Re: [PATCH] enable buildable targets by default

[Pavel Roskin]

On Fri, 2009-07-17 at 11:21 +0200, Vladimir 'phcoder' Serbinenko wrote:
> On Fri, Jul 17, 2009 at 2:37 AM, Pavel Roskin<address@hidden> wrote:
> > Quoting Vladimir 'phcoder' Serbinenko <address@hidden>:
> >
> >> comitted
> >
> > I wish I had time to review it :-(
> >
> > I have fixed the easy stuff (spelling, wrong variables, use of "$no",
> Thanks
> > use of "-c" in CFLAGS).
> This was done on purpose so that we don't need gcc-multilib for tests

AC_COMPILE_IFELSE already means that the program is compiled but not
linked.  I tried removing glibc-devel.i586, and everything is still OK.

> > The remaining problem is that efiemu only compiles on
> > i386-pc, but configure not only checks for the necessary flags on other
> > platforms, but also announces that it will be built.
> Actually the check is necessary only for efiemu64. Perhaps we should
> always build efiemu32 and make grub fallback to efiemu32 even on
> 64-bit platform if efiemu64 is unavailable. Then enable-efiemu would
> be renamed to enable-efiemu64

That's a separate issue.  And then there is a problem that efiemu is
undocumented, so users cannot decide whether they need it or not.

> > Perhaps we can compile efiemu on i386 and x86_64, but certainly not on
> > powerpc and sparc.
> >
> > Once everything is compiled by default, some of the configure options become
> > less meaningful, so we can consider removing them.  It's not so important to
> > disable some feature as it is to enable it.
> IMO if user does some not so useful things it's his problems. It's not
> a reason to limit flexibility

I think we should provide meaningful options.  For example, an option to
disable compiling debugging tools, or an option to add extra sanity
checks.  An option to disable efiemu64 is not meaningful for someone who
doesn't know how to use efiemu64.

-- 
Regards,
Pavel Roskin