[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Monotone-devel] minor heads up for Windows people
From: |
Stephen Leake |
Subject: |
Re: [Monotone-devel] minor heads up for Windows people |
Date: |
Mon, 26 May 2008 08:41:02 -0400 |
User-agent: |
Gnus/5.1008 (Gnus v5.10.8) Emacs/22.2 (windows-nt) |
"Zack Weinberg" <address@hidden> writes:
> On mainline there is now a small program called "check_net" that the
> testsuite invokes to decide whether or not it should run netsync
> tests. This was needed to deal with a problem with some of the Debian
> build daemons; they disallow use even of the loopback interface by
> build jobs. The source is unix/tester-check-net.c, if anyone would
> like to see what it does. It's completely self-contained.
>
> The program is only implemented for Unix at present; it uses fork()
> and the sockets API. There is a stub in the win32 directory which
> always claims that network tests are fine. I do not know nearly enough
> about Windows network programming to implement it for that platform.
> I also don't know if a real implementation would be useful to have.
> So I'm throwing the question over to you, Windows people - take a
> look, please, and implement if you think it useful.
I guess if you are running a network firewall, it might object to the
local loopback connection.
I should think just attempting to establish a connection in exactly
the same way netsync does would be the best test. However, with some
firewall implementations, that could pop up a user dialog every time.
The user would have to choose "always permit this program"; then it
would be fine.
But we would get that behavior now from the normal netsync tests.
So until someone encounters an actual problem on Windows, I think we
can leave things as they are.
> (I'm hoping that in the medium term, like sometime this year, we'll be
> able to run these tests over local domain sockets instead of IP
> loopback, which would make the need for this go away; but netxx's code
> for that is broken and I would prefer to replace it with a
> non-moribund library, but that in turn is blocked on the library-build
> project... I assume there is *some* equivalent of AF_UNIX on
> Windows...)
That would be named pipes, and they are sufficiently different from
unix sockets that things break; that's why file: and ssh: are not
supported on Windows.
We need to switch to boost aio (or some equivalent non-moribund
library) before we try that.
--
-- Stephe