[Cluster-devel] Re: Problems with recent Makefile changes

[Fabio Massimo Di Nitto]

Robert Peterson wrote:
> Fabio Massimo Di Nitto wrote:
>> It would be lovely if you could show me the full error next time it happens
>> and if you have a way to reproduce it.
>>
>> Thanks
>> Fabio
>>
> Hi Fabio,
> 
> The problem seems to be that all the include files are somehow specified on 
> the gcc command if only one file changes.  Thus, if I change rgrp.c and do 
> a make, I get:
> 
> [root at trin-10 ../cluster/gfs/gfs_fsck]# make
> gcc -Wall -I/home/devel/HEAD.intermediate/cluster/config -D_FILE_OFFSET_BITS=64 -DHELPER_PROGRAM -DGFS_RELEASE_NAME=\"DEVEL.1178030462\" -MMD -O2 -I../include -c -o rgrp.o rgrp.c util.h fsck_incore.h ondisk.h ../include/osi_list.h ../include/osi_user.h ../include/global.h ../include/linux_endian.h fs_bits.h rgrp.h fsck.h log.h block_list.h bitmap.h bio.h fs_inode.h inode.h
> gcc: cannot specify -o with -c or -S with multiple files
> make: *** [rgrp.o] Error 1
> 
> However, I can do "make clean ; make" and it works just fine. 
> This is a gfs userland example, but I think gfs2 is the same.
> 
> Regards,
> 
> Bob Peterson
> Red Hat Cluster Suite

This is pretty special, but i am ready to bet that you have seen the problem
only in gfs/gfs_fsck and gfs2/fsck and I am now able to reproduce this by
issuing twice in a row:
make clean all

I am going to investigate it, but I would like to know if there is a special
reason why we use -MMD and -include for *.d files (that are actually leaked in
the above command and hence the world fall apart in an endless blackhole).

Thanks
Fabio

-- 
I'm going to make him an offer he can't refuse.