about design consideration of strcpy()

[steve yee]

strcpy() returns it's input parameter. i asked a similar question on
c++, most people seem to not think it as a problemic design. how do you
think of this?

http://groups.google.com/group/comp....0aa4307b91bad8

[steve yee]

> most people seem to not think it as a problemic design.

sorry. it should be:
most people seem to think it as a problemic design.

[Andrew Poelstra]

On 2006-05-25, steve yee <yi******@gmail.com> wrote:

most people seem to not think it as a problemic design.

sorry. it should be:
most people seem to think it as a problemic design.

I can't imagine what the C++ folks know about C. There is absolutely
nothing wrong with strcpy's design.

--
Andrew Poelstra < http://www.wpsoftware.net/blog >
To email me, use "apoelstra" at the above address.
It's just like stealing teeth from a baby.

[jacob navia]

Andrew Poelstra a écrit :

On 2006-05-25, steve yee <yi******@gmail.com> wrote:

most people seem to not think it as a problemic design.

sorry. it should be:
most people seem to think it as a problemic design.

I can't imagine what the C++ folks know about C. There is absolutely
nothing wrong with strcpy's design.

Everything is wrong in that design:

1) No provision for bounds of the character strings. strcpy is the most
often found culprit of buffer overflows and memory overwrites.
2) The return value makes no provision for error reporting. This is
in most cases a bad design.
3) The return value returns no useful information.

[Keith Thompson]

Andrew Poelstra <ap*******@wpsoftware.net> writes:

On 2006-05-25, steve yee <yi******@gmail.com> wrote:

most people seem to not think it as a problemic design.

sorry. it should be:
most people seem to think it as a problemic design.

I can't imagine what the C++ folks know about C. There is absolutely
nothing wrong with strcpy's design.

Presumably they know about the parts of C that are included by
reference in the C++ standard. That includes strcpy().

I think the point of having strcpy() return its first argument is to
allow calls to be chained together. For example:

strcat(strcpy(s1, s2), s3);

Personally, I'm not convinced this is all that useful; I find it
clearer to do one call at a time:

strcpy(s1, s2);
strcat(s1, s3);

which can also make it easier to confirm there's enough room for the
result.

I think the real answer is historical precedent.

--
Keith Thompson (The_Other_Keith) ks***@mib.org <http://www.ghoti.net/~kst>
San Diego Supercomputer Center <*> <http://users.sdsc.edu/~kst>
We must do something. This is something. Therefore, we must do this.

[Richard Heathfield]

Andrew Poelstra said:

On 2006-05-25, steve yee <yi******@gmail.com> wrote:

most people seem to not think it as a problemic design.
sorry. it should be:
most people seem to think it as a problemic design.

I can't imagine what the C++ folks know about C.

Some of them are pretty bright, you know - and some are even C experts who
happen not to like the language as much as they like C++. (Weird, I know,
but then there's nowt as queer as folk.)
There is absolutely nothing wrong with strcpy's design.

Yeah there is. It ought to return a pointer to the null terminator. Giving
back the pointer you gave it in the first place is a complete waste of
time.

--
Richard Heathfield
"Usenet is a strange place" - dmr 29/7/1999
http://www.cpax.org.uk
email: rjh at above domain (but drop the www, obviously)

[Tomás]

Richard Heathfield posted:

Some of them are pretty bright, you know - and some are even C experts
who happen not to like the language as much as they like C++. (Weird,
I know, but then there's nowt as queer as folk.)

I know my way inside out through the C subset of C++. I can do everything
I want to do in C, but I haven't got its Standard memorised as much as
the C++ Standard. For instance, yesterday I was hesitant to state the
following:
int array[5] = { 3, 4 };

/* First two elements are 3, 4, and all others are zero */
I know without a shadow of a doubt that this is true for C++, but didn't
want to post misinformation just in case I was wrong about C.

Even though I'm an expert C++ programmer, and know its every fancy
feature inside out, I'm still an expert at the lower-level stuff which
are held in common with C.

(Before I say anything: I'm not waving the C++ flag on a C newsgroup, but
since the topic's been brought up:) I myself prefer C++. That said, a
lot of my C++ code will compile as C code (i.e. I like to keep things
basic), but there are times when the more fancy features are definitely
preferable. Even when I'm writing C code, I can find plenty of places
where I'd like to use a template, but can't.

