In article <20050430233729.GA27406@spathi.chuq.com>,
Chuck Silvers  <chuq@chuq.com> wrote:
>I've noticed for a long time now that "build.sh -j 8 release" on i386
>usually dies in the middle with errors like:
>
>...
>--- dependall-ne2000_isa ---
>ln -s /build/obj/build/src/sys/arch/i386/stand/netboot/lib .
>--- dependall-3c90xb ---
>--- __always_make_zlib ---
>--- dependall-ne2000_isa ---
>[ -d /build/obj/build/src/sys/arch/i386/stand/netboot/lib ] || mkdir
>/build/obj/build/src/sys/arch/i386/stand/netboot/lib
>--- dependall-pcnet_isapnp ---
>--- lib ---
>ln: ./lib: File exists
>--- dependall-ne2000_isa ---
>--- __always_make_kernlib ---
>--- dependall-3c590 ---
>--- netif_small.o ---
>--- dependall-pcnet_isapnp ---
>*** [lib] Error code 1
>
>
>more recently I noticed that this problem is described by these PRs:
>
>9566	.BEGIN target does not follow dependencies
>9567	.BEGIN targets use depencencies
>
>
>so should we apply the changes from the PRs, or does someone want to change
>make to actually process dependencies of .BEGIN?
>
>here's the list of makefiles under sys/arch that have this problem:
>
>./arc/stand/boot/Makefile:.BEGIN: machine mips
>./evbarm/stand/gzboot/Makefile.gzboot:.BEGIN: machine
>./hp700/stand/Makefile.inc:.BEGIN: machine hp700 hppa
>./i386/stand/boot/Makefile.boot:.BEGIN: machine x86 lib
>./i386/stand/bootxx/Makefile.bootxx:.BEGIN: machine x86 lib
>./i386/stand/mbr/Makefile.mbr:.BEGIN: machine x86
>./i386/stand/Makefile.booters:.BEGIN: machine x86 lib
>./mvme68k/stand/installboot/Makefile:.BEGIN: machine
>./news68k/stand/boot/Makefile:.BEGIN: machine m68k
>./news68k/stand/bootxx/Makefile:.BEGIN: machine m68k
>./prep/stand/boot/Makefile:.BEGIN: machine powerpc
>./sun68k/stand/Makefile.inc:.BEGIN: machine m68k sun68k
>
>
>I see that at least one of the instances mentioned in the PR,
>pmax/stand/Makefile.booters, has been fixed differently than the PR suggests.
>it would be good to be consistent with this.
>
>if we opt to leave .BEGIN the way it is now and not process its dependencies,
>can we at least make it an error to specify dependencies for .BEGIN?

We should....

christos