[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Bison no longer builds with K&R C.
From: |
Paul Eggert |
Subject: |
Re: Bison no longer builds with K&R C. |
Date: |
Wed, 16 Oct 2002 00:17:27 -0700 (PDT) |
> From: Akim Demaille <address@hidden>
> Date: 16 Oct 2002 08:36:05 +0200
>
> | I propose that we officially discontinue support for building Bison with
> | K&R C.
>
> This is fine with me, unless the problem is in Automake: they want to
> keep maintaining ansi2knr, so if the problem was there, we ought to
> know what went wrong.
No, the problems were independent of ansi2knr. Here are the
first three problems with Bison 1.75 that Peter Klein reported
for SunOS 4.1.4 cc:
"./quotearg.c", line 631: no automatic aggregate initialization
"./quotearg.c", line 631: operands of = have incompatible types
"./quotearg.c", line 639: no automatic aggregate initialization
"./quotearg.c", line 639: operands of = have incompatible types
"./xstrndup.c", line 38: syntax error at or near type word "char"
"./abitset.c", line 904: syntax error at or near word "union"
"./abitset.c", line 905: syntax error at or near symbol (
"./abitset.c", line 906: syntax error at or near word "struct"
At that point he gave up, and I don't blame him. It's just not worth
the maintenance effort any more.
> Also, GCC does keep the K&R requirement, although in a limited form
> (IIRC the stage 1 compiler must compile with a K&R compiler). I
> suspect that the bitset lib must remain K&R.
That's OK: we can run it through ansi2knr before giving it to the GCC
folks. This is safe so long as we continue stick to a subset of ANSI
C that will safely become K&R C after being passed through ansi2knr.
> Did you have any news from its author?
Not since his patch of 2002-10-04.