Re: Are Goto's in the Linux Kernel necessarily bad?

[Mike Connor <mikec@silug.org>]

Huh?  Well, I didn't think a big answer was wanted, so I used the word 
"usually".  I like generalities.

Speaking of kernels and hardware, I spent four years hacking on embedded 
systems drivers and applications for VxWorks & and a proprietary kernel.  
I've also done things like porting a kernel (something like 3000 C & 68k 
assy. files  worth) from a 68k processor board to run as a process on 
Solaris.  Helped with about 10k lines of DSP assembly too-- dealing with 2 
memory buses was pretty novel (don't ask me which chip).  So I think I'm 
qualified to open my mouth here. 

Mike



On Sat, 15 Feb 2003 mike808@users.sourceforge.net wrote:

> > It depends on why they're being used.  In the kernel, they are being used 
> > for error handling, and pretty much only error handling in complex 
> > situations.
> > 
> > Any other use of a goto usually means bad code or intentionally 
> > obfuscated.
> 
> Not true. They are prefectly suited to state machine execution and
> jump tables (such as interrupt vectors). You really only need to use
> gotos when you're close to the "metal". Which is where the kernel is.
> The closer you are to the assembly and wiring, a "goto" is merely an integral
> part of branching logic. Since in modern computers, instructions are executing
> sequentially, any form of "If this, then execute A else execute B" means that
> someone (A or B) necessarily *cannot* possibly be the next sequential set
> of instructions to be executed. So, if the branch is taken to execute the
> "other" code (non-sequential), the only mechanism a CPU has to fetch that
> set of instructions is necessarily a "goto" instruction.
> 
> But, back to the kernel. It has to run on *lots* of hardware. So, it has 
> constraints being placed on it that do not apply to general user application 
> programming, which is what the people who gave lectures about "Goto considered
> harmful" and such in the early 80's of Computer Science.
> 
> You should read a more recent book, Jonathan... :=)
> 
> Mike808/
> 
> ---------------------------------------------
> http://www.valuenet.net
> 
> 
> 
> -
> To unsubscribe, send email to majordomo@silug.org with
> "unsubscribe silug-discuss" in the body.
> 


-
To unsubscribe, send email to majordomo@silug.org with
"unsubscribe silug-discuss" in the body.