Re: building faillure

[Phil Holmes]

----- Original Message ----- 
From: "David Kastrup" <address@hidden>
To: <address@hidden>
Cc: <address@hidden>
Sent: Thursday, September 22, 2011 3:47 PM
Subject: Re: building faillure


> "Phil Holmes" <address@hidden> writes:
>
> > ----- Original Message ----- 
> > From: "Graham Percival" <address@hidden>
> > To: "Phil Holmes" <address@hidden>
> > Cc: "Bernardo Barros" <address@hidden>; "Devel"
> > <address@hidden>; "lilypond-user" <address@hidden>
> > Sent: Tuesday, September 20, 2011 11:29 PM
> > Subject: Re: building faillure
> >
> >
> > > On Tue, Sep 20, 2011 at 10:08:26AM +0100, Phil Holmes wrote:
> > > > $(outdir)/general-scheme.o: $(outdir)/version.hh
> > > > $(outdir)/lily-guile.o: $(outdir)/version.hh
> > > > $(outdir)/lily-version.o: $(outdir)/version.hh
> > > ...
> > > > Graham: git grep version.hh gives:
> > > >
> > > > lily/general-scheme.cc:#include "version.hh"
> > > > lily/lexer.ll:#include "version.hh"
> > > > lily/lily-guile.cc:#include "version.hh"
> > > > lily/lily-version.cc:#include "version.hh"
> > > > lily/main.cc:#include "version.hh"
> > > > lily/relocate.cc:#include "version.hh"
> > > > lily/warn-scheme.cc:#include "version.hh"
> > > >
> > > > I think it could make the build system more robust to make ordering
> > > > by adding lexer.ll, main.cc, relocate.cc and warn-scheme.cc to the
> > > > target list for version.hh?
> > >
> > > Yes please.  Add the lines, check that you can compile from
> > > scratch, then push directly.
> >
> > Now pushed as 29447b3a224f52444f0ec74225eb9e6af0591223
>
> There is no such thing as lexer.ll.o and most of the stuff seems already
> mentioned in out/*.dep.  Isn't this activism?  I think the following is
> much more glaring:

Apologies.  I had noticed and corrected that in the patch I intended to 
push, but my problems with git push meant my correction got lost.  I'll get 
rid of the line, since lexer actually has a rule a bit higher up.

If you can find the mail trail, you'll see that this does actually correct a 
failing build on Fedora, owing to the missing rule for relocate.

> # list parser.hh first: making parser.hh removes parser.cc
> OUT_DIST_FILES=$(addprefix $(outdir)/,parser.hh parser.cc)
>
>
> We can't really have a situation like that described in the comment if
> we want to have parallel compilation to work.
>
> Either we need a _single_ rule for parser.hh and parser.cc (not much
> else makes sense), or one needs to juggle with -o and -d in order to
> make sure that generating the header file does not stomp over the C file
> and vice versa.  Since non-matching output files are a bad idea anyway,
> I'd prefer a single rule.

I've no problem with that being corrected, too.


--
Phil Holmes