Add handling for resolving special URLs

Takes inspiration from https://github.com/lightpanda-io/browser/pull/2030 and
fixes https://github.com/lightpanda-io/browser/issues/1994

A url like http:/test gets special treatment. If the scheme, `http:` matches
the base scheme, then it's treated as relative to the base. If it doesn't match
the base scheme, then it's normalized to http://test, e.g. the path becomes
the host.

src/Notification.zig

@@ -74,8 +74,6 @@ const EventListeners = struct {
     page_network_idle: List = .{},
     page_network_almost_idle: List = .{},
     page_frame_created: List = .{},
-    page_dom_content_loaded: List = .{},
-    page_loaded: List = .{},
     http_request_fail: List = .{},
     http_request_start: List = .{},
     http_request_intercept: List = .{},
@@ -93,8 +91,6 @@ const Events = union(enum) {
     page_network_idle: *const PageNetworkIdle,
     page_network_almost_idle: *const PageNetworkAlmostIdle,
     page_frame_created: *const PageFrameCreated,
-    page_dom_content_loaded: *const PageDOMContentLoaded,
-    page_loaded: *const PageLoaded,
     http_request_fail: *const RequestFail,
     http_request_start: *const RequestStart,
     http_request_intercept: *const RequestIntercept,
@@ -141,18 +137,6 @@ pub const PageFrameCreated = struct {
     timestamp: u64,
 };
 
-pub const PageDOMContentLoaded = struct {
-    req_id: u32,
-    frame_id: u32,
-    timestamp: u64,
-};
-
-pub const PageLoaded = struct {
-    req_id: u32,
-    frame_id: u32,
-    timestamp: u64,
-};
-
 pub const RequestStart = struct {
     transfer: *Transfer,
 };

src/browser/HttpClient.zig

@@ -1325,15 +1325,15 @@ pub const Transfer = struct {
             }
         }
 
-        transfer.req.notification.dispatch(.http_response_header_done, &.{
-            .transfer = transfer,
-        });
-
         const proceed = transfer.req.header_callback(transfer) catch |err| {
             log.err(.http, "header_callback", .{ .err = err, .req = transfer });
             return err;
         };
 
+        transfer.req.notification.dispatch(.http_response_header_done, &.{
+            .transfer = transfer,
+        });
+
         return proceed and transfer.aborted == false;
     }
 

src/browser/Page.zig

@@ -487,6 +487,7 @@ pub fn navigate(self: *Page, request_url: [:0]const u8, opts: NavigateOpts) !voi
                 return error.InjectBlankFailed;
             };
         }
+        self.documentIsComplete();
 
         session.notification.dispatch(.page_navigate, &.{
             .frame_id = self._frame_id,
@@ -518,8 +519,6 @@ pub fn navigate(self: *Page, request_url: [:0]const u8, opts: NavigateOpts) !voi
 
         // force next request id manually b/c we won't create a real req.
         _ = session.browser.http_client.incrReqId();
-
-        self.documentIsComplete();
         return;
     }
 
@@ -739,12 +738,6 @@ pub fn _documentIsLoaded(self: *Page) !void {
         self.document.asEventTarget(),
         event,
     );
-
-    self._session.notification.dispatch(.page_dom_content_loaded, &.{
-        .frame_id = self._frame_id,
-        .req_id = self._req_id,
-        .timestamp = timestamp(.monotonic),
-    });
 }
 
 pub fn scriptsCompletedLoading(self: *Page) void {
@@ -803,6 +796,19 @@ pub fn documentIsComplete(self: *Page) void {
     self._documentIsComplete() catch |err| {
         log.err(.page, "document is complete", .{ .err = err, .type = self._type, .url = self.url });
     };
+
+    if (self._navigated_options) |no| {
+        // _navigated_options will be null in special short-circuit cases, like
+        // "navigating" to about:blank, in which case this notification has
+        // already been sent
+        self._session.notification.dispatch(.page_navigated, &.{
+            .frame_id = self._frame_id,
+            .req_id = self._req_id,
+            .opts = no,
+            .url = self.url,
+            .timestamp = timestamp(.monotonic),
+        });
+    }
 }
 
 fn _documentIsComplete(self: *Page) !void {
@@ -821,12 +827,6 @@ fn _documentIsComplete(self: *Page) !void {
         try self._event_manager.dispatchDirect(window_target, event, self.window._on_load, .{ .inject_target = false, .context = "page load" });
     }
 
