Re: old bootstrap error emerges again

[Eli Zaretskii <eliz@gnu.org>]

> From: Dieter Deyke <deyke@comcast.net>
> Cc: emacs-pretest-bug@gnu.org, emacs-devel@gnu.org
> Date: Sat, 18 Mar 2006 12:55:35 -0700
> 
> I'm not sure on how to debug this. Let me start to point out that
> there were 3 errors about missing DOC

Yes, I've seen those as well in your transcript.  Needless to say, in
my builds none of them are present.

> but the first 2 were not fatal:

No, they are all fatal.  It's just that the shell's for loop invoked
by Make doesn't stop until it finishes (you will see that
url/vc-dav.el is the last Lisp file that is compiled), so Make only
sees the non-zero exit status when the list of files to compile is
exhausted and the subsidiary shell exits.

> Second, after that build stopped, a find | grep DOC came up empty, so
> if anything would expect to find DOC-X, there was none to be found.

In my builds, I don't have DOC-X either, at this stage of the build.
But the problems still don't happen for me.  I think Emacs shouldn't
look for the doc string file at all at this place.  The DOC vs DOC-X
thing was just another strange thing, so I mentioned it.

> So, could you please guide me in what I have to do to debug this?

Well, one way to try to debug this is to say "make bootstrap" again
(without the workaround), let it fail, then run Emacs under GDB and
let it byte-compile the offending file vc-dav.el with precisely the
same command-line arguments as when Emacs is run from Make.  But
before you run Emacs, after starting GDB, set a breakpoint in doc.c,
in the function get_doc_string, on the line which says:

	error ("Cannot open doc string file \"%s\"", name);

(this is the line that displays the error message and aborts the
compilation).  When this breakpoint breaks, please post here the
backtrace.  Maybe that would give us some ideas.

TIA