Re: CVS commit: src/usr.bin/mklocale

["T.SHIOZAKI" <tshiozak%netbsd.org@localhost>]

From: "Valeriy E. Ushakov" <uwe%ptc.spbu.ru@localhost>
Subject: Re: CVS commit: src/usr.bin/mklocale
Date: Mon, 10 Mar 2003 19:05:20 +0300
Message-ID: <20030310160520.GA6668%snark.ptc.spbu.ru@localhost>

> On Mon, Mar 10, 2003 at 11:37:05 +0200, Takuya SHIOZAKI wrote:
> 
> > Module Name:        src
> > Committed By:       tshiozak
> > Date:               Mon Mar 10 09:37:04 UTC 2003
> > 
> > Modified Files:
> >     src/usr.bin/mklocale: yacc.y
> > 
> > Log Message:
> > revoke using PRIu32 macro and add fallback for UINT32_C macro,
> > because these do not exist on rather older systems.
> 
> UINT32_C is still used by lex.l (via _CTYPE_* macros).  Building on
> FreeBSD 4.x fails with:
> 
> lex.lo: In function `yylex':
> lex.lo(.text+0x96f): undefined reference to `UINT32_C'
> *** Error code 1

I see.


> There's also a consistency problem - when mklocale is compiled as a
> host tool it pulls in host's <ctype.h> but our <runetype.h>.  On
> FreeBSD 4.x it results in a clash between _CTYPE_* macros they define
> in ctype.h and _CTYPE_* macros we define in runetype.h.
> 
> In file included from /nb/src/tools/mklocale/../../usr.bin/mklocale/yacc.y:66:
> /usr/include/ctype.h:54: warning: `_CTYPE_A' redefined
> /nb/src/lib/libc/locale/runetype.h:66: warning: this is the location of the 
> previous definition
> 
> I think that while other host tools should be using host's ctype.h,
> mklocale should make sure to use the in-tree ctype.h.

No, ctype.h is used only for debug dump (`isprint') in mklocale.
The real problem was what _CTYPE_x (defined in runetype.h) macros
just conflict to host's ctype.h under FreeBSD.

I have fixed them.  Thanks for your report.


--
Takuya SHIOZAKI