«

"santosh" <> writes:
> Default User wrote:
>> Kenneth Brody wrote:
>> > santosh wrote:
>> > > Though Turbo C++ 3.0, which the OP uses, is not given, it's 1.0
>> > > version is provided and it exhibits the same bug. The C compiler
>> > > however is correct. I wonder for how many more rarely used
>> > > constructs like these, they trip up on? Maybe, they didn't do all
>> > > the rigorous software testing that's done these days.
>> >
>> > Does the bug only occur with "for(;0"? What about:
>>
>>
>> > for ( ; a>b ; )
>> > printf("Yup, the bug is still here.\n");
>> > return(0);
>> > }
>>
>> That's a good question. What about something like:
>>
>> char s[] = "";
>>
>> for(;*s
>> printf("Yup, the bug is still here.\n");
>
> The "bug" doesn't appear, both for global and local s.
>
>> That seems common enough that it should have turned up in common usage.
>
> The bug seems to be produced only with a numeric literal. There may be
> other possibilities too, but whose's going to bother testing
> exhaustively such an outdated compiler/s.

I'll bet it's an optimization gone awry. The author(s) probably tried
to optimize the case of a constant condition and got the logic
backwards.

I wonder whether "for (;1" also misbehaves.

--
Keith Thompson (The_Other_Keith) kst- <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.

Keith Thompson wrote:
> "santosh" <> writes:
> > Default User wrote:
> >> Kenneth Brody wrote:
> >> > santosh wrote:
> >> > > Though Turbo C++ 3.0, which the OP uses, is not given, it's 1.0
> >> > > version is provided and it exhibits the same bug. The C compiler
> >> > > however is correct. I wonder for how many more rarely used
> >> > > constructs like these, they trip up on? Maybe, they didn't do all
> >> > > the rigorous software testing that's done these days.
> >> >
> >> > Does the bug only occur with "for(;0"? What about:
> >>
> >>
> >> > for ( ; a>b ; )
> >> > printf("Yup, the bug is still here.\n");
> >> > return(0);
> >> > }
> >>
> >> That's a good question. What about something like:
> >>
> >> char s[] = "";
> >>
> >> for(;*s
> >> printf("Yup, the bug is still here.\n");
> >
> > The "bug" doesn't appear, both for global and local s.
> >
> >> That seems common enough that it should have turned up in common usage.
> >
> > The bug seems to be produced only with a numeric literal. There may be
> > other possibilities too, but whose's going to bother testing
> > exhaustively such an outdated compiler/s.
>
> I'll bet it's an optimization gone awry. The author(s) probably tried
> to optimize the case of a constant condition and got the logic
> backwards.
>
> I wonder whether "for (;1" also misbehaves.

Nope. It results in an infinite loop as it should. Thus far it only
seems to occur for literal constant 0. I'm rather surprised though that
nobody seems to have spotted it in about 15 years.

"santosh" <> writes:
> Keith Thompson wrote:
>> "santosh" <> writes:
[...]
>> > The bug seems to be produced only with a numeric literal. There may be
>> > other possibilities too, but whose's going to bother testing
>> > exhaustively such an outdated compiler/s.
>>
>> I'll bet it's an optimization gone awry. The author(s) probably tried
>> to optimize the case of a constant condition and got the logic
>> backwards.
>>
>> I wonder whether "for (;1" also misbehaves.
>
> Nope. It results in an infinite loop as it should. Thus far it only
> seems to occur for literal constant 0. I'm rather surprised though that
> nobody seems to have spotted it in about 15 years.

Are you sure nobody has ever spotted it? It appears to have been
fixed in later releases, so either somebody noticed it or it vanished
in a code rewrite.

--
Keith Thompson (The_Other_Keith) kst- <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.

santosh wrote:
[...]
> > >> > Does the bug only occur with "for(;0"? What about:
[...]
> > > The bug seems to be produced only with a numeric literal. There may be
> > > other possibilities too, but whose's going to bother testing
> > > exhaustively such an outdated compiler/s.
> >
> > I'll bet it's an optimization gone awry. The author(s) probably tried
> > to optimize the case of a constant condition and got the logic
> > backwards.
> >
> > I wonder whether "for (;1" also misbehaves.
>
> Nope. It results in an infinite loop as it should. Thus far it only
> seems to occur for literal constant 0. I'm rather surprised though that
> nobody seems to have spotted it in about 15 years.

Well, the bug was fixed in a later release, either intentionally (because
someone found it) or unintentionally (in which case it was no longer there
to be "spotted").

In either case, how often does one write "for(;0"? (BTW, does "while(0)"
fail the same way?)

--
+-------------------------+--------------------+-----------------------------+
| Kenneth J. Brody | www.hvcomputer.com | |
| kenbrody/at\spamcop.net | www.fptech.com | #include <std_disclaimer.h> |
+-------------------------+--------------------+-----------------------------+
Don't e-mail me at: <private.php?do=newpm&u=>

santosh wrote:

> Default User wrote:

> > That's a good question. What about something like:
> >
> >
> > char s[] = "";
> >
> > for(;*s
> > printf("Yup, the bug is still here.\n");
>
> The "bug" doesn't appear, both for global and local s.
>
> > That seems common enough that it should have turned up in common
> > usage.
>
> The bug seems to be produced only with a numeric literal. There may be
> other possibilities too, but whose's going to bother testing
> exhaustively such an outdated compiler/s.

Ok, that makes more sense. That's a fairly unlikely construct in "real"
code, so I'm not too surprised they didn't catch it. Keith's suggestion
of an optimization bug is possible.



Brian

Kenneth Brody wrote:
> santosh wrote:
> [...]
> > > >> > Does the bug only occur with "for(;0"? What about:
> [...]
> > > > The bug seems to be produced only with a numeric literal. There may be
> > > > other possibilities too, but whose's going to bother testing
> > > > exhaustively such an outdated compiler/s.
> > >
> > > I'll bet it's an optimization gone awry. The author(s) probably tried
> > > to optimize the case of a constant condition and got the logic
> > > backwards.
> > >
> > > I wonder whether "for (;1" also misbehaves.
> >
> > Nope. It results in an infinite loop as it should. Thus far it only
> > seems to occur for literal constant 0. I'm rather surprised though that
> > nobody seems to have spotted it in about 15 years.
>
> Well, the bug was fixed in a later release, either intentionally (because
> someone found it) or unintentionally (in which case it was no longer there
> to be "spotted").
>
> In either case, how often does one write "for(;0"? (BTW, does "while(0)"
> fail the same way?)

No. Works correctly.