5 Feb
2002
5 Feb
'02
9:24 p.m.
On Feb 6, 1:16, Chris wrote:
Get a roll-your-own. There are several pinouts for RJ45-DE9, so if you buy
one ready-made, you're going to end up muttering to Murphy. They're
usually in the form of a shell made to hold an RJ45 socket, which comes
with 8 wires connected, each with a pin (or socket pin) crimped on the end,
ready to insert into the DE9 body that comes with it. They're also cheap,
less than 1 UKP here (about $1.00) from CPC.
--
Pete Peter Turnbull
Network Manager
University of York
>I thought I had one, but I don't. Someone out
there have/know where to
find
>a cable with an RJ-45 plug on one side and a
female DE-9 (DB-25 okay but
>prefer DE-9) on the other? This is to plug my Commodore into the serial
port
on my Lantronix
EPS4+1 and attempt to get it on the network by reverse
Telnet.
I have seen RJ45 to DE9 adaptors before. I have one someplace, it came
with my DSL router. But I have seen them for sale at computer shows
before, and would think they should be readily available at any decent
electronics or network parts dealer. I think Rat Shack sells a roll your
own pinout one that goes to a DB25 (I know they have an RJ14, 6 pin to
DB25). 
6 Feb
6 Feb
7:20 a.m.
New subject: The importance of showing ALL of the code in C
an example:
int A[10];
int N;
N = 3;
A[10] = 2;
printf("%d",N);
In many other languages, A[10] = 2 would not have any effect on N.
7:34 a.m.
New subject: The importance of showing ALL of the code in C
Two more common mistakes: (that means ones that I do frequently)
N = 2;
X = 3;
. . .
if (x > 10);
{
return(N);
}
N = 7;
return(N);
and:
N = 2;
X = 3;
. . .
if (x = 10)
{
return(N);
}
N = 7;
return(N);
Look for why both sections return 2 instead of 7.
7:43 a.m.
New subject: The importance of showing ALL of the code in C
"Fred Cisin (XenoSoft)" wrote:
Two more common mistakes: (that means ones that I do
frequently)
N = 2;
X = 3;
. . .
if (x > 10);
^ BAD semicolon
{
return(N);
}
N = 7;
return(N);
and:
N = 2;
X = 3;
. . .
if (x = 10)
^ BAD equal
{
return(N);
}
N = 7;
return(N);
Look for why both sections return 2 instead of 7.
:-)
--
,,,,
/'^'\
( o o )
-oOOO--(_)--OOOo-------------------------------------
| Ram Meenakshisundaram |
| Senior Software Engineer |
| OpenLink Financial Inc |
| .oooO Phone: (516) 227-6600 x267 |
| ( ) Oooo. Email: rmeenaks(a)olf.com |
---\ (----( )--------------------------------------
\_) ) /
(_/
7:49 a.m.
New subject: The importance of showing ALL of the code in C
Capitalization.
Peace... Sridhar
On Wed, 6 Feb 2002, Fred Cisin (XenoSoft) wrote:
Two more common mistakes: (that means ones that I do
frequently)
N = 2;
X = 3;
. . .
if (x > 10);
{
return(N);
}
N = 7;
return(N);
and:
N = 2;
X = 3;
. . .
if (x = 10)
{
return(N);
}
N = 7;
return(N);
Look for why both sections return 2 instead of 7.
7:59 a.m.
New subject: The importance of showing ALL of the code in C
On Wed, 6 Feb 2002, Julius Sridhar wrote:
Capitalization.
Peace... Sridhar
Even without my failure to maintain case in the variable, it would still
return 2 instead of 7 due to premature termination of conditional
expression (semicolon), and due to assignment instead of comparison in
condition (= instead of ==). Amongst folks as careless as myself, those
two seem to be the two most common typos in C programs.
> Two more common mistakes: (that means ones that
I do frequently)
>
> N = 2;
> X = 3;
> . . .
> if (x > 10);
> {
> return(N);
> }
> N = 7;
> return(N);
>
> and:
>
> N = 2;
> X = 3;
> . . .
