RE: patch 20031120

> > 
> > perhaps this can be a better solution... add a field to TDSICONV to 
> > store old errors and caller should say if this is the first
> > and/or the last chunk (in no-chunk conversions it's always 
> > the first and the last chunk). 
> 
> This might turn out to be the best answer.  In effect, 
> TDSICONV would hold an "error mask" of messages to suppress.  
> tds_iconv() would add to the mask as messages accumulated.  
> The problem would be to remember to clear the mask when done.  
> 
> tds_iconv.h:
>       struct _suppress {
>               unsigned int ei2big:1;
>               unsigned int eilseq:1;
>               unsigned int einval:1;
>       } suppress;
>       /* ... */
> 
> /* caller: */
>       memset(&iconv_info->suppress, 0, sizeof(iconv_info->suppress));
>       iconv_info->suppress.eilseq = *wire_size > 0;
>       irreverisible = tds_iconv(...);
> 
> > So you can also reset iconv state (iconv(cd, NULL,
> > NULL, NULL, NULL) )... Does this sound reasonable ?
> 
> I don't know if this is important.  It might be just as easy to call
> iconv(3) instead.  
> 

I'd prefer a tds_iconv_reset... 
I don't understand "iconv_info->suppress.eilseq = *wire_size > 0;"
reason...
Perhaps the best solution should be a tds_iconv_done called after every
conversion. So to reset state for next conversion, issue
EINVAL/EILSEQ/whatever warnings/errors and prepare everything else for
conversion. 
So tds_iconv is able to detect first/not first conversion. 
If I have time I'll try to add a tds_iconv_done and report a patch
today...

freddy77