standards/177742: conflict of dd's bs= option with use of conv=sparse

[Matthew Rezny]

The following reply was made to PR standards/177742; it has been noted by GNATS.

From: Matthew Rezny <mrezny at hexaneinc.com>
To: Konstantin Belousov <kostikbel at gmail.com>
Cc: bug-followup at FreeBSD.org
Subject: Re: standards/177742: conflict of dd's bs= option with use of
 conv=sparse
Date: Tue, 23 Apr 2013 13:00:43 +0200

 On Tue, 23 Apr 2013 09:21:23 +0300
 Konstantin Belousov <kostikbel at gmail.com> wrote:
 
 > On Thu, Apr 11, 2013 at 05:44:59AM +0200, Matthew Rezny wrote:
 > > Looking not just at my problem but the overall situation, it seems
 > > the logic in dd_in() does not hold up. The last line, (*cfunc)();,
 > > actually invokes the appropriate conversion(s) and then calls
 > > dd_out(0). With the simple test on line 361, we will never get down
 > > to that call if bs= option is used. Really, to make the code fit
 > > the comment, it should probably be something like
 > > if ((ddflags & ~(C_NOERROR | C_NOTRUNC | C_SYNC)) == C_BS) {
 > >     out.dbcnt = in.dbcnt;
 > >     dd_out(0);
 > >     in.dbcnt = 0;
 > >     continue;
 > > }
 > I do not see why did you changed dd_out(1) to dd_out(0).
 > 
 > Anyway, I am going to commit the following patch, unless I get some
 > objection in two-three days:
 > 
 > commit 37664b1393db30d97aa9eefa70360cd550bf9dd9
 > Author: Konstantin Belousov <kib at freebsd.org>
 > Date:   Tue Apr 23 09:18:31 2013 +0300
 > 
 >     Literally follow POSIX:
 >     If the bs= expr operand is specified and no conversions other
 > than sync, noerror, or notrunc are requested, the data returned from
 > each input block shall be written as a separate output block.
 >     
 >     PR:	standards/177742
 >     Submitted by:	Matthew Rezny <mrezny at hexaneinc.com>
 > 
 > diff --git a/bin/dd/dd.c b/bin/dd/dd.c
 > index 01b66fc..9d0f9b1 100644
 > --- a/bin/dd/dd.c
 > +++ b/bin/dd/dd.c
 > @@ -358,7 +358,7 @@ dd_in(void)
 >  		 * than noerror, notrunc or sync are specified, the
 > block
 >  		 * is output without buffering as it is read.
 >  		 */
 > -		if (ddflags & C_BS) {
 > +		if ((ddflags & ~(C_NOERROR | C_NOTRUNC | C_SYNC)) ==
 > C_BS) { out.dbcnt = in.dbcnt;
 >  			dd_out(1);
 >  			in.dbcnt = 0;
 > 
 
 On review, I think the dd_out(0) was an oversight. I put the more
 precise condition up top but then forgot to put it back to dd_out(1) as
 it had been. The one line patch you plan to commit should be effective.
 
 Thank you for taking the time to read my perhaps overly verbose report
 on the issue.