Re: make bootstrap in separate build directory will overwrite loaddefs.el

[David Kastrup]

"Eli Zaretskii" <address@hidden> writes:

> > From: address@hidden (David Kastrup)
> > Date: 19 Feb 2002 08:48:01 +0100
> > 
> > The question would be whether an initial copy of loaddefs.el is
> > necessary for make bootstrap in the first place.  If it isn't, the
> > easy way out would be to remove it from CVS altogether.
> 
> It's a long-standing issue that didn't get resolved.
> 
> I don't think we want to force people to "make bootstrap" every time
> they resync with the CVS.  Bootstrap is only required the first time
> you check-out the whole tree.  Thereafter, "make recompile" in the
> lisp directory should be enough in the absolute majority of cases.

*Now* you tell me.

> > Since on my machine make bootstrap takes hours, I have not tested
> > this right now.
> 
> The time it takes to bootstrap is one reason not to require it.

Here is my take on that:
make bootstrap
from a separate build tree *will* cause changes to the source tree.
In particular, all the elc files will be compiled in the source
tree.  Those elc files are *not* in the CVS, and they shouldn't.  If
you delete your separate build tree, you don't need to redo
make bootstrap
Now the question is whether
make bootstrap
can get off the ground without requiring an initial loaddefs.el.  If
it does, then its absence in the CVS tree would not imply that it
would be removed from the source tree once you have run "make
bootstrap".

Since it suffers the same consequences of "bit rot" as do the .elc
files, it would be sensible to regenerate it in the same manner,
namely in the course of running
"make recompile"
in the lisp directory.

This is how it appears to me at first glance.  Not being intimate
with the details of the build process, I might well be mistaken.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum
Email: address@hidden