Skip to content

Lightpanda: the headless browser designed for AI and automation

cdp: report a non-UTF-8 status text and response headers as strings

Lucien Coffefdb3bf247158
diff --git a/src/server/cdp/domains/fetch.zig b/src/server/cdp/domains/fetch.zig
index 14401cf5a..6b085a0df 100644
--- a/src/server/cdp/domains/fetch.zig
+++ b/src/server/cdp/domains/fetch.zig
@@ -25,6 +25,7 @@ const HttpClient = @import("../../../network/HttpClient.zig");
const id = @import("../id.zig");
const CDP = @import("../CDP.zig");
+const SafeString = @import("../SafeString.zig");
const network = @import("network.zig");
@@ -186,8 +187,8 @@ pub fn responseIntercept(arena: Allocator, bc: *CDP.BrowserContext, intercept: *
.request = network.RequestWriter.init(arena, transfer, bc.network_limits.post_data),
.resourceType = fetchResourceType(transfer.req.resource_type),
.responseStatusCode = head.status,
- .responseStatusText = transfer.statusText() orelse "",
- .responseHeaders = transfer.responseHeaders(),
+ .responseStatusText = SafeString.wrap(transfer.statusText() orelse ""),
+ .responseHeaders = HeaderEntriesWriter{ .headers = transfer.responseHeaders() },
.networkId = &id.toRequestId(transfer), // matches the Network REQ-ID
}, .{ .session_id = session_id });
},
@@ -204,6 +205,25 @@ pub fn responseIntercept(arena: Allocator, bc: *CDP.BrowserContext, intercept: *
intercept.wait_for_interception.* = true;
}
+// A Fetch.HeaderEntry array. Names and values are the server's octets, which
+// need not be UTF-8: SafeString keeps them strings.
+const HeaderEntriesWriter = struct {
+ headers: []const http.Header,
+
+ pub fn jsonStringify(self: *const HeaderEntriesWriter, jws: anytype) !void {
+ try jws.beginArray();
+ for (self.headers) |hdr| {
+ try jws.beginObject();
+ try jws.objectField("name");
+ try jws.write(SafeString.wrap(hdr.name));
+ try jws.objectField("value");
+ try jws.write(SafeString.wrap(hdr.value));
+ try jws.endObject();
+ }
+ try jws.endArray();
+ }
+};
+
// The transfer a response-stage intercept id holds, left in the intercept
// state. Null when the id is unknown, the transfer is gone, or the pause is
// not at the response stage.
@@ -1034,6 +1054,42 @@ test "cdp.Fetch: response stage pauses a final response and continueResponse del
try testing.expectEqual(0, bc.cdp.browser.http_client.intercepted);
}
+test "cdp.Fetch: response stage reports