> if (x = 10)
> {
> return(N);
> }
> N = 7;
> return(N);
>
> Look for why both sections return 2 instead of 7.
8:20 a.m.
New subject: The importance of showing ALL of the code in C
Or it would possibly cause a segfault, since A can only be legally indexed
by 0..9. (This is assuming it didn't get optimized out).
Sometimes, you need to specify the address or pointers to addresses of
hardware registers as 'volatile', so that the compiler does not optimize
them out of doing what you expect.
And yes, seeing all the code is important. Otherwise, it's just pure
speculation.
--John
-----Original Message-----
From: owner-classiccmp(a)classiccmp.org
[mailto:owner-classiccmp@classiccmp.org]On Behalf Of Fred Cisin
(XenoSoft)
Sent: Wednesday, February 06, 2002 13:21 PM
To: classiccmp(a)classiccmp.org
Subject: The importance of showing ALL of the code in C
an example:
int A[10];
int N;
N = 3;
A[10] = 2;
printf("%d",N);
In many other languages, A[10] = 2 would not have any effect on N.
9:19 a.m.
New subject: The importance of showing ALL of the code in C
On Wed, 6 Feb 2002, John Chris Wren wrote:
Or it would possibly cause a segfault, since A can
only be legally indexed
by 0..9. (This is assuming it didn't get optimized out).
No. In C, that is actually NOT considered an error by the language! It's
"bad form", but permitted by the language to use array notation to access
memory locations that were NOT allocated to the array!
And yes, seeing all the code is important.
Otherwise, it's just pure
speculation.
9:52 a.m.
New subject: The importance of showing ALL of the code in C
"Fred Cisin (XenoSoft)" wrote:
On Wed, 6 Feb 2002, John Chris Wren wrote:
Thats because array notations are nothing more that pointer arithmetics. That is
why it is valid. It eventually becomes *(A + 10) which is *VALID* like any other
pointer operations in C. Of course, this may SEGV in certain systems. I remember
that * (A - 1) used to have the size of the array that was allocated! Dont know if
that
is true today....
Ram
--
,,,,
/'^'\
( o o )
-oOOO--(_)--OOOo-------------------------------------
| Ram Meenakshisundaram |
| Senior Software Engineer |
| OpenLink Financial Inc |
| .oooO Phone: (516) 227-6600 x267 |
| ( ) Oooo. Email: rmeenaks(a)olf.com |
---\ (----( )--------------------------------------
\_) ) /
(_/
Or it would possibly cause a segfault,
since A can only be legally indexed
by 0..9. (This is assuming it didn't get optimized out).
No. In C, that is actually NOT considered an error by the language! It's
"bad form", but permitted by the language to use array notation to access
memory locations that were NOT allocated to the array! 
11:33 a.m.
New subject: The importance of showing ALL of the code in C
I remember that * (A - 1) used to have the size of
the array that was allocated! Dont know if that is true today....
I can't imagine there's any reason that would be true except
for (compiler dependent quirk) or (argc, only, preceding argv).
In general, 'c' very often Never cares what the bounds of an array
are, and leaves it to you to stuct-ify or class-ify the array & size
together. Taking array bounds too lightly is probably the number
one problem of programmers of medium & large 'c' programs.
Allocate/Deallocate is probably the number two problem.
Yeah, and "if (a=b)..." and "a[max dimension]=b;" still do get
me from time to time, but they're the easy ones.
John A.
11:45 a.m.
New subject: The importance of showing ALL of the code in C
"Fred Cisin (XenoSoft)" wrote:
> On Wed, 6 Feb 2002, John Chris Wren wrote:
> > Or it would possibly cause a segfault, since A can only be legally
indexed
> > by 0..9. (This is assuming it didn't get optimized out).
>
> No. In C, that is actually NOT considered an error by the language! It's
> "bad form", but permitted by the language to use array notation to access
> memory locations that were NOT allocated to the array!
It is not permitted in standard C.
