Date: Sun, 10 Jan 1999 22:20:10 +1100 (EST) From: Darren Reed <darrenr@reed.wattle.id.au> To: jkh@zippy.cdrom.com (Jordan K. Hubbard) Cc: phk@critter.freebsd.dk, des@flood.ping.uio.no, committers@FreeBSD.ORG Subject: Re: sysctl descriptions Message-ID: <199901101120.WAA15560@avalon.reed.wattle.id.au> In-Reply-To: <50085.915954825@zippy.cdrom.com> from "Jordan K. Hubbard" at Jan 9, 99 11:53:45 pm
next in thread | previous in thread | raw e-mail | index | archive | help
In some email I received from Jordan K. Hubbard, sie wrote:
>
> Actually, there is a legitimate syncronization issue to be dealt with
> here. If I have a sysctl implemented in one area of the code with a
> doc string that describes it, I'm going to be more inclined to stick
> it in the macro definition or something itself so that I have doc
> string and associated sysctl code together. If I have to go remember
> to edit a file somewhere as well, I'm going to forget just as
> developers have been forgetting such extraneous details for years and
> the problem is acute enough that entire programming paradigms
> (tangle/weave) have been evolved in an effort to deal with the
> problem. I'd sooner have a larger kernel if it meant that my
> doc strings had a better chance of being actually *correct*. :)
If you have such problems then the tools you are using to generate
code for your object are not suitable to the task at hand. There
is a plethora of extremely useful text manipulation tools ("perl"
for one) which could easily automate any such documentation task if
it were merely based on the contents of a .c file.
As I'm not familiar with the KLD issue, I cannot say what would and
wouldn't be a good way to resolve the problem. If it were applied
to KLM's, then my response would be to have the "modload process"
involve more than just "modload foo.o". It might be that a wrapping
program or script is needed (modload already supports something similar
with the -e command line option). If the interface is truely lacking
in a way to support this feature then perhaps the interface needs to
change.
I'm not trying to say that the descriptions shouldn't be there, but at
least try to be smart about considering alternatives and the work involved.
We should be at least trying to move out of the stone age of programming
now (it is, however, hard when that is what you're used to).
Personally, I view the sysctl description code as done by DES as the
"easy solution". It's more than likely what most people would do if
they had the same problem to solve. I don't know whether I would go
on to say that this is the "correct" solution for "this problem" (but
then people may not be at all interested in the "correct" solution).
Darren
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199901101120.WAA15560>