Ingo Schwarze wrote:
> Oh well, i'm finally catching up...
> Replying to myself:
> 
> Ingo Schwarze wrote on Sat, Jul 31, 2010 at 07:57:17PM +0200:
>> kristaps@mdocml.bsd.lv wrote on Tue, Jul 27, 2010 at 04:38:04AM -0400:
> 
>>> Log Message:
>>> -----------
>>> Fix how `Bd -unfilled' and `Bd -literal' break lines.  This unbreaks
>>> displays to work as old groff shows them; however, new groff still does
>>> some fancy shit.
> 
>> What exactly is this supposed to fix?
> 
> I see now, i missed one of kristaps@' messages.
> It fixes awk(1) and boot_config(8).
> But it causes regressions with respect to blank lines.
>  
>>> Modified Files:
>>> --------------
>>>     mdocml:
>>>         mdoc_term.c
>>>         mdoc_html.c
>> I don't care that much about the HTML part, so i'm snipping that,
>> but i disagree with the following chunk:
>>
>> [...]
>>> Index: mdoc_term.c
>>> ===================================================================
>>> RCS file: /usr/vhosts/mdocml.bsd.lv/cvs/mdocml/mdoc_term.c,v
>>> retrieving revision 1.178
>>> retrieving revision 1.179
>>> diff -Lmdoc_term.c -Lmdoc_term.c -u -p -r1.178 -r1.179
>>> --- mdoc_term.c
>>> +++ mdoc_term.c
>> [...]
>>> @@ -1654,12 +1653,9 @@ termp_bd_pre(DECL_ARGS)
>>>  	p->rmargin = p->maxrmargin = TERM_MAXMARGIN;
>>>  
>>>  	for (nn = n->child; nn; nn = nn->next) {
>>> -		p->flags |= TERMP_NOSPACE;
>>> +		if (nn->prev && nn->prev->line < nn->line)
>>> +			term_newln(p);
>>>  		print_mdoc_node(p, pair, m, nn);
>>> -		if (NULL == nn->prev ||
>>> -		    nn->prev->line < nn->line ||
>>> -		    NULL == nn->next)
>>> -			term_flushln(p);
>>>  	}
>>>  
>>>  	p->tabwidth = tabwidth;
> 
>> For me, this breaks .Bd -literal vertical and horizontal spacing
>> in multiple ways, see
>>
>>   mdoc/Bd/blank.in
>>   char/space/tab.in
>>
>> in the OpenBSD regression suite.
>>
>> To be precise, i see at least the following problems:
>>
>>  * Removing the TERMP_NOSPACE adds an additional space character
>>    at the beginning of display lines which we don't want.
>>  * Changing term_flushln to term_newln has the effect of
>>    ignoring blank lines, but blank lines must be preserved
>>    in -literal mode.
>>  * Removing the NULL == nn->next clause kills trailing
>>    blank lines in -literal displays.  Even these must be
>>    preserved.
>>
>> So, for now, i'm not merging any of the changes of this chunk
>> to OpenBSD.
> 
> Here is an alternative patch which fixes the issues, too,
> but does not cause regressions.  Apply to OpenBSD, or revert
> the above before applying to bsd.lv:
> 
> 
> Index: mdoc_term.c
> ===================================================================
> RCS file: /cvs/src/usr.bin/mandoc/mdoc_term.c,v
> retrieving revision 1.99
> diff -u -p -r1.99 mdoc_term.c
> --- mdoc_term.c	31 Jul 2010 21:43:07 -0000	1.99
> +++ mdoc_term.c	4 Aug 2010 00:19:28 -0000
> @@ -1649,12 +1649,11 @@ termp_bd_pre(DECL_ARGS)
>  	p->rmargin = p->maxrmargin = TERM_MAXMARGIN;
>  
>  	for (nn = n->child; nn; nn = nn->next) {
> -		p->flags |= TERMP_NOSPACE;
>  		print_mdoc_node(p, pair, m, nn);
> -		if (NULL == nn->prev ||
> -		    nn->prev->line < nn->line ||
> -		    NULL == nn->next)
> -			term_flushln(p);
> +		if (nn->next && nn->next->line == nn->line)
> +			continue;
> +		term_flushln(p);
> +		p->flags |= TERMP_NOSPACE;
>  	}
>  
>  	p->tabwidth = tabwidth;

Ingo et al.,

First note that I'm back by a machine, so can answer questions and so on.

This patch seems to work by my small tests, so ok kristaps.  I've 
enclosed a patch for mdoc_html.c, which you can throw in to the commit 
for completion's sake.

Thanks,

Kristaps