On Thu, Apr 26, 2007 at 10:25:47PM +0200, Daniel Stenberg wrote:
> On Thu, 26 Apr 2007, Peter O'Gorman wrote:

> >which is not so pleasant. At the moment, we are thinking that gnutls
> >should have some api to allow its initialization to fail without exiting
> >the application.
>
> I completely agree.
>

When I looked closer, I decided to patch curl so that it would only
attempt to init GnuTLS if it were actually used. Thus, `curl
--version' will not look for an egd socket, but `curl
https://example.com' will.

That patch is attached.

Thanks,
Peter