Re: Another new build problem on nt branch

From: Robert Collins <robertc@dont-contact.us>
Date: 12 Mar 2003 09:03:22 +1100

On Tue, 2003-03-11 at 23:33, Guido Serassio wrote:
> Hi Robert,

> So, for example, in squid-mswin.h connect will be:
>
> namespace Squid {
>
> inline
> int connect(int s, const struct sockaddr * n, int l)
> {
> if (::connect(_get_osfhandle(s),n,l) == SOCKET_ERROR) {
> if (WSAEMFILE == (errno = WSAGetLastError()))
> errno = EMFILE;
> return -1;
> }
> else
> return 0;
> }
>
> (I think that other similar functions comes here too)

Yes, if there are other conflicts.

> };
>
> >Then, in squid.h, add :
> >'using namespace Squid;
> >
> >What this does is makes unguarded calls to a global connect(a,b,c) use
> >that inline function. Calls that need the original connect can use
> >::connect(a,b,c).
>
> OK, but there is some risk of future namespace conflicts, aka a "Squid"
> namespace will be safe ? or it will be better something like a Squid_MSWIN
> namespace ?

It needs to be a namespace that we can use throughout squid, yet is
private to squid as far as OS includes goes : thus Squid is a good name
for it.

this is what we will end up with :
::connect(a,b,c)
::Squid::connect(a,b,c)
::ConnectState::connect()

and you can see that there are no conflicts. As long as the code that
uses connect(a,b,c) is scoped correctly (thus the using namespace Squid;
in squid.h) we will call the correct connect all the time.

Rob

-- 
GPG key available at: <http://users.bigpond.net.au/robertc/keys.txt>.

Received on Tue Mar 11 2003 - 15:03:32 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:19:32 MST