Lexical State Swiching, SCAN and odd error messages: Page 2

Lexical State Swiching, SCAN and odd error messages

revusky

adMartem
Oh, by the way, do make sure to test against the absolutely latest code. In the last couple of days, I really did some review of the lookahead generation logic and nailed certain bugs. Now, there was a pretty flagrant bug in the lexical state handling, where it could generate a faulty lookahead when the lookahead involved a lexical state switch. That might have been something you were hitting. So, there's that. But, just yesterday, like less than 24 hours ago, I came across another bug in the lookahead generation logic and fixed that.

Actually, I started writing here a description of the bug (now fixed) but then I figured it would be better to write a top-level post about it...

adMartem

revusky
I have tested against the latest version, and it seems to be fine. I assume I can now use ENSURE to denote what were previously ASSERTions annotated with # and use ENSURE ASSERT together if necessary (but I don't think I ever need to do that). This is a great change, BTW. I think it is much clearer than the effective (mis)use of ASSERT before. Thanks.

vsajip

Can you please add tests for the bugs you've found, so that we can test for regressions in the future?

revusky

adMartem
Aha! You are definitely like one of the students sitting in the front row, eager to raise his hand and make a point, not sitting in the back row napping!

So, yes, you saw the new feature before I announced it! I just put up an announcement here: https://parsers.org/announcements/revisiting-assertions-introducing-the-ensure-keyword/

« Previous Page