Bug in Arduino compiler.

[Pommie]

I recently had a bug in my program and couldn't figure it out.
Turns out I'd typed count==25; instead of count=25;
I thought this would be an error or at least a warning. It isn't.

As an experiment I placed the following in a new sketch,

void setup(){
  uint8_t count;
  count==25;
  Serial.begin(115200);
  Serial.println(count);
}

void loop(){}

It compiles fine but count is displayed as 0.

Mike.
Edit, is this normal compiler behaviour? Just tried it in XC8 with the same result.

 

[danadak]

In C count=25 is an assign a value statement, whereas count==25 is a test for equality
of the two operands.

Excellent book on C for ref library :

https://venkivasamsetti.github.io/ebookworm.github.io/Books/cse/C%20Programming%20Language%20(2nd%20Edition).pdf

Regards, Dana.

 

[Pommie]

I know the difference, I was expressing my surprise that it doesn't generate an error or warning.

Mike.

 

[danadak]

I am not sure about this, from an ANSI C point of view, but the ==
statement does not have to be in parenthesis so compiler cannot
figure out what its supposed to be. And vice versa, if its single =
is it supposed to be test, a == ?

Maybe a C expert can comment on this.


Regards. Dana.

 

[Pommie]

I just consider it a syntax error and it should be flagged as such.

I'm just wondering if there is a construct along the lines of strcpy(buff,count==25?"is 25":"isn't") that I'm not aware of.

Mike.

 

[Papabravo]

It is not a syntax error. As written, it is syntactically correct. It is a logical expression that evaluates to TRUE or FALSE. It is semantically wrong, because a free-standing logical expression has no effect on the overall program. I do agree that a "smart" compiler could throw a warning for useless expressions masquerading as statements. A really smart compiler would just optimize it away into nothingness and maybe mention the fact that you're being wasteful by coding free standing expressions that have no effect.

 

[Nigel Goodwin]

I just consider it a syntax error and it should be flagged as such.

It's not a syntax error, it's perfectly correct C code - so can't be flagged as a syntax error (and yes, I do it a LOT as well )

I'm just wondering if there is a construct along the lines of strcpy(buff,count==25?"is 25":"isn't") that I'm not aware of.

Mike.

If you write it slightly differently though, it WILL trigger an error:

'25==count' does exactly the same, but '25=count' generates an error (because you're writing to a constant).

If you look at MCC created code, it's often done that way.

 

[granddad]

Turns out I'd typed count==25; instead of count=25;

Caught me out a few times too, Should be an error I guess. what about .. if(count=2) .... not sure XC8 errors ?

 

[Pommie]

Using 25==count makes sense when used as a comparison but using an assignment backwards doesn't make sense. You are simply introducing a deliberate error!

Note, the construct I wrote, strcpy(buff,count==25?"is 25":"isn't") is valid in C.

I was wondering what use count==25 might have (if any).

and yes, I do it a LOT as well

I have only ever done this once. I often do if(count=25) (which I suspect you mean) but that's a completely different problem.

The problem is a line which only contains count==25; is valid in C but I've no idea why.
XC8 does the same thing if(x=5) gives a warning but x==5; doesn't!!!

Mike.

 

[Pommie]

Caught me out a few times too, Should be an error I guess. what about .. if(count=2) .... not sure XC8 errors ?

In that case xc8 gives a warning.
warning: using the result of an assignment as a condition without parentheses [-Wparentheses]

But not if you type (only) LATA==0; It just compiles to nothing.

Mike.

 

[Papabravo]

Using 25==count makes sense when used as a comparison but using an assignment backwards doesn't make sense. You are simply introducing a deliberate error!

Note, the construct I wrote, strcpy(buff,count==25?"is 25":"isn't") is valid in C.

I was wondering what use count==25 might have (if any).


I have only ever done this once. I often do if(count=25) (which I suspect you mean) but that's a completely different problem.

The problem is a line which only contains count==25; is valid in C but I've no idea why.
XC8 does the same thing if(x=5) gives a warning but x==5; doesn't!!!

Mike.

It is valid BECAUSE any expression is a valid statement period FULL STOP.
This is similar to x++; which is a valid expression that can turn into a statement. This is also similar to a function call, is evaluated as an expression and returns a value. They are all syntactically the same, but semantically different.

A smart compiler that realizes that an evaluated expression has no effect is free to optimize the evaluation away without effecting the operation; it assumes you knew what you were doing, which is not always the case. If it was really benevolent it might throw a warning to draw your attention to the matter.

 

[Pommie]

Yes, still doesn't help catching a typo. Most other typos I can think of result in an Error or Warning.
Think I'll look for a lint type program.

Mike.

 

[Papabravo]

Yes, still doesn't help catching a typo. Most other typos I can think of result in an Error or Warning.
Think I'll look for a lint type program.

Mike.

I'll be curious to learn what you find. I've never had the need to take such an action.

 

[Pommie]

I'll probably wait until next time I've got a bug I can't find. However, forewarned is forearmed. As I said above, I've only done this once.

Mike.

 

[Papabravo]

I'll probably wait until next time I've got a bug I can't find. However, forewarned is forearmed. As I said above, I've only done this once.

Mike.

I cross my fingers in your favor and hope for the best result possible.

 

[granddad]

MPLABX IDE warnings , must be written for programmers.

 

[tepalia02]

It's a typo but not a syntax error. That's why Arduino IDE could not catch it.