The reason I read and post on this C newsgroup is that there's a lot more
posts about the "lower-level" stuff, than you would find on a C++
newsgroup. In general, I prefer the lower-level stuff to the higher-level
stuff.

-Tomás

[Andrew Poelstra]

On 2006-05-25, Richard Heathfield <in*****@invalid.invalid> wrote:

Andrew Poelstra said:

On 2006-05-25, steve yee <yi******@gmail.com> wrote:

most people seem to not think it as a problemic design.

sorry. it should be:
most people seem to think it as a problemic design.

I can't imagine what the C++ folks know about C.

Some of them are pretty bright, you know - and some are even C experts who
happen not to like the language as much as they like C++. (Weird, I know,
but then there's nowt as queer as folk.)

Absolutely true. That's not what I meant. strcpy doesn't fit in with C++'s
idea of everything being simple and beautiful and not having pointers.
Therefore, even if a person is a C expert, if they prefer C++, their opinion
will be skewed in a C++ey way.

--
Andrew Poelstra < http://www.wpsoftware.net/blog >
To email me, use "apoelstra" at the above address.
It's just like stealing teeth from a baby.

[CBFalconer]

Richard Heathfield wrote:

Andrew Poelstra said:

On 2006-05-25, steve yee <yi******@gmail.com> wrote:

most people seem to not think it as a problemic design.

sorry. it should be:
most people seem to think it as a problemic design.

I can't imagine what the C++ folks know about C.

Some of them are pretty bright, you know - and some are even C
experts who happen not to like the language as much as they like
C++. (Weird, I know, but then there's nowt as queer as folk.)

There is absolutely nothing wrong with strcpy's design.

Yeah there is. It ought to return a pointer to the null
terminator. Giving back the pointer you gave it in the first
place is a complete waste of time.

Except that is the wrong problem. Such routines are better off
returning void, to prevent the sort of chaining that often gets
done. Consider a hypothetical strrev(s) routine, that returns the
value of s, and reverses the string in place. It can then be used
as:

printf("%s\n%s\n", s, strrev(s));

which looks clean. However the results will surprise the user,
because both printed strings are the reverse form. If the routine
returns void then the user would have written:

puts(s); strrev(s); puts(s);

and the birds would sing, the sun would shine...

--
"If you want to post a followup via groups.google.com, don't use
the broken "Reply" link at the bottom of the article. Click on
"show options" at the top of the article, then click on the
"Reply" at the bottom of the article headers." - Keith Thompson
More details at: <http://cfaj.freeshell.org/google/>
Also see <http://www.safalra.com/special/googlegroupsreply/>

[Charles Richmond]

CBFalconer wrote:

Richard Heathfield wrote:

Andrew Poelstra said:

On 2006-05-25, steve yee <yi******@gmail.com> wrote:

> most people seem to not think it as a problemic design.

sorry. it should be:
most people seem to think it as a problemic design.

I can't imagine what the C++ folks know about C.

Some of them are pretty bright, you know - and some are even C
experts who happen not to like the language as much as they like
C++. (Weird, I know, but then there's nowt as queer as folk.)

There is absolutely nothing wrong with strcpy's design.

Yeah there is. It ought to return a pointer to the null
terminator. Giving back the pointer you gave it in the first
place is a complete waste of time.

Except that is the wrong problem. Such routines are better off
returning void, to prevent the sort of chaining that often gets
done. Consider a hypothetical strrev(s) routine, that returns the
value of s, and reverses the string in place. It can then be used
as:

printf("%s\n%s\n", s, strrev(s));

which looks clean. However the results will surprise the user,
because both printed strings are the reverse form. If the routine
returns void then the user would have written:

puts(s); strrev(s); puts(s);

and the birds would sing, the sun would shine...

....or with the "strrev()" that returns the pointer, you could write:

printf("%s\n%s\n",strrev(s),strrev(s));

Now *no* matter which way things are evaluated, the strings will
be the reverse of each other... ;-)
--
+----------------------------------------------------------------+
| Charles and Francis Richmond richmond at plano dot net |
+----------------------------------------------------------------+

[Default User]

Richard Heathfield wrote:

Andrew Poelstra said:

I can't imagine what the C++ folks know about C.

