Re: bin/41072: dd ignores bs= when if= comes from stdin

["Perry E. Metzger" <perry%piermont.com@localhost>]

The following reply was made to PR bin/41072; it has been noted by GNATS.

From: "Perry E. Metzger" <perry%piermont.com@localhost>
To: SODA Noriyuki <soda%yuruyuru.net@localhost>
Cc: gnats-bugs%NetBSD.org@localhost,  gnats-admin%netbsd.org@localhost,  
netbsd-bugs%netbsd.org@localhost
Subject: Re: bin/41072: dd ignores bs= when if= comes from stdin
Date: Wed, 25 Mar 2009 09:10:33 -0400

 SODA Noriyuki <soda%yuruyuru.net@localhost> writes:
 >> try
 >>     something.... | dd of=/dev/somepartition bs=10m progress=1
 >> and note that you get way too many dots printed. 
 >
 > That's not a bug, but documented behavior:
 >
 >      bs=n     Set both input and output block size, superseding
 >               the ibs and obs operands.  If no conversion values
 >               other than noerror, notrunc or sync are specified,
 >               then each input block is copied to the output as
 >               a single block without any aggregation of short
 >               blocks.
 >
 > Please note the following sentence, especially:
 >      If no conversion values other than noerror, notrunc or sync
 >      are specified, then each input block is copied to the output
 >      as a single block without any aggregation of short blocks.
 > i.e. In this case, dd preserve the return size of the read(2)
 > system calls at the corresponding write(2) operation.
 >
 > This behavior is needed to copy a variable-blocksize tape with
 > preserving each size of tape blocks.
 
 Okay, the phrase "without any aggregation of short blocks" is somewhat
 confusing to me. I would suggest that it be explained a bit
 better. How about adding, at the end of the paragraph:
 
     This behavior is needed to copy a variable-blocksize tape with
     preserving each size of tape blocks. However, it may result in
     unexpected behavior when short reads are likely, such as when
     reading from a pipe on stdin. In such cases, set obs= alone.
 
 Actually, though, is there a conv= setting that says "always
 aggregate"? (I don't seem to see one.)
 
 Perry