Subject: Re: libcurses ABI change
To: None <source-changes@netbsd.org>
From: Christos Zoulas <christos@zoulas.com>
List: source-changes
Date: 06/26/2002 18:40:21
In article <20020626182835.05B184B27@coconut.itojun.org>,
 <itojun@iijlab.net> wrote:
>
>>Module Name:	basesrc
>>Committed By:	christos
>>Date:		Wed Jun 26 18:14:09 UTC 2002
>>
>>Modified Files:
>>	basesrc/lib/libcurses: color.c cr_put.c ctrace.c cur_hash.c
>>	    curses_private.h refresh.c scanw.c setterm.c tscroll.c
>>
>>Log Message:
>>PR/17352: David Laight: Curses core-dumps on windows > 1024 wide.
>
>	this commit includes an ABI change: __hash(x, y) to
>	__hash_more(x, y, z).  the previous symbol have gone.
>	don't we need a shlib major bump?   or, should we provide __hash(x, y)
>	as function (not as a macro) and do a shlib minor bump?

I don't think so. __hash(x, y) is a private function and no portable code should
be using it [it is not advertised to be there, so a program needs to know it
exists in order to use it]. I would be more worried about the change in the
structure size 1024 -> 128 in the terminal name buffer, but again all this is
private curses space and does not affect the binaries.

christos