Re: C_ASSERT portability fixes and simplification [PATCH 1/2]

[Alexandre Julliard]

Gerald Pfeifer  writes:
> This patch moves these into main() in main.c, which has several 
> advantages:
> - The asserts are only evaluated once, not for every #inclusion
> of request.h, which simplifies/speeds up the compilation.

It doesn't make any difference, that stuff is inside a #ifdef.
Also it's auto-generated, so moving it manually won't help.
> - More importantly, these were the only occurences of C_ASSERT at
> file level.  By moving these at function level use thereof becomes
> consistent which will allow the fix from PATCH 2/2 to work at all.

C_ASSERT is supposed to work at the file level too.

C_ASSERT portability fixes and simplification [PATCH 1/2]

[Gerald Pfeifer]

On Mon, 11 May 2009, Alexandre Julliard wrote:
> > - More importantly, these were the only occurences of C_ASSERT at
> > file level.  By moving these at function level use thereof becomes
> > consistent which will allow the fix from PATCH 2/2 to work at all.
>
> C_ASSERT is supposed to work at the file level too.

The current implementation doesn't work with GCC 4.5 (or ISO C)
and I failed to find any other way. :-(

Your point on this being autogenerated clearly is strong one --
either we need to change that generation or find a different way
to implement C_ASSERT.

Any idea how to best approach this?

Gerald

C_ASSERT portability fixes and simplification [PATCH 1/2]

[Alexandre Julliard]

Gerald Pfeifer  writes:
> The current implementation doesn't work with GCC 4.5 (or ISO C)
> and I failed to find any other way. :-(
>
> Your point on this being autogenerated clearly is strong one --
> either we need to change that generation or find a different way
> to implement C_ASSERT.
>
> Any idea how to best approach this?

I don't have gcc 4.5 to test this, but you could try using a function
instead, something like this:

diff --git a/include/winnt.h b/include/winnt.h
index abcc502..aaa4112 100644
--- a/include/winnt.h
+++ b/include/winnt.h
@@ -292,7 +292,7 @@ extern "C" {
 #if defined(_MSC_VER)
 # define C_ASSERT(e) typedef char __C_ASSERT__[(e)?1:-1]
 #elif defined(__GNUC__) 
-# define C_ASSERT(e) extern char __C_ASSERT__[(e)?1:-1] __attribute__((unused))
+# define C_ASSERT(e) extern void __C_ASSERT__(int [(e)?1:-1])
 #else
 # define C_ASSERT(e)
 #endif

C_ASSERT portability fixes and simplification [PATCH 1/2]

[Gerald Pfeifer]

On Mon, 11 May 2009, Alexandre Julliard wrote:
> I don't have gcc 4.5 to test this, but you could try using a function
> instead, something like this:
>
> diff --git a/include/winnt.h b/include/winnt.h
> index abcc502..aaa4112 100644
> --- a/include/winnt.h
> +++ b/include/winnt.h
> @@ -292,7 +292,7 @@ extern "C" {
>  #if defined(_MSC_VER)
>  # define C_ASSERT(e) typedef char __C_ASSERT__[(e)?1:-1]
>  #elif defined(__GNUC__) 
> -# define C_ASSERT(e) extern char __C_ASSERT__[(e)?1:-1] __attribute__((unused))
> +# define C_ASSERT(e) extern void __C_ASSERT__(int [(e)?1:-1])
>  #else
>  # define C_ASSERT(e)
>  #endif

You are a clever man, Sir! :-)

It took me a bit, since incidently the GCC 4.5 development tree was
broken wrt. Wine on Monday (I reported this bug upstream), but a
build with your patch went through both with GCC 4.2.1 and GCC 4.5.0 
(experimental) tonight.

I assume you will commit this now?

Thanks,
Gerald