root@qa73 ~# gcc -o test test.c
test.c:5: warning: no semicolon at end of struct or union
root@qa73 ~# cat test.c
#include
struct test {
int a;
char b
};
int main()
{
struct test t1;
printf(" %d %c ",t1.a,t1.b);
return 0;
}
root@qa73 ~# ./test
10476704 ¼ root@qa73 ~#


Hi, PLEASE notice, there is not semicolon after char b in declaration of "test" structure. Yet GCC doesnot give any compile time error?
WHy is it so? Do not counter question for gcc version ... It's RedHAt 4.0.0

Can someone shed light on it.......please?

Thanks,

this compiles even with nothing after the #include?
some weird code

Why should it ? It already gave a warning about it and this missing last colon doesn't prevent the compiler to understand the code.
Use the gcc -Werror option to have this warning abort compilation.
Of course, as already stated, gcc should have failed because of the bogus include.

Finally, don't compile or do anything as root unless there is no other choice.

It's looking for the semicolon .It doesn't matter if you put it on the 4th line of the 100th line as long as it's the next thing it sees.

99 posts hehe.

Well I know that. But you see that there is a "}" in between....

One more thing #include is proper... It was just tyoe error.... I just wanted to concetrate n below code fragment only...


what does it mean? This is a parsing error which should be reflected at compile time in any case...

Please be clear....

There is no parsing error but a code error. Please clarify.
The compiler detects the error and is trying to be helpful by accepting the code as is.
Despite the missing semicolon, there is no doubt about what the code should have been.

If it is not parsing error and compiler is trying to be helpful. Then it is giving me the "wrong warning". It gives warning as "no semicolon at end of struct" which is wrong... Semicolon is there at the end but semicolon is not at the end of one last variable declaration. So, it parses wrong...
HOwever, what if write
#include <stdio.h>
void main()
{
int i
}

It gives parsing error. It should not in this case also.

You are misinterpreting the warning message.
The compiler is complaining about the missing semicolon after the last structure field. It hasn't parsed yet the semicolon after the closing brace, which is by the way mandatory.
This has nothing to do with your previous example, a function block is not a structure.

Its not an error what the compiler said, just a convention in error handling .

Basically the compiler see this

char b};

And theres mulitude of errors it could throw back.

Invalid identifier, expected semicolon after b, no semicolon at end of struct or union

It throws errors based on the order of the keyword. Theres two key words that expect ';' at the end of statement.

struct came before the char variable. so the compiler gives an error message related to the struct keyword. which is true. There really is no semilcolon at the the end of struct. because }; denotes the end of a struct. Not char b};.

Compilers are somewhat complicated tools. <satire>If you see a problem with this behaviour, then I suggest you write a patch for it and submit it to the GCC team.</satire>

It's hard to write a patch for a problem that doesn't exist.
The compiler is correct in telling there is a missing semicolon on line 5 at the end of the structure.
The fact another semicolon is also present in this very same line but outside of the structure doesn't change anything.

Hello all,

Just to put things a bit clear... In C, any instruction should be terminated by a semicolon. The semicolon is not mandatory for block structures ( i.e {} and all between it). Notice also that the instructions within a block follow the normal rule again.

For that reason :

- a complain about no semicolon in line 5 ("c:5" in the message) is logical.

- semicolon after } is optional and treated as an empty statement. Every string, ending with a semicolon is treated as statement by the parser.

- The fact that the "}" separates the two statements is not to the point : the parser doesn't know anything about separators.

- patching this behaviour would violate the standard for C and for C++

I hope I helped to clarify out something. Enjoy programming !!!

Greetings,

Ronald

No it isn't. A semi-colon is mandatory after a structure definition, as I believe jlliagre has already pointed out.

A semi-colon is mandatory after a structure definition

Checked my own code... It seems to be so, yes... Sorry for the mistake. Not mandatory for functions although

Greetings,

ronald

Hi,
but I still dont get why wouldn't this give an error?
I mean "expected ; at line 4" should be error,