front panel display for a modern PC

Vincent Slyngstad v.slyngstad at verizon.net
Sun Mar 2 10:06:37 CST 2008


From: "Chuck Guzis" <cclist at sydex.com>
> On 1 Mar 2008 at 21:44, Vincent Slyngstad wrote:
>> I think of blinkenlights as way ahead of a ROM with a debugger in it, 
>> and not for just the aesthetic reasons folks are talking about.  It's 
>> a little like Tony's fondness for "things he can understand".  
> 
> Well, if I were debugging hardware, I'll take a logic analyzer and a 
> scope, thank you.   I'm not sure how broad the range of hardware 
> faults is that a bunch of blinking LEDs will indicate.  I'm not above 
> using a bus card with indicators and switches on it--but the card 
> goes as soon as the hardware gets debugged.

I'm not talking about debugging hardware.  I'm not really very good at 
that, so wouldn't presume to tell you which tools to use :-).  

We always assumed our machines were in good repair, and that the 
hardware was essentially perfect.  (Our PDP-8's never contradicted 
us, either :-).)

What I am saying is that, when debugging software, it is useful to 
be really sure about what is happening, an the front panels gave 
that reassurance, in a way that debuggers and emulators just don't.
Since debuggers and emulators crash or malfunction, they don't 
mimic the underlying hardware perfectly, whereas the front panels 
gave visibility into the real thing.

> I earned my stripes in the big-iron days.  Hands-on-dedicated time on 
> systems was hard to come by.  Come in at 2AM or 3AM, in hopes of 
> getting some work done, hang around for the 9AM status report, go 
> home, eat supper and go to bed.  A worse case happens when the 
> closest available machine is half across the country--and you have to 
> catch a flight to check out your system changes.  It made me hate 
> travel something awful.

Sure.

> With hurdles like that, you tend to develop a habit of debugging your 
> code *away* from the machine.  Even when time was available, often 
> there simply weren't any good tools.  Try debugging a CDC 6600 PPU 
> program.  No lights, displays or anything else to help you.  If 
> you're patient and the CEs aren't too nosy (they don't like non-CE 
> types going behind the skins), you can grab a scope and read the P-
> counter of a specific PPU--assuming that your code hasn't trashed its 
> memory and rendered any postmortem dump you get meaningless.  

The fact that you ever had to grap a 'scope and read the P-counter 
seems to support my argument for real lights, not refute it :-).

> That isn't to say that I've never used a machine with an elaborate 
> front panel; I just never found one to be terribly useful in the long 
> run.

I don't claim they are useful in the long run.  They are useful in 
certain difficult debug situations, similar to what is done nowdays 
with ICE.

    Vince


More information about the cctalk mailing list