Re: make fails for staging branch

[Graham Percival]

On Thu, Dec 22, 2011 at 01:35:27PM +0100, David Kastrup wrote:
> David Kastrup <address@hidden> writes:
> 
> >> In file included from 
> >> /home/jlowe/lilypond-git/lily/lily-lexer-scheme.cc:21:
> >> /home/jlowe/lilypond-git/lily/include/lily-lexer.hh:69: error: ISO C++
> >> forbids declaration of 'YYSTYPE' with no type
> >> /home/jlowe/lilypond-git/lily/include/lily-lexer.hh:69: error:
> >> expected ';' before '*' token
> >
> > That looks like something I would have committed.  I'll research.
> 
> Does not make all that much sense: it complains about YYSTYPE not being
> defined.  But YYSTYPE is defined in lily/out/parser.hh (and has been
> defined there before) and lily-lexer.hh
> has
> #include "parser.hh"

Looking in Patchy's logs, I see:

In file included from
/main/large-tmp/lilypond-autobuild/lily/lily-lexer-scheme.cc:21:
/main/large-tmp/lilypond-autobuild/lily/include/lily-lexer.hh:28:21:
error: parser.hh: No such file or directory
bison -d -o out/parser.cc
/main/large-tmp/lilypond-autobuild/lily/parser.yy
In file included from
/main/large-tmp/lilypond-autobuild/lily/lily-lexer-scheme.cc:21:
/main/large-tmp/lilypond-autobuild/lily/include/lily-lexer.hh:69:
error: ISO C++ forbids declaration of 'YYSTYPE' with no type


That order of commands suggests that lily-lexer-scheme.cc does not
depend on parser.cc.  I see that it contains such an #include, so
I would assume that make would catch it... but apparently not.

hmm... are the dependencies regenerated all the time?  What if
that #include was a recent one, but the .deps from an old build
were still used?

I'd kind-of like the Patchy build to fail as soon as it has *any*
missing include files.  Does anybody know how to make g++ and/or
make(1) behave that way?  Is that what -Werror would do?


Anyway, I've completely removed the build dir and started another
run, just in case it was something weird.

Cheers,
- Graham