-    self._session.notification.dispatch(.page_loaded, &.{
-        .frame_id = self._frame_id,
-        .req_id = self._req_id,
-        .timestamp = timestamp(.monotonic),
-    });
-
     if (self._event_manager.hasDirectListeners(window_target, "pageshow", self.window._on_pageshow)) {
         const pageshow_event = (try PageTransitionEvent.initTrusted(comptime .wrap("pageshow"), .{}, self)).asEvent();
         try self._event_manager.dispatchDirect(window_target, pageshow_event, self.window._on_pageshow, .{ .context = "page show" });
@@ -879,19 +879,6 @@ fn pageHeaderDoneCallback(transfer: *HttpClient.Transfer) !bool {
         });
     }
 
-    if (self._navigated_options) |no| {
-        // _navigated_options will be null in special short-circuit cases, like
-        // "navigating" to about:blank, in which case this notification has
-        // already been sent
-        self._session.notification.dispatch(.page_navigated, &.{
-            .frame_id = self._frame_id,
-            .req_id = self._req_id,
-            .opts = no,
-            .url = self.url,
-            .timestamp = timestamp(.monotonic),
-        });
-    }
-
     return true;
 }
 

src/browser/URL.zig

@@ -77,6 +77,41 @@ pub fn resolve(allocator: Allocator, base: [:0]const u8, path: anytype, comptime
         return processResolved(allocator, result, opts);
     }
 
+    if (path.len >= 4) { // Minimum: "ws:x"
+        if (std.mem.indexOfScalar(u8, path[0..@min(path.len, 6)], ':')) |pos| {
+            // we know this isn't a complete URL, else the very first check in
+            // this function would have handled it.
+            const possible_special_protocol = path[0..pos];
+            const special_schemes = [_][]const u8{ "https", "http", "ws", "wss", "file", "ftp" };
+            for (special_schemes) |special_scheme| {
+                if (std.ascii.eqlIgnoreCase(possible_special_protocol, special_scheme)) {
+                    const rest = path[pos + 1 ..];
+
+                    // Check if base has the same scheme
+                    const base_scheme_end = std.mem.indexOf(u8, base, "://") orelse 0;
+                    if (base_scheme_end > 0 and std.ascii.eqlIgnoreCase(base[0..base_scheme_end], special_scheme)) {
+                        // Same scheme - strip it and resolve rest as relative
+                        return resolve(allocator, base, rest, opts);
+                    }
+
+                    // Different scheme - construct absolute URL
+                    // Skip any leading slashes in rest
+                    var rest_start: usize = 0;
+                    while (rest_start < rest.len and (rest[rest_start] == '/' or rest[rest_start] == '\\')) {
+                        rest_start += 1;
+                    }
+                    const rest_trimmed = rest[rest_start..];
+
+                    // file: scheme needs empty host (triple slash)
+                    const separator = if (std.mem.eql(u8, special_scheme, "file")) ":///" else "://";
+                    const normalized = try std.mem.joinZ(allocator, "", &.{ special_scheme, separator, rest_trimmed });
+                    return resolve(allocator, "", normalized, opts);
+                }
+            }
+            // Don't know what this is, just try to resolve it through our normal logic
+        }
+    }
+
     const scheme_end = std.mem.indexOf(u8, base, "://");
     const authority_start = if (scheme_end) |end| end + 3 else 0;
     const path_start = std.mem.indexOfScalarPos(u8, base, authority_start, '/') orelse base.len;
@@ -1570,3 +1605,84 @@ test "URL: getOrigin" {
         }
     }
 }
+
+test "URL: resolve path scheme" {
+    const Case = struct {
+        base: [:0]const u8,
+        path: [:0]const u8,
+        expected: [:0]const u8,
+    };
+
+    const cases = [_]Case{
+        .{
+            .base = "https://www.example.com/example",
+            .path = "https:/about",
+            .expected = "https://www.example.com/about",
+        },
+        .{
+            .base = "https://www.example.com/example",
+            .path = "https:about",
+            .expected = "https://www.example.com/about",
+        },
+        .{
+            .base = "https://www.example.com/example",
+            .path = "https://about",
+            .expected = "https://about",
+        },
+        .{
+            .base = "https://www.example.com/example",
+            .path = "http:about",
+            .expected = "http://about",
+        },
+        .{
+            .base = "https://www.example.com/example",
+            .path = "http:/about",
+            .expected = "http://about",
+        },
+        .{
+            .base = "https://www.example.com/example",
+            .path = "http://about",
+            .expected = "http://about",
+        },
+        .{
+            .base = "https://site/",
+            .path = "https://path",
+            .expected = "https://path",
+        },
+        .{
+            .base = "http://localhost/",
+            .path = "data:test",
+            .expected = "data:test",
+        },
+        .{
+            .base = "https://www.example.com/example",
+            .path = "ws://about",
+            .expected = "ws://about",
+        },
+        .{
+            .base = "https://www.example.com/example",
+            .path = "wss://about",
+            .expected = "wss://about",
+        },
+        .{
+            .base = "https://www.example.com/example",
+            .path = "ftp://about",
+            .expected = "ftp://about",
+        },
+        .{
+            .base = "https://www.example.com/example",
+            .path = "file://path/to/file",
+            .expected = "file://path/to/file",
+        },
+        .{
+            .base = "https://www.example.com/example",
+            .path = "file:/path/to/file",
+            .expected = "file:///path/to/file",
+        },
+    };
+
+    for (cases) |case| {
+        const result = try resolve(testing.arena_allocator, case.base, case.path, .{});
+        try testing.expectString(case.expected, result);
+    }
+}

