[Date Prev][Date Next][Thread Prev][Thread Next][Author Index][Date Index][Thread Index]
Re: Something to do While Waiting for Compiles
- To: <vlad!mark>
- Subject: Re: Something to do While Waiting for Compiles
- From: Michael McClary <michael>
- Date: Sat, 11 Aug 90 17:18:30 PDT
- Cc: <xtech>
> and hack the translator to produce PROTECTED: instead of PRIVATE:
> if/when/as long as you must? That way you're not polluting the
> ENTIRE universe, just that part translated from smalltalk.
>
> That's where we started, and it seemed worse (although I sheepishly
> can't reconstruct why). Hopefully we will all fix our protection
> violations soon enough that we need not worry about which kludgy work-
> around we need to use till our code is unbroken.
Why not just fix the classes and be done with it? You've got two
separate tools that identify them all. That way we could pull out
the kludge in the next merge.
This isn't like fixing the regression tests, which can require deep
analysis. Replacing the improper accesses with calls to accessor
functions, and adding protected accessor functions when necessary,
is straightforward, and when the compiler shuts up it's nearly
certain the result is as correct as what it replaced. (And if your
regression test was working properly before you started, you know
your code is converted correctly if it passes when you're done.)
Leaving them around until it's "convenient" makes things three times
as bad, because you have:
- The old style,
- The new style, and
- The kludge switch
to worry about, as well as trying to remember that you need to make
the change, extra vulnerability to bugs (because each style attracts
a slightly different set of global bugs), code that sets a bad example
for new programmers, and I could go on.
curiously
michael