[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: multiline substitutions
From: |
Stepan Kasal |
Subject: |
Re: multiline substitutions |
Date: |
Thu, 20 Oct 2005 10:36:50 +0200 |
User-agent: |
Mutt/1.4.1i |
[ removed automake-patches ]
Hello,
On Thu, Oct 20, 2005 at 09:31:30AM +0200, Alexandre Duret-Lutz wrote:
> We already have AC_SUBST_FILE. Shouldn't we also have
> AC_SUBST_MULTILINE or some such? This way we can tell the
> difference from --trace,
sure, this would be possible. But I think it's not needed.
(We should decide this before Autoconf 2.60 is released.)
> and use something like @\n@ only for
> the variables that need it, without unnecessary bloat.
I think four bytes per AC_SUBSTed variable in each Makefile.in is not
any "bloat".
> If VAR contains newlines that matters to me (like in a Makefile
> rule), then there is no way I can use $(VAR) in Makefile and
> have these newlines, so I might as well use @VAR@ directly and
> in this case I don't need the VAR = @VAR@ definition.
Yes.
> Changing the definition of LIBOBJS in a way that requires @\n@
> or in a way that doesn't use AC_SUBST will break the
> compatibility with previous Automake versions.
... unless a special compatibility hack is added.
> Maybe the safest
> way would be to add the backslash-newlines before the AC_SUBST.
Yes, this is a possible solution of the problem of too long LIBOBJS.
But I still think that my proposal is more elegant: no matter what
are the values of AC_SUBSTed variables, the generated Makefiles will
be correct.
I hope I'll post an actual implementation, before you get to this
long thread, so that you can have better image.
Have a nice day,
Stepan Kasal