Subject: Re: Tinderbox tyrant (wv issues)
From: Thomas Fletcher (thomasf@qnx.com)
Date: Sun Oct 15 2000 - 15:25:02 CDT
Thomas opens mouth and inserts foot ...
On Sun, 15 Oct 2000, Mike Nordell wrote:
> Thomas Fletcher wrote:
> > On Fri, 13 Oct 2000, Mike Nordell wrote:
> ["fd", tmpnam and portability snipped]
> >
> > Well whatever ... read/write/open are standard C file operations
>
> "standard C file operations"? Really? ;->
>
> Last time I checked open/read/write were not ANSI C, and I would be really,
> *really*, surprised if ANSI C 9x made it part of the library. That is the
> *only* "standard" I know about when discussing C. And even if they did vote
> these in it wouldn't affect the incorrectness of the code discussed. C++
> does not derive from C9x.
Absolutely correct, I was wrong open/read/write are not part
of the ANSI standard, but are POSIXish standard. My mistake,
I was absolutely unabashedly wrong!
> > (as much as fread/fwrite/fopen) with less overhead than the
> > stdio FILE *.
>
> Overhead to create a FILE* from a _Unix-only_ "fd" compared to time time it
> actually takes to do the (probable) disk seek to locate the file on disk on
> most current platforms? Any recent profiling figures? ;->
Not necessarily overhead to create, but overhead to use. No
doubt this will turn into a "but you can configure the buffer
to do be appropriate for you" type of discussion. They both
have their uses ... my comment was an off the cuff dig at
stdio.
Thomas
This archive was generated by hypermail 2b25 : Sun Oct 15 2000 - 15:24:14 CDT