[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: bison 3.4.2 fails to handle large model
From: |
Akim Demaille |
Subject: |
Re: bison 3.4.2 fails to handle large model |
Date: |
Sat, 28 Sep 2019 17:09:49 +0200 |
Hi Tom,
> Le 27 sept. 2019 à 23:13, Tom Kramer via Bug reports for Bison, the GNU
> parser generator <address@hidden> a écrit :
>
> My request is: please enable bison to handle much larger state numbers. I am
> guessing that all that is required is to change some numerical type from 16
> bits to 32 bits, which should be very easy. But that's a guess and I realize
> "should be very easy" is often a dream.
No. I agree with you, it should be easy :) However, I don't want to just bump
to 32 bits, for some of our user, space matters, so I'll have to make sure we
generate parsers that use "the right size" for states. As a matter of fact,
that's already what we do for the constant tables: they use the "best" int
type. Stupidly enough, we did not do that for the variables storing state
numbers... Even more amazing is that you seem to be the first one to notice.
> I will be happy to send you more information, but I think a quick glance at
> the dump data shown below is probably all that is required to understand the
> problem. Since the problem occurs when yydebug is not set, I do not think the
> problem is limited to printing the dump data.
Sure. Could you please try this tarball, quick and dirty, just to make sure
that it does address your problem?
https://www.lrde.epita.fr/~akim/private/bison/bison-3.4.2.117-e8e3f.tar.gz
https://www.lrde.epita.fr/~akim/private/bison/bison-3.4.2.117-e8e3f.tar.xz
Are you in a hurry? In the sense of: do you need a release really soon, or can
it wait for Bison 3.5 (which should be before 2020) and you can work with the
git version meanwhile?
Cheers!