Re: untangling the remaining regressions

[David Kastrup]

Mike Solomon <address@hidden> writes:

> The short term solution may be to create an internal property called
> "pure-relevant", set it to #t as default for all grobs in the
> Grob::Grob constructor and set to false only when necessary in
> define-grobs.cc.  Then, write a big fat comment saying that the goal
> eventually is to remove this test and make sure to say that the
> property is _strictly_ internal.
>
> This is more elegant than the Scheme functions in the old
> implementation but less elegant than the currently (albeit broken)
> solution of removing the pure-relevant distinction.

It's window dressing on bugs if the design does actually work properly,
and if the design _doesn't_ actually work properly, it lands us with a
non-working design and related code that is even harder to revert or
fix.  Papering over and nailing shut known fundamental problems would
create an even larger maintenance nightmare than what we have now.  We
have enough "oh, we'll fix that properly some time I'm sure" permanently
rotting code as it stands.

> I can do this tonight - does this sound like a reasonable
> middle-ground solution?

No.  Two half-solutions don't make a full one.

-- 
David Kastrup