Date: Mon, 24 Nov 1997 02:48:56 +0100 (MET)
From: Hans-Bernhard Broeker <broeker AT physik DOT rwth-aachen DOT de>
To: DJ Delorie <dj AT delorie DOT com>
cc: eliz AT is DOT elta DOT co DOT il, djgpp-workers AT delorie DOT com
Subject: Re: alpha-971114: Makefiles revisited
In-Reply-To: <199711240132.UAA16504@delorie.com>
Message-ID: <Pine.LNX.3.93.971124023511.23046C-100000@acp3b9>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Precedence: bulk

On Sun, 23 Nov 1997, DJ Delorie wrote:

[This will be the last answer from me for now. It's 02:30h in the morning
here, gotta get some sleep :-)]

> Well, yes.  I meant everything that *should* be built with the new
> libc should be built with the as-yet-uninstalled one.

The current off-$(DJDIR) solution already grants this, I think. At least,
the "ident | grep '$Id' | sort | uniq -c" output says it does.

> >   * bin2h is in the sources, but doesn't seem to be in djdev, why?

> I don't know why.  Perhaps it isn't used by anything, so I thought it
> was built just for doing the build itself.

I don't think it's used in the build process (that's what 'stub2inc' is
used for, IIRC). But it's a handy tool for the users, so I think it should
go into djdev*.zip. 

> >   * A .txh bug noticed in passing: 'remque.txh' has 'putenv' as the
> >     function name in the syntax section  (fixed)
> 
> Fixed where?

In the version I have sitting on my machine at home. The patch will be
in the next set of diffs I'll send, promised.

> >   * The current lsr dist contains some non-source files, and at least 
> >     one temporary file (djasm-n.c, built from djasm.y). 
> 
> Please identify them and I'll add the exclusion to my packaging
> routines.

Even easier: just 'make clean' after the build, and they'll be deleted,
so the packaging routine will not be able to package them any more :-)

(But yes, you will get that list, ASAP)

> >     MY_LIBGCC_A = $(shell $(CROSS_GCC) -print-libgcc-file-name)
[doesn't work ...]
> >     Any ideas how to get around that?
> You'd have to change gcc to use forward slashes exclusively.

That's about as far as inspecting gcc.c took me, yes. But doing that
(it's a compile time option only) would render
it less usable for many other, more important tasks, of course. And I
cannot count on everyone having a gcc.exe changed this way, anyway.

> >   * Currently, 'libc2.tex' is rebuilt every time 'make' is run. Not a
> >     big issue, but a fixable one.
> 
> Hard to get right, since make can't detect that a target no longer
> depends on a file that's been deleted.  That's why libc's dependencies
> are so complicated - it needs to detect the case where an object was
> removed, even when all remaining objects are older than the library.

I had something working in the July version of the makefiles. It was
a copy of the .o->.a file machinery, essentially. The same could be
done again, of course, in makemake.c.

BTW: congratulations to the idea with misc.c. Seeing it immediately made
me think "Yes, that's the method! What a pity you didn't imagine that
one." :-) 

Bye, for today.

Hans-Bernhard Broeker (broeker AT physik DOT rwth-aachen DOT de)
Even if all the snow were burnt, ashes would remain.