[9fans] type signatures

[9fans] type signatures

[erik quanstrom]

i added a function to 9load's lib.h today declared incorrectly.
9load uses lib.h so one doesn't haul in the entire c library by
accident. there were no link errors. it didn't crash, but did some
difficult to explain things.  now fixed.

i believe other times i've declared an extern function incorrectly
and the linker has given me diagnostics.  but when i looked at
/386/lib/libc.a, there are no type signatures and it turns out the
default is that they are not generated.

so, am i wrong that i have seen diagnostics for incompatable
type signatures?  and, why aren't the signatures put into
object files by default?

- erik

[9fans] type signatures

[Steve Simon]

Hi,

I get a different type sygnature depending on weather the arg to a function
contains members which are in scope or not - is this expected or a bug?

for example:

/* junk.c */
typedef struct unknown unknown;
#ifdef DEF
struct unknown {
	int a;
};
#endif

typedef struct arg arg;
struct arg {
	unknown *u;		// Note pointer, not instance
};

void
func(arg a)
{
	USED(a);
}

larch% 8c -T t.c && nm -T junk.8
b0b8ed37  T func
larch% 8c -T -DDEF t.c && nm -T junk.8
763b103c  T func

-Steve

[9fans] type signatures

[Fco. J. Ballesteros]

Isn't this what
#pragma incomplete
is for?

>  From: steve@quintile.net
>  To: 9fans@9fans.net
>  Reply-To: 9fans@9fans.net
>  Date: Mon Mar  2 16:24:23 CET 2009
>  Subject: [9fans] type signatures
>
>  Hi,
>
>  I get a different type sygnature depending on weather the arg to a function
>  contains members which are in scope or not - is this expected or a bug?
>
>  for example:
>
>  /* junk.c */
>  typedef struct unknown unknown;
>  #ifdef DEF
>  struct unknown {
>  	int a;
>  };
>  #endif
>
>  typedef struct arg arg;
>  struct arg {
>  	unknown *u;		// Note pointer, not instance
>  };
>
>  void
>  func(arg a)
>  {
>  	USED(a);
>  }
>
>  larch% 8c -T t.c && nm -T junk.8
>  b0b8ed37 T func
>  larch% 8c -T -DDEF t.c && nm -T junk.8
>  763b103c T func
>
>  -Steve

Re: [9fans] type signatures

[Steve Simon]

> Isn't this what
> #pragma incomplete
> is for?

yep, I completely forgot about it.

Thanks,

-Steve

Re: [9fans] type signatures

[Charles Forsyth]

[-- Attachment #1: Type: text/plain, Size: 99 bytes --]

that seems correct. if you want an incomplete type, you can specify that
(with pragma incomplete).

[-- Attachment #2: Type: message/rfc822, Size: 2160 bytes --]

From: "Steve Simon" <steve@quintile.net>
To: 9fans@9fans.net
Subject: [9fans] type signatures
Date: Mon, 2 Mar 2009 15:23:01 +0000
Message-ID: <307e6ea2aed293668f8a7caacb1b5e2b@quintile.net>

Hi,

I get a different type sygnature depending on weather the arg to a function
contains members which are in scope or not - is this expected or a bug?

for example:

/* junk.c */
typedef struct unknown unknown;
#ifdef DEF
struct unknown {
	int a;
};
#endif

typedef struct arg arg;
struct arg {
	unknown *u;		// Note pointer, not instance
};

void
func(arg a)
{
	USED(a);
}

larch% 8c -T t.c && nm -T junk.8
b0b8ed37  T func
larch% 8c -T -DDEF t.c && nm -T junk.8
763b103c  T func

-Steve