Re: Porting from visual C to gcc problem

On Fri, 19 Aug 2005 22:05:48 -0000, Gustin.Kiffney@xxxxxxxxx wrote in
comp.lang.c.moderated:

I will add some detail interspersed in your code.

> This code is in a file dnis.h:
>
> typedef DNIS_ERROR (*DNIS_TIMER_CALLBACK) (struct DNIS_TIMER_CB
> *timerCB);

The declaration above defines the identifier 'DNIS_TIMER_CALLBACK' as
an alias for "a pointer to a function that returns an object of type
'DNIS_ERROR', and takes a single parameter that is a pointer to a
'struct DNIS_TIMER_CS'.

Somewhere before this, the compiler must have come across the typedef
or macro defining 'DNIS_ERROR', otherwise you would have gotten
another error first.

> typedef struct DNIS_TIMER_CB
> {
> UTIL_LIST_ENTRY listEntry;
> DNIS_SIGNATURE_TYPE signature;
> DNIS_TIMER_ID timerID;
> DNIS_TIME timeout;
> DNIS_TIME expiration;
> DNIS_TIMER_CALLBACK callBack;
> ULONG context;
> } DNIS_TIMER_CB;

Now here, after that typedef, is a typedef that includes a definition
of a structure type named 'struct DNS_TIMER_CB'.

> gcc doesn't like this.
>
> When the compiler hits this definition, it says
> ../Hdr/dnis.h:68: warning: `struct DNIS_TIMER_CB' declared inside
> parameter list
> ../Hdr/dnis.h:68: warning: its scope is only this definition or
> declaration, which is probably not what you want

gcc is correct here, and Visual C++ is wrong. But the gcc error
message might have been a little clearer.

C is a scoped language, and one of the scopes is "prototype scope".
That basically means that there is a scope that begins at the '(' in a
function declaration, and ends at the ')'. Any types or identifiers
defined inside the '(' and the ')' go out of scope at the end. The
only exception to this is when the function declaration also defines
the body of the function, then the scope lasts to through the body to
the closing '}'.

That is why you can have this in a C file:

int func(int a, int b);
int func(int c, int d);

int func(int x, int y)
{
return x + y;
}

There are three declarations of 'func', but there is no conflict with
different names for the parameters. That is because the names in the
two prototypes are forgotten as soon as the ';' closes the declaration
without a body.

So the typedef for 'DNIS_TIMER_CALLBACK' includes a declaration of an
incomplete type 'struct DNIS_TIMER_CB' and says the function accepts a
pointer to this type. Well and good. But if the compiler has not
already seen a definition of 'struct DNIS_TIMER_CB', this declaration
has prototype scope, and ends at the ';'. In particular, it is a
completely different type and completely unrelated to anything defined
in an outer scope afterwards.

> and all references to DNIS_TIMER_CB explode with, for example,
>
> ../Hdr/switch.h:39: error: syntax error before "DNIS_TIMER_CB"
> ../Hdr/switch.h:39: warning: no semicolon at end of struct or union
>
> I am trying to get this code to run under Unix/Linux and I'm not very
> experienced in C. I'm not at all sure what this
> typedef DNIS_ERROR (*DNIS_TIMER_CALLBACK) (struct DNIS_TIMER_CB
> timerCB);
>
> is supposed to mean or do; it looks frankly like a mess to me. Anyone
> know what the programmer meant by this? Any way to make this portable
> to gcc/Linux?

The programmer who wrote this depended on non-standard behavior by his
compiler. It might be a VC++ bug, or a deliberate Microsoft
extension. Either way, your experience with gcc shows the dangers in
depending on non-standard extensions, especially when you try to port
the code.

In any case, the fix is quite simple. Move the complete definition
with typedef of the 'DNIS_TIMER_CB' structure above the typedef of the
'DNIS_TIMER_CALLBACK' function pointer. That's all it takes. Then
when the compiler sees the 'struct DNIS_TIMER_CB *timerCB', it will
not take it as a new definition of a new type, because it already has
a definition of the structure type in scope.

--
Jack Klein
Home: http://JK-Technology.Com
FAQs for
comp.lang.c http://www.eskimo.com/~scs/C-faq/top.html
comp.lang.c++ http://www.parashift.com/c++-faq-lite/
alt.comp.lang.learn.c-c++
http://www.contrib.andrew.cmu.edu/~ajo/docs/FAQ-acllc.html
--
comp.lang.c.moderated - moderation address: clcm@xxxxxxxxxxxx -- you must
have an appropriate newsgroups line in your header for your mail to be seen,
or the newsgroup name in square brackets in the subject line. Sorry.
.

Relevant Pages

  • compiling kernel
    ... scripts/kconfig/qconf.h:51: error: `e' was not declared in this scope ... scripts/kconfig/qconf.h:73: error: `int updateList' redeclared as different ... scripts/kconfig/qconf.h:8: error: forward declaration of `class ConfigList' ... ConfigLineEdit' ...
    (alt.os.linux.suse)
  • Re: syntax errror
    ... >>> int maxGrey' ... >> declaration of maxGrey. ... typedef double; ... struct some_tag_name { ...
    (comp.lang.c)
  • Re: mutually referential (Pg 140 K&R2)
    ... >In this example the vacuous declaration doesn't make any difference; ... >needed when you want a name in an inner scope to match a tag defined ... >in the inner scope instead of a tag from the outer scope. ... the vacuous declarations of struct ...
    (comp.lang.c)
  • Re: Porting from visual C to gcc problem
    ... the compiler must have come across the typedef ... >>function declaration, ... then the scope lasts to through the body to ... >>int func; ...
    (comp.lang.c.moderated)
  • Re: void * vs char *
    ... cast p to a (struct s*) to avoid a compile-time error. ... struct s {int a;}; ... function call consists solely of an identifier, and if no declaration ... And see what your compiler has to say. ...
    (comp.lang.c)