It was thus said that the Great Tony Duell once stated:
It was thus said that the Great Tony Duell once stated:
But how ywould _you_ feel if sombody kept on saying 'Oh, go on, you'll
like this steak'. Becasue that's _exactly_ how I feel when people offer
to fix my things by boardswapping. It is against the way I choose to live
my life.
Okay, I'm curious. You are fine with replacing chips and and far as I
Actually, I am not 'fine with prpleacing chips'. It does irritate me a
little that I have to rpelace a lot of still-wroking parts in soem cases.
I do prefver (as I have said many times) machines built from simpler ICs...
Of coruse ICs can't sensibly be repaired and it is not uncommon for al
lthe circuitry on an IC to get daamged at ocne (e.g. by power supply
overvoltage).
So yes, I do repalce ICs, it doesn't mean I wouldn't prefer to repace
indivdual transsitors
Heh. Okay, at least you are consistent, which I can respect. Thanks.
There are atualyl 2 aspects to baoad-swapping,
soemthign that is not
generally realised. The first is that you repalce a lot more than is
necessary, soemthing that you have already mentioned. But more important
is that people swap boards (modules, whatever) wihtout ever doing
sufficient tests to find out what is wrong -- you know 'let's try a new
video card, if that doesn't work we'll try a new motherboard, then a new
PSU, etc). The problem is that you never really know waht the fault was.
It appears to have been cured, the machien works again, but has it? Was
it perhaps jsut a bad connection? Or some marginal parameter in a part
you didn't replace that works with the nbenewew example of somethign you
did repalce but not wit hthe origianl one. And so on.
Ah, root cause analysis. I've been currently tasked with reproducing a
crash that is literally a one-in-a-million type crash [1]. We have a core
file. We have the actual location of the crash (dereferencing a NULL
pointer). The issue then becomes---why is that pointer NULL? And this
isn't code like:
do_a();
do_b();
do_c(); // crash here
do_d();
But more like
send_a();
// later
receive_a();
// later
send_b();
// later, different thread
receive_b();
// later on, different thread
send_c();
where each function may be in a different source file, but all the code is
definitely in different functions. Fun stuff.
-spc (I'm hoping that by reading email I will magically figure out what
is causing the issue 8-P
[1] We handle thousands of transactions per hour, with at least one
crash per day [2].
[2] The system itself is redundant---there are something like 100 copies
of the program running, and if one crashes, the monitoring program
will automatically restart it.