Re: emacs.pdmp not always rebuilt

[Gregory Heytings]

> > > What do you mean by "debugged or updated", and how would that affect 
> > > what we do at loadup time, but not what we do at any other time, 
> > > including in interactive sessions?
> >
> > The problem that Lars, you and possibly others experienced (make 
> > continuing when it should have stopped) was due to the fact that "call1 
> > (Qsubstitute_command_keys, Fget (errname, Qerror_message));" in 
> > print_error_message() failed when it was called while loading the first 
> > files in loadup.  Are you sure that this call to call1 cannot possibly 
> > fail when fboundp is t?
>
> If fboundp returns non-nil, it means the symbol's function cell is not 
> void, i.e. the symbol can be used as a function.  Right?  So the call 
> can only fail if the function itself has a bug, right?  And in the 
> latter case, how is that different from any other Lisp function called 
> during loadup?

It's different, precisely because it is not a function among other 
functions, it's the function that signals an error.  And as we've just 
seen, if the function that signals an error itself fails, unexpected 
behavior can happen, especially when this happens during bootstrap.

I would have taken extra care to avoid that possibility, especially given 
that the importance of substitute-command-keys on error messages during 
bootstrap is close to zero.  Perhaps I did not look close enough, but I 
could not find a single error message where substitute-command-keys would 
have had an effect, and even if there are a few, seeing "Command 
\[foobar\] blah blah" instead of "Command M-x foobar blah blah" or 
"Command C-x % blah blah" does not make a big difference.