[openib-general] Re: [PATCH] add LDFLAGS to perftest/Makefile

[Michael S. Tsirkin]

Quoting r. Nishanth Aravamudan <nacc at us.ibm.com>:
> Subject: Re: [PATCH] add LDFLAGS to perftest/Makefile
> 
> On 15.12.2005 [08:57:24 +0200], Michael S. Tsirkin wrote:
> > Quoting Nishanth Aravamudan <nacc at us.ibm.com>:
> > > Is there a reason the perftest/Makefile doesn't use LDFLAGS?
> > > Specifically, in automating userspace build & test, I put the IB
> > > libraries in a temporary directory, and exporting CFLAGS and LDFLAGS
> > > works with all other Makefiles (well, the ones I expect to work),
> but
> > > perftest does not seem to pick up my exports.
> > > 
> > > Would something like the following make sense (sorry if a different
> -p
> > > is preferred)? Or does it need to be +=?
> > > 
> > > Description: Add LDFLAGS to the perftest Makefile to allow library
> > > directories in non-standard locations to be specified.
> > 
> > Are you using gnu make? which version?
> 
> GNU Make 3.80 on SLES 9 SP2.
> 
> > Gnu make should use LDFLAGS automatically:
> > 
> > 	Linking a single object file
> > 	`N' is made automatically from `N.o' by running the linker
> > 	(usually called `ld') via the C compiler.  The precise command
> > 	used is `$(CC) $(LDFLAGS) N.o $(LOADLIBES) $(LDLIBS)'.
> 
> I thought this would be the case as well, but it didn't seem to work
> without the Makefile modification.
> 
> > > Signed-off-by: Nishanth Aravamudan <nacc at us.ibm.com>
> > > 
> > > --- Makefile	2005-12-14 14:57:04.000000000 -0800
> > > +++ Makefile.ldflags	2005-12-14 14:57:23.000000000 -0800
> > > @@ -2,6 +2,7 @@ TESTS = rdma_lat rdma_bw
> > >  
> > >  all: ${TESTS}
> > >  
> > > +LDFLAGS +=
> > >  CFLAGS += -Wall -O2 -g -D_GNU_SOURCE 
> > >  LOADLIBES += -libverbs
> > >  EXTRA_FILES = get_clock.c
> > 
> > This really does nothing. Does this patch help you?
> 
> I didn't think it should do anything either, but it did allow the make
> to work on both ppc32 and ppc64 with LDFLAGS exported in the
> environment. Without the change, the build would fail as it would not
> have the appropriate -L flags.

Looks like a work around for bug in make. I'll have a look.

-- 
MST