[antlr-interest] Serious Bug when using BitSetgeneration

[Ric Klaren]

Geir Ove Skjaervik wrote:
> First of all, I have stated that I WILL try to provide an example that
> reproduces the error. I am currently at a conference in Malmø Sweeden
> and do not have the time nor opportuinity to do that; It will surely be
> time consuming.

I'll look forward to the example. In this case it will be enough if it's
a .g file that generates code it doesn't need to compile. (if it does
it's nice though)

> Secondly, I sincerely hope you are NOT Serious when you say: QUOTE:
> <<Also you seem to be the only one that ran into this so far>>  Well,
> someone surely must be the first one to report / encounter a bug, but in
> your book that means neglecting it? How do you know that others haven't
> encountered this bug? It took me 2 at least 2 months of trying to
> rewrite code since I thought it was me, when it was ANTLR all the time.

There's a difference between prioritizing it in a way *you* don't like
and neglecting. We can spent our time bughunting much more efficiently
with a clue where to look if we get a snippet that triggers the problem.

> Thrirdly, the workaround is not good performance wise.

But it 'works'. E.g. you can continue developing. Not in an optimal way
but you can.

> What's worst, is that as long as the Bug is there, you really do not
> knows when it bites you: The only way to find out, is to try BOTH with
> and without Bitsets whenever a piece of code does not parse as expected.
> THAT IS NOT a good solution timewise, and ANTLR is supposed to save us
> time, isn't it?

Well develop with bitsets turned off? Or modify your build to create
both versions so you can directly see if the behaviour differs between
the two. Producing a snippet that triggers the bug is probably more time
efficient though.

Anyway shouting at eachother won't solve anything.

Cheers,

Ric