Speed up net_unittests by re-using one FTP test server instance.

Speed up net_unittests by re-using one FTP test server instance.

I managed to save 30s with this change!

I had to change the interface of the test server a bit. Now the credentials
to be used are passed (optionally) for TestPage request, not in the ctor.

BUG=none
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=24942

[wtc, 2009-08-31 20:38:44]

LGTM.  It's always nice to speed up tests!

http://codereview.chromium.org/182031/diff/1001/1002
File net/url_request/url_request_unittest.cc (right):

http://codereview.chromium.org/182031/diff/1001/1002#newcode1924
Line 1924: static void SetUpTestCase() {
Where do we call SetUpTestCase() and TearDownTestCase()?

http://codereview.chromium.org/182031/diff/1001/1003
File net/url_request/url_request_unittest.h (right):

http://codereview.chromium.org/182031/diff/1001/1003#newcode266
Line 266: return GURL(scheme_ + "://" + host_name_ + ":" + port_str_ + "/" +
path);
Is there a GURL method that combines these components into
a URL?

If host_name_ may be an IP address literal, we should be
careful about IPv6 address literals, which need to be put
inside square brackets [].  This also applies to the next
constructor.

http://codereview.chromium.org/182031/diff/1001/1003#newcode270
Line 270: const std::string& user, const std::string& password) {
This is the second form in the Style Guide:
http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Function_Decla...

We're supposed to list one parameter on each line.

http://codereview.chromium.org/182031/diff/1001/1003#newcode274
Line 274: else
Nit: don't use else after a return statement.

http://codereview.chromium.org/182031/diff/1001/1004
File third_party/pyftpdlib/README.chromium (right):

http://codereview.chromium.org/182031/diff/1001/1004#newcode3
Line 3: Chromium-specific changes are in google.patch file.
Why don't we name the file chromium.patch?

I guess we're not going to submit this patch upstream
because it is not useful in general?  I believe your change
removes the artificial 5-second delay after a login failure.
While the 5-second delay makes it hard to try different
passwords, we don't need this protection in a ftp server
used only for testing.  Correct?

Perhaps we can submit a patch upstream in which the delay
after a login failure is a parameter rather than a hardcoded
5 seconds??

You should explain this patch in this file or in the patch
itself.

http://codereview.chromium.org/182031/diff/1001/1006
File third_party/pyftpdlib/pyftpdlib/ftpserver.py (right):

http://codereview.chromium.org/182031/diff/1001/1006#newcode2549
Line 2549: auth_failed()
Why are these changes related to re-using the same FTP
test server instance?  If not, you should describe these
changes in the CL's description.