Re: Google broke recently with 3.0.4 causing repush loop

All of this is with the cookiesrc below, which lets google have
cookies on two domains.

If I:
- log out of google
-- it stalls on a meta refresh tag
- select home, which for me is https://mail.google.com/?ui=html
- provide a username and password
-- I am eventually shown my email messages
then the code below redirects:
- once during the logout
- twice before asking for a password
- eight (8) times before showing me my mail

Adding back the hack to share cookies probably makes no difference to
these figures, but I'm not sure, because I think that I remember
inconsistently getting seven and nine redirects, so I think that
google is A/B testing me.

This code cannot be put into production, because it produces messages
about redirecting 0 times reasonably often, for reasons that I haven't
tracked down yet.

--------------------------------
#host         action
DEFAULT DENY
.google.com	ACCEPT_SESSION
.google.co.nz	ACCEPT_SESSION
--------------------------------

render/dillo-jfw01$ hg diff
diff -r 07c0c28f8960 src/cache.c
--- a/src/cache.c	Fri Mar 06 13:12:11 2015 +1300
+++ b/src/cache.c	Sat May 09 22:03:31 2015 +1200
@@ -1007,11 +1007,24 @@
       return 0;
    }

-   /* if there's a redirect loop, stop now */
-   if (bw->redirect_level >= 5)
-      entry->Flags |= CA_RedirectLoop;
+//   if (0 == bw->redirect_level % 5) {
+   if (0 == bw->redirect_level % 1) {
+      char *redirMessage;
+      int choice;
+
+      asprintf(&redirMessage, "Dillo: redirected %d times",
bw->redirect_level);
+
+      choice = a_Dialog_choice(redirMessage, URL_STR(entry->Url),
+                               "Stop", "Continue", NULL);
+
+      if (1 == choice)
+         entry->Flags |= CA_RedirectLoop;
+
+      free(redirMessage);
+   }

    if (entry->Flags & CA_RedirectLoop) {
+      MSG("Redirect loop for: %s\n", URL_STR_(entry->Url));
       a_UIcmd_set_msg(bw, "ERROR: redirect loop for: %s",
URL_STR_(entry->Url));
       bw->redirect_level = 0;
       return 0;
diff -r 07c0c28f8960 src/cookies.c
--- a/src/cookies.c	Fri Mar 06 13:12:11 2015 +1300
+++ b/src/cookies.c	Sat May 09 22:03:31 2015 +1200
@@ -196,6 +196,15 @@

    if (requester == NULL) {
       /* request made by user */
+#if 0
+   } else if (!strcmp("accounts.google.co.nz", URL_HOST(requester)) &&
+              !strcmp("mail.google.com", URL_HOST(query_url))) {
+      MSG("Cookies: sent for google NZ special case 1.\n");
+   } else if (!strcmp("accounts.google.com", URL_HOST(requester)) &&
+              !strcmp("accounts.google.co.nz", URL_HOST(query_url))) {
+      MSG("Cookies: sent for google NZ special case 2.\n");
+
+#endif
    } else if (!a_Url_same_organization(query_url, requester)) {
       MSG("Cookies: not sent for request by '%s' for '%s'\n",
           URL_HOST(requester), URL_HOST(query_url));


On 09/05/2015, eocene <[email protected]> wrote:
> James wrote:
>> Here's a guess without justification: it's trying to get confirmation
>> that you accepted a cookie, and failing.
>
> I did make a quick attempt with accept_session in cookiesrc, and it
> didn't appear to make a difference...
>
>> The basis for my guess is that login to gmail is possible, but
>> requires more than five redirects, so I run a hacked up dillo 3.1,
>> with a dialog about whether to keep redirecting.
>
> How many redirects does it need? The only site that I tend to run into
> redirect problems with is nytimes.com (Not logging into any account;
> just looking at articles that people post links to...), and the last
> time I checked it was something like 12 redirects before it will give
> up and provide the page. And I wonder whether raising the redirect
> limit might be good, overall.
>
> Does gmail still need all of those redirects if you let it have a
> cookie?
>
> _______________________________________________
> Dillo-dev mailing list
> [email protected]
> http://lists.dillo.org/cgi-bin/mailman/listinfo/dillo-dev
>