Some of them are pretty bright, you know - and some are even C
experts who happen not to like the language as much as they like C++.
(Weird, I know, but then there's nowt as queer as folk.)

And some are old C programmers who happen to work in jobs that require
C++.


Brian

[CBFalconer]

Charles Richmond wrote:

CBFalconer wrote:

.... snip ...
Except that is the wrong problem. Such routines are better off
returning void, to prevent the sort of chaining that often gets
done. Consider a hypothetical strrev(s) routine, that returns the
value of s, and reverses the string in place. It can then be used
as:

printf("%s\n%s\n", s, strrev(s));

which looks clean. However the results will surprise the user,
because both printed strings are the reverse form. If the routine
returns void then the user would have written:

puts(s); strrev(s); puts(s);

and the birds would sing, the sun would shine...

...or with the "strrev()" that returns the pointer, you could write:

printf("%s\n%s\n",strrev(s),strrev(s));

Now *no* matter which way things are evaluated, the strings will
be the reverse of each other... ;-)

Surprise. No they won't be reverses of each other. Try it and
see. That is the evil caused by returning the parameter pointer as
a result. Think about it.

Here's a strrev routine for you to test with: The comments
indicate how to make it return the input pointer.

/* ======================= */
/* reverse string in place */
void revstring(char *string)
{
char *last, temp;
/* char *s = string; */

last = string + strlen(string); /* points to '\0' */
if (*string)
while (last-- > string) {
temp = *string; *string++ = *last; *last = temp;
}
/* return s */
} /* revstring */

--
"If you want to post a followup via groups.google.com, don't use
the broken "Reply" link at the bottom of the article. Click on
"show options" at the top of the article, then click on the
"Reply" at the bottom of the article headers." - Keith Thompson
More details at: <http://cfaj.freeshell.org/google/>
Also see <http://www.safalra.com/special/googlegroupsreply/>

[pete]

CBFalconer wrote:

Charles Richmond wrote:

CBFalconer wrote:

... snip ...
Except that is the wrong problem. Such routines are better off
returning void, to prevent the sort of chaining that often gets
done. Consider a hypothetical strrev(s) routine, that returns the
value of s, and reverses the string in place. It can then be used
as:

printf("%s\n%s\n", s, strrev(s));

which looks clean. However the results will surprise the user,
because both printed strings are the reverse form. If the routine
returns void then the user would have written:

puts(s); strrev(s); puts(s);

and the birds would sing, the sun would shine...

...or with the "strrev()" that returns the pointer, you could write:

printf("%s\n%s\n",strrev(s),strrev(s));

Now *no* matter which way things are evaluated, the strings will
be the reverse of each other... ;-)

Surprise. No they won't be reverses of each other.

That's a good one.
The side effects of both strrev calls are complete,
before printf's side effects begin.

--
pete

[Charles Richmond]

CBFalconer wrote:

Charles Richmond wrote:

CBFalconer wrote:

... snip ...
Except that is the wrong problem. Such routines are better off
returning void, to prevent the sort of chaining that often gets
done. Consider a hypothetical strrev(s) routine, that returns the
value of s, and reverses the string in place. It can then be used
as:

printf("%s\n%s\n", s, strrev(s));

which looks clean. However the results will surprise the user,
because both printed strings are the reverse form. If the routine
returns void then the user would have written:

puts(s); strrev(s); puts(s);

and the birds would sing, the sun would shine...

...or with the "strrev()" that returns the pointer, you could write:

printf("%s\n%s\n",strrev(s),strrev(s));

Now *no* matter which way things are evaluated, the strings will
be the reverse of each other... ;-)

You are right. I am caught again in a familiar trap. The "printf()"
just passes the pointer to the string twice, and the string remains as
it was the *last* time that "strrev()" was run.

At a PPoE, I fixed a c program that someone else had written. There
was a function to do something to a copy of a string. The function
copied the string into a static local string variable, did the
conversion there, and returned the pointer to the static local string.

The problem resulted from calling this function *twice* in the same
parameter list. It is strikingly similar to what happened with
"strrev()" above. Programmer beware!!!
--
+----------------------------------------------------------------+
| Charles and Francis Richmond richmond at plano dot net |
+----------------------------------------------------------------+