Thats because array notations are nothing more that
pointer arithmetics. That is
why it is valid. It eventually becomes *(A + 10) which is *VALID* like any other
pointer operations in C. Of course, this may SEGV in certain systems. I remember
that * (A - 1) used to have the size of the array that was allocated! Dont know if
that
is true today....
It all depends upon what your definition of valid is. If you use "will
compile without error" then any of the above are valid. If you use "is
specified in the language standard as conforming code" then it is definitely
not valid.
IMHO, and according to the C standard, *(A+10) or A[10] are only valid if A[10]
points to a member of the same object as A[0]. *(A-1) is only valid if
*(A-1) and *A are contained in the same object. Compilers will allow a
programmer to violate the standards, but such violations are done at the
expense of portability. Addition of pointers beyond the end of an object
isn't guaranteed to work, even if you don't dereference them.
Eric
12:06 p.m.
New subject: The importance of showing ALL of the code in C
"Eric J. Korpela" wrote:
It is not permitted in standard C.
Thats because array notations are nothing more
that pointer arithmetics. That is
why it is valid. It eventually becomes *(A + 10) which is *VALID* like any other
pointer operations in C. Of course, this may SEGV in certain systems. I remember
that * (A - 1) used to have the size of the array that was allocated! Dont know if
that
is true today....
It all depends upon what your definition of valid is. If you use "will
compile without error" then any of the above are valid. If you use "is
specified in the language standard as conforming code" then it is definitely
not valid.
IMHO, and according to the C standard, *(A+10) or A[10] are only valid if A[10]
points to a member of the same object as A[0]. *(A-1) is only valid if
*(A-1) and *A are contained in the same object. Compilers will allow a
programmer to violate the standards, but such violations are done at the
expense of portability. Addition of pointers beyond the end of an object
isn't guaranteed to work, even if you don't dereference them.
I agree with you but I have seen several DDJ articles and books about *(A - 1) as most of
the older C compilers were built from K&R C. I strongly doubt it would work today.
The
reason
why this was done IIRC was because passing arrays as arguments in a function, there were
no
way to determine the size of the array , so this is how a lot of programmers got
the size in runtime. I personally NEVER did this....
Ram
--
,,,,
/'^'\
( o o )
-oOOO--(_)--OOOo-------------------------------------
| Ram Meenakshisundaram |
| Senior Software Engineer |
| OpenLink Financial Inc |
| .oooO Phone: (516) 227-6600 x267 |
| ( ) Oooo. Email: rmeenaks(a)olf.com |
---\ (----( )--------------------------------------
\_) ) /
(_/
11:02 a.m.
New subject: The importance of showing ALL of the code in C
Perhaps I worded that poorly. I know that we vould index it by 666, if we
like, but it's liable to generate a runtime error (one a machine with a MMU,
at any rate. A non-MMU processor the effects might never show up. Or you
may tell the CP/M BIOS that your disk has 10 heads instead of 2).
I was thinking "legal" as in "I'm not so stupid as to permit this to
happen."
--John
-----Original Message-----
From: owner-classiccmp(a)classiccmp.org
[mailto:owner-classiccmp@classiccmp.org]On Behalf Of Fred Cisin
(XenoSoft)
Sent: Wednesday, February 06, 2002 15:19 PM
To: classiccmp(a)classiccmp.org
Subject: RE: The importance of showing ALL of the code in C
On Wed, 6 Feb 2002, John Chris Wren wrote:
Or it would possibly cause a segfault, since A
can only be
legally indexed
by 0..9. (This is assuming it didn't get
optimized out).
No. In C, that is actually NOT considered an error by the language! It's
"bad form", but permitted by the language to use array notation to access
memory locations that were NOT allocated to the array!
And yes, seeing all the code is important.
Otherwise, it's
just pure
speculation.
11:23 a.m.
New subject: The importance of showing ALL of the code in C
On Wed, 6 Feb 2002, John Chris Wren wrote:
Perhaps I worded that poorly. I know that we vould
index it by 666, if we
like, but it's liable to generate a runtime error (one a machine with a MMU,
at any rate. A non-MMU processor the effects might never show up. Or you
may tell the CP/M BIOS that your disk has 10 heads instead of 2).
