cvs commit: src/bin/cat

[Jason Smethers]

Matthew Dillon wrote:
:
:Liam J. Foy wrote:
:> liamfoy     2004/11/04 13:31:37 PST
:> 
:> DragonFly src repository
:> 
:>   Modified files:
:>     bin/cat              cat.c 
:>   Log:
:>   - Remove unnecessary selection statement to see if buf == NULL.
:>   - Don't use fileno to get stdout, just use STDOUT_FILENO
:>   - Fix malloc err message
:>   
:>   OK'ed by: Joerg and Jeroen
:>   
:>   Revision  Changes    Path
:>   1.12      +8 -10     src/bin/cat/cat.c
:> 
:> 
:> http://www.dragonflybsd.org/cvsweb/src/bin/cat/cat.c.diff?r1=1.11&r2=1.12&f=u
:
:
:Removing the check for (buf == NULL):
:1. Introduces a memory leak -- equal to (bsize * (n - 1)) where n is the 
:number of times raw_cat is called.
:2. Adds unecessary work -- stdout never changes
:
:
:Also, the constant 1024 in the buffer size check should be replaced by 
:the posix constant BUFSIZ.
:
:- Jason
    
    Whoops, quite right.  I missed the fact that 'buf' was a static when
    I review Liam's commit.

    The original code isn't so good either.  While I don't expect we
    will be cat'ing multiple block devices I think it would probably still
    be a good idea to fstat/resize the buffer on a per-descriptor basis.
    The expensive is minimal.
    I'll just go ahead and fix it.  Don't worry about it, Liam.

					-Matt
					Matthew Dillon 
					<dillon at xxxxxxxxxxxxx>

But there is only one descriptor to stat -- stdout. Cat always writes to 
 stdout, and as per POSIX stdout is never closed and reopened. 
Shouldn't fstat therefore always return the same value? Or, do you mean 
you want to optimize the code for the input buffer instead of the output 
buffer?

- Jason