WIN32 improvements about polipo

[jech]

diskcache: don't check for leading '/' on WIN32, else serving files from the webserver never works.

Please see #7.

dns: don't even try to parse resolv.conf on WIN32

Makes sense.

io: REUSEADDR has completely different semantics on WIN32, don't use it; for better security use SO_EXCLUSIVEADDRUSE

As far as I can tell from the doc, REUSEADDR has the same semantics as under Unix. We use it in order to bind to sockets in the CLOSE_WAIT state. So I won't apply that, unless you can make sure that Windows doesn't have problems with CLOSE_WAIT on the server side.

util: try detaching from the console

Makes sense.

xnoreq, could you please generate discrete patches for the two changes that I'm accepting, and send them to me either through a github pull request or using "git send-email"?

--jch

[jech]

dns: don't even try to parse resolv.conf on WIN32

Makes sense.

Doesn't Cygwin or msys include a resolv.conf file?

[xnoreq]

Doesn't Cygwin or msys include a resolv.conf file?

In cygwin polipo is not compiled with WIN32 defined. Cygwin is like an emulation layer. Opening "/etc/resolv.conf" will actually open C:\Cygwin\etc\resolv.conf if cygwin is installed directly on C:.

With mingw it is compiled with WIN32 set and there is no emulation, so it is compiled like a normal, native windows application. Opening a path like "/etc" will always fail.

[xnoreq]

As far as I can tell from the doc, REUSEADDR has the same semantics as under Unix. We use it in order to bind to sockets in the CLOSE_WAIT state. So I won't apply that, unless you can make sure that Windows doesn't have problems with CLOSE_WAIT on the server side.

It doesn't. From my tests on WIN32 you can always bind during CLOSE_WAIT state. Setting SO_REUSEADDR will allow other applications to bind to the same address even during LISTEN state. So without the patch you can start as many polipo.exe as you want - they will all bind successfully.

In my pull request I removed the SO_EXCLUSIVEADDRUSE because it might still be desirable for multiple polipo processes to run on the same port but on different interfaces.

[jech]

Fixed by #25. THanks.