1mach.o, , $(CRUFT_OBJ2))
	CRUFT_OBJ = $(CRUFT_OBJ1) misc/d1mach.o
 
	../libcruft.a: $(CRUFT_OBJ)
        	rm -f ../libcruft.a
        	$(AR) $(ARFLAGS) ../libcruft.a $(CRUFT_OBJ)
        	$(RANLIB) ../libcruft.a
 
Given the above Makefile configuration, the output from make in ./libcruft  
yields:

	rm -f ../libcruft.a
	ar rv ../libcruft.a  blas/dasum.f blas/daxpy.f ....etc 

which is at best NONSENSE. 

Since I don't have a manual for GNU make handy, I changed CRUFT_FSRC from 

  	CRUFT_FSRC = $(foreach dir, $(SUBDIRS), $(wildcard $(srcdir)/$(dir)/*.f))

to 

  	CRUFT_FSRC = $(foreach dir, $(SUBDIRS), $(wildcard $(srcdir)/$(dir)/*.o))

and I was able to generate a libcruft.a archive and complete the linking.


===============================================================================
Kenneth Simpson                        		NASA
Internet: simpson at ptolemy dot arc dot nasa dot gov         	Ames Research Center, MS/269-1
UUCP: ames!ptolemy!simpson      		Moffett Field, CA 94035-1000  
AT&T: 1-415-604-3593                        	USA, Earth      
===============================================================================