[V3dot10] Image before 0002788: Bug: Use of == for arithmetic
equality problem
Ralph Johnson
johnson at cs.uiuc.edu
Wed Oct 17 13:06:30 UTC 2007
For people who don't know what this is about, look at
http://bugs.squeak.org/view.php?id=2788
Briefly, Dan Ingalls suggested that == should generally be replaced by
=, and so someone did this, probably automatically. The changes went
through a few iterations, but finally they got installed in 3.10.
This caused a few failing tests, some of which got fixed, and produced
some bugs that didn't show up in the tests.
> My proposal is do the rest of updates without this, name image 3.10.1 and
> warn people.
Edgar, are you saying that you will take this change out of 3.10 and,
once 3.10 is done, put it in as 3.10.1?
I prefer Jerome's suggestion that the changes should be put in
piecemeal, and with tests. Replacing == with = is only right 98% of
the time, and putting bugs in the code and then taking them out is a
lot more work than thinking about the changes you are going to make
before you make them. Now, if the transformation were 99.9% accurate
then it might be better just to go ahead, but there are too many cases
where == is right and = is wrong.
-Ralph
More information about the V3dot10
mailing list