With, OR WITHOUT, an MMU, it is legal C code!
int A[10];
int N;
. . .
A[10] = 0;
will NOT generate a runtime error.
"If you want to shoot yourself in the foot, C provides you with unlimited
ammunition."
Depending on memory model, we could overwrite other parts of our data, and
parts of THAT could be disastrous (such as following that by dividing
something by N).
With certain memory models, and with an ubelievably badly written OS, it
might be possible to step on memory that doesn't even belong to the
program. Then we could, indeed bluescreen it.
I was thinking "legal" as in "I'm
not so stupid as to permit this to
happen."
C accepts as "legal" LOTS of stuff that we hopefully would not be so
stupid as to permit to happen. And there are some professional
programmers who will do stuff like that! "By the time that we use A[10],
we no longer need N, so we just let it become an extension of the array"!
But many other languages will permit "legal" language structures that are
suicidally stupid, such as:
X = 0
Y = Z / X
--
Grumpy Ol' Fred cisin(a)xenosoft.com
11:34 a.m.
New subject: The importance of showing ALL of the code in C
--- "Fred Cisin (XenoSoft)" <cisin(a)xenosoft.com> wrote:
On Wed, 6 Feb 2002, John Chris Wren wrote:
...it's liable to generate a runtime error
With, OR WITHOUT, an MMU, it is legal C code!
int A[10];
int N;
. . .
A[10] = 0;
will NOT generate a runtime error.
True, but additionally, *depending* on A[10]=0 setting N to 0 as a side-
effect is not reliable. C won't trap that as an error, but if you
are trying it on purpose and a particular implementation does not
order variables in memory in the order you expect, or puts gaps in to
maintain data alignment, etc., you have no right to be surprised.
"If you want to shoot yourself in the foot, C
provides you with unlimited
ammunition."
As does assembler. ;-)
-ethan
__________________________________________________
Do You Yahoo!?
Send FREE Valentine eCards with Yahoo! Greetings!
http://greetings.yahoo.com
11:50 a.m.
New subject: The importance of showing ALL of the code in C
On Wed, 6 Feb 2002, John Chris Wren wrote:
This may or may not generate a runtime error. It may or may not cause
your computer to burst into flames. Do a google search on the term
"nasal demons". Upon encountering the above code, the compiler is allowed
to make demons fly out of your nose.
Eric
Perhaps I worded that poorly. I know that we
vould index it by 666, if we
like, but it's liable to generate a runtime error (one a machine with a MMU,
at any rate. A non-MMU processor the effects might never show up. Or you
may tell the CP/M BIOS that your disk has 10 heads instead of 2).
With, OR WITHOUT, an MMU, it is legal C code!
int A[10];
int N;
. . .
A[10] = 0;
will NOT generate a runtime error. 
1:22 p.m.
New subject: The importance of showing ALL of the code in C
It was thus said that the Great "Fred Cisin (XenoSoft)" once stated:
On Wed, 6 Feb 2002, John Chris Wren wrote:
C allows the *pointer* to exceed the limit by one element, but not to
*dereference* said pointer.
-spc (Well, dereferencing such a behavior invokes (I think) undefined
behavior)
Or it would possibly cause a segfault, since A
can only be legally indexed
by 0..9. (This is assuming it didn't get optimized out).
No. In C, that is actually NOT considered an error by the language! It's
"bad form", but permitted by the language to use array notation to access
memory locations that were NOT allocated to the array!
7 Feb
7 Feb
7:07 a.m.
New subject: The importance of showing ALL of the code in C
--- Sean 'Captain Napalm' Conner <spc(a)conman.org> wrote:
It was thus said that the Great "Fred Cisin
(XenoSoft)" once stated:
C allows the *pointer* to exceed the limit by one element, but not to
*dereference* said pointer.
-spc (Well, dereferencing such a behavior invokes (I think) undefined
behavior)
Exactly, but "undefined behavior" is not the same thing as illegal.
-ethan
__________________________________________________
Do You Yahoo!?
Send FREE Valentine eCards with Yahoo! Greetings!
http://greetings.yahoo.com
9:53 a.m.
