Re: bugs in CP1258 converter (2)

[Ulrich Drepper <drepper at redhat dot com>]

Bruno Haible <haible@ilog.fr> writes:

> 2001-05-14  Bruno Haible  <haible@clisp.cons.org>
> 
> 	* iconv/gconv.c (__gconv): For flush without output, pass do_flush = 2.
> 	* iconv/skeleton.c: Distinguish do_flush = 1 and do_flush = 2. In the
> 	first case, set outbuf, outstart, outend, and call PREPARE_LOOP before
> 	EMIT_SHIFT_TO_INIT; then pass the output produced by this step down to
> 	the next step. In the second case, clear the state without calling
> 	EMIT_SHIFT_TO_INIT.
> 	* iconvdata/ibm930.c (EMIT_SHIFT_TO_INIT): Use outbuf instead of
> 	data->__outbuf, and outend instead of data->__outbufend.
> 	* iconvdata/ibm933.c (EMIT_SHIFT_TO_INIT): Likewise.
> 	* iconvdata/ibm935.c (EMIT_SHIFT_TO_INIT): Likewise.
> 	* iconvdata/ibm937.c (EMIT_SHIFT_TO_INIT): Likewise.
> 	* iconvdata/ibm939.c (EMIT_SHIFT_TO_INIT): Likewise.
> 	* iso-2022-cn.c (EMIT_SHIFT_TO_INIT): Likewise.
> 	* iso-2022-cn-ext.c (EMIT_SHIFT_TO_INIT): Likewise.
> 	* iso-2022-jp.c (EMIT_SHIFT_TO_INIT): Likewise.
> 	* iso-2022-kr.c (EMIT_SHIFT_TO_INIT): Likewise.
> 	* utf-7.c (EMIT_SHIFT_TO_INIT): Likewise.

This change introduces far too much unconditionally used code.  For
instance, the do_flush parameter is almost never 2 but still the test
is there in every use of the code.

For has to happen is to introduce (yet) another flag/macro which
announces the necessity of this code.  The modules which need the code
will have to define it.

And once again: do not send context diffs.  I can't read this unusable
format.

-- 
---------------.                          ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Red Hat          `--' drepper at redhat.com   `------------------------