cURL / Mailing Lists / curl-library / Single Mail

curl-library

Re: NSS and configure and pkg-config

From: Guenter <lists_at_gknw.net>
Date: Mon, 07 Sep 2009 00:00:05 +0200

Hi,
Kamil Dudka schrieb:
> On Sunday 06 of September 2009 23:07:21 Daniel Stenberg wrote:
>> The argument seems to be that they don't want to specify the arguments as
>> that function pointer is used as a generic function pointer which can call
>> functions using different signatures. Similar to how a plain void * would
>> be used for generic data.
>
> Any idea how to solve this? I don't think void* could solve this as
> sizeof(void*) != sizeof (PRFuncPtr) on some obscure archs.
not tested, but a friend suggested (...), will soon try if that works...

>> Personally I think they solve that problem the wrong way, but that doesn't
>> help us. I figure we'll need to live with it. The NSS project clearly
>> doesn't think compiler warnings are as important to fix as we do.
>
> I don't think so since the bug is still opened. I can see bugs with higher
> priority and/or severity which are being solved now instead. You are always
> welcome with an idea how to solve it. The only requirement is to not break
> binary/source compatibility.
>
> I personally like the idea 1) from comment #11. Challenge? Note I haven't yet
> dug through the gcc mailing list if it has been already proposed or not.
I did yesterday already, and came to same thread which is mentioned in
the RedHat BZ, but this isnt helpful - either -Wno-strict-prototypes or
void was suggested there (which lead me quickly to the void fix);
however if void is not acceptable then we should figure out what's the
counterpart pragma for:
#pragma GCC diagnostic ignored "-Wstrict-prototypes"
if we find out about that we could just switch off the warnings before
we include *bad* headers, and afterwards turn on again ...

Gün.
Received on 2009-09-07