New subject: The importance of showing ALL of the code in C
On Thu, 7 Feb 2002, Ethan Dicks wrote:
--- Sean 'Captain Napalm' Conner
<spc(a)conman.org> wrote:
I agree. Even "stupid and suicidal" is not the same thing as illegal
IN C!
Folks,
PLEASE be careful about attribution!
The above quotes falsely attributes to me a statement that an "out of
range" pointer can not be dereferenced. I did NOT say that! C has no
objections to dereferencing pointers to garbage. You can put ANY RANDOM
bit pattern into a pointer variable, and get garbage out.
There is a presumption in many of these posts that "stupid and suicidal"
is the same as illegal.
C has NO compunctions about permitting you to do things that you really
shouldn't do!
If you want to say that one should NEVER do something, that's fine; but do
NOT say that it is "illegal" in C.
Non-portable assumptions are NOT illegal. Stupid, yes; but NOT illegal.
A REAL C compiler does NOT have array bounds checking!
Nor do any C compilers that are on-topic for this list.
It certainly might be true that MICROS~1 might currently sell compilers
that might protect you from your mistakes, or wrap you and your program in
a safe cocoon, but that is contrary to the fundamental spirit of C.
Even ANSI C does not adhere to K&R's basic spirit of C.
It is one (of many) reasons why C is NOT necessarily the right language to
use for some projects.
"If you want to shoot yourself in the foot,
a REAL C compiler will provide you with unlimited ammunition."
--
Grumpy Ol' Fred cisin(a)xenosoft.com
It was thus said that the Great "Fred Cisin
(XenoSoft)" once stated:
C allows the *pointer* to exceed the limit by one element, but not to
*dereference* said pointer.
-spc (Well, dereferencing such a behavior invokes (I think) undefined
behavior)
Exactly, but "undefined behavior" is not the same thing as
illegal. 
6 Feb
6 Feb
9:39 a.m.
New subject: The importance of showing ALL of the code in C
At 10:20 AM 2/6/2002 -0800, you wrote:
an example:
int A[10];
int N;
In many other languages, A[10] = 2 would not have any effect on N.
And, delightfully, the trashing of the stack need not
happen so close and so obviously. :-)
- John
1:13 p.m.
New subject: The importance of showing ALL of the code in C
It was thus said that the Great "Fred Cisin (XenoSoft)" once stated:
an example:
int A[10];
int N;
N = 3;
A[10] = 2;
printf("%d",N);
In many other languages, A[10] = 2 would not have any effect on N.
And it might not have an effect on N in C either. The C compiler is free
to organize local storage any way it sees fit. So for example:
{
int x[1];
char c;
int y;
c = 'C';
y = 1;
x[2] = 0x40;
/* ... */
}
The compiler is free to organize the layout of the local storage as
(assuming ints are four bytes, represented by [ ]):
x --> [ ][ ][ ][ ]
y --> [ ][ ][ ][ ]
c --> [ ]
or:
x --> [ ][ ][ ][ ]
c --> [ ] - - - ( three bytes of padding )
y --> [ ][ ][ ][ ]
or possibly:
c --> [ ]
y --> [ ][ ][ ][ ]
x --> [ ][ ][ ][ ]
depending on various things like alignment restrictions or variable usage
(heavily used variables might be pushed closer to the base of the frame
pointer (pointer to local storage) to take advantage of smaller offsets for
instance). Such bugs are typically very hard to find. Heck, the assignment
to x[2] may be optimized out entirely if the compiler sees that nothing is
done with the result so it might go unnoticed for a long time until the code
is modified.
-spc (Been There, Done That, Have The Bug Report ... )
8197
days inactive
8198
days old
20 comments
10 participants
participants (10)
-
allain＠panix.com -
cisin＠xenosoft.com -
erd_6502＠yahoo.com -
jcwren＠jcwren.com -
jfoust＠threedee.com -
korpela＠ssl.berkeley.edu -
pete＠dunnington.u-net.com -
rmeenaks＠olf.com -
spc＠conman.org -
vance＠ikickass.org