This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: gcc/Makefile.in dependency messup
On Wed, May 26, 2010 at 10:48 PM, Steven Bosscher <stevenb.gcc@gmail.com> wrote:
> On Thu, May 27, 2010 at 7:29 AM, Ralf Wildenhues <Ralf.Wildenhues@gmx.de> wrote:
>> Hello,
>>
>> one of the recent patches to the build system broke dependencies:
>>
>> make[3]: Entering directory `/tmp/build-gcc/gcc'
>> make[3]: Circular build/gencondmd.o <- insn-flags.h dependency dropped.
>> make[3]: Circular build/gencondmd.o <- insn-flags.h dependency dropped.
>>
>> Something like this is bad, because it usually has the potential to
>> break parallel builds in subtle ways. ?Please check for such output when
>> bootstrapping. ?It is very simple to do afterwards, too, in an up to
>> date build tree:
>>
>> $ cd gcc && make
>> make: Circular build/gencondmd.o <- insn-flags.h dependency dropped.
>> make: Circular build/gencondmd.o <- insn-flags.h dependency dropped.
>> $
>
> This is probably mine, gencondmd depends on FUNCTION_H, which includes
> $TM_H, which includes insn-flags.h. ?The TM_H dependency is new in
> Makefile, but it always really existed. I think this "worked" before
> only because everything that includes function.h already inherited the
> TM_H dependency in other ways.
>
> Could you please try and see if removing TM_H from FUNCTION_H removes
> the warning from make?
>
> The fix would be to make function.h independent of tm.h again. ?Right
> now, function.h includes tm.h for CUMULATIVE_ARGS, but in fact the
> whole part of function.h with rtl data structure definitions could be
> split off to another place (function-rtl.h, or emit-rtl.h -- tbd).
>
> Sorry for the inconvenience.
>
I opened
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44315
to keep track of it.
--
H.J.