src/cdp/CDP.zig

@@ -427,8 +427,6 @@ pub fn BrowserContext(comptime CDP_T: type) type {
             try notification.register(.page_navigate, self, onPageNavigate);
             try notification.register(.page_navigated, self, onPageNavigated);
             try notification.register(.page_frame_created, self, onPageFrameCreated);
-            try notification.register(.page_dom_content_loaded, self, onPageDOMContentLoaded);
-            try notification.register(.page_loaded, self, onPageLoaded);
         }
 
         pub fn deinit(self: *Self) void {
@@ -604,16 +602,6 @@ pub fn BrowserContext(comptime CDP_T: type) type {
             return @import("domains/page.zig").pageFrameCreated(self, msg);
         }
 
-        pub fn onPageDOMContentLoaded(ctx: *anyopaque, msg: *const Notification.PageDOMContentLoaded) !void {
-            const self: *Self = @ptrCast(@alignCast(ctx));
-            return @import("domains/page.zig").pageDOMContentLoaded(self, msg);
-        }
-
-        pub fn onPageLoaded(ctx: *anyopaque, msg: *const Notification.PageLoaded) !void {
-            const self: *Self = @ptrCast(@alignCast(ctx));
-            return @import("domains/page.zig").pageLoaded(self, msg);
-        }
-
         pub fn onPageNetworkIdle(ctx: *anyopaque, msg: *const Notification.PageNetworkIdle) !void {
             const self: *Self = @ptrCast(@alignCast(ctx));
             return @import("domains/page.zig").pageNetworkIdle(self, msg);

src/cdp/domains/page.zig

@@ -385,6 +385,7 @@ pub fn pageNavigated(arena: Allocator, bc: anytype, event: *const Notification.P
     // things, but no session.
     const session_id = bc.session_id orelse return;
 
+    const timestamp = event.timestamp;
     const frame_id = &id.toFrameId(event.frame_id);
     const loader_id = &id.toLoaderId(event.req_id);
 
@@ -436,9 +437,9 @@ pub fn pageNavigated(arena: Allocator, bc: anytype, event: *const Notification.P
     const page = bc.session.currentPage() orelse return error.PageNotLoaded;
 
     // When we actually recreated the context we should have the inspector send
-    // this event, see: resetContextGroup. Sending this event will tell the
-    // client that the context ids they had are invalid and the context should
-    // be dropped. The client will expect us to send new contextCreated events,
+    // this event, see: resetContextGroup Sending this event will tell the
+    // client that the context ids they had are invalid and the context shouls
+    // be dropped The client will expect us to send new contextCreated events,
     // such that the client has new id's for the active contexts.
     // Only send executionContextsCleared for main frame navigations. For child
     // frames (iframes), clearing all contexts would destroy the main frame's
@@ -448,18 +449,6 @@ pub fn pageNavigated(arena: Allocator, bc: anytype, event: *const Notification.P
         try cdp.sendEvent("Runtime.executionContextsCleared", null, .{ .session_id = session_id });
     }
 
-    // frameNavigated event
-    try cdp.sendEvent("Page.frameNavigated", .{
-        .type = "Navigation",
-        .frame = Frame{
-            .id = frame_id,
-            .url = event.url,
-            .loaderId = loader_id,
-            .securityOrigin = bc.security_origin,
-            .secureContextType = bc.secure_context_type,
-        },
-    }, .{ .session_id = session_id });
-
     {
         const aux_data = try std.fmt.allocPrint(arena, "{{\"isDefault\":true,\"type\":\"default\",\"frameId\":\"{s}\",\"loaderId\":\"{s}\"}}", .{ frame_id, loader_id });
 
@@ -509,22 +498,18 @@ pub fn pageNavigated(arena: Allocator, bc: anytype, event: *const Notification.P
     // chromedp client expects to receive the events is this order.
     // see https://github.com/chromedp/chromedp/issues/1558
     try cdp.sendEvent("DOM.documentUpdated", null, .{ .session_id = session_id });
-}
-
-pub fn pageDOMContentLoaded(bc: anytype, event: *const Notification.PageDOMContentLoaded) !void {
-    const session_id = bc.session_id orelse return;
-    const timestamp = event.timestamp;
-    var cdp = bc.cdp;
 
+    // domContentEventFired event
+    // TODO: partially hard coded
     try cdp.sendEvent(
         "Page.domContentEventFired",
         .{ .timestamp = timestamp },
         .{ .session_id = session_id },
     );
 
+    // lifecycle DOMContentLoaded event
+    // TODO: partially hard coded
     if (bc.page_life_cycle_events) {
-        const frame_id = &id.toFrameId(event.frame_id);
-        const loader_id = &id.toLoaderId(event.req_id);
         try cdp.sendEvent("Page.lifecycleEvent", LifecycleEvent{
             .timestamp = timestamp,
             .name = "DOMContentLoaded",
@@ -532,23 +517,16 @@ pub fn pageDOMContentLoaded(bc: anytype, event: *const Notification.PageDOMConte
             .loaderId = loader_id,
         }, .{ .session_id = session_id });
     }
-}
-
-pub fn pageLoaded(bc: anytype, event: *const Notification.PageLoaded) !void {
-    const session_id = bc.session_id orelse return;
-    const timestamp = event.timestamp;
-    var cdp = bc.cdp;
-
-    const frame_id = &id.toFrameId(event.frame_id);
 
+    // loadEventFired event
     try cdp.sendEvent(
         "Page.loadEventFired",
         .{ .timestamp = timestamp },
         .{ .session_id = session_id },
     );
 
+    // lifecycle DOMContentLoaded event
     if (bc.page_life_cycle_events) {
-        const loader_id = &id.toLoaderId(event.req_id);
         try cdp.sendEvent("Page.lifecycleEvent", LifecycleEvent{
             .timestamp = timestamp,
             .name = "load",
@@ -557,6 +535,7 @@ pub fn pageLoaded(bc: anytype, event: *const Notification.PageLoaded) !void {
         }, .{ .session_id = session_id });
     }
 
+    // frameStoppedLoading
     return cdp.sendEvent("Page.frameStoppedLoading", .{
         .frameId = frame_id,
     }, .{ .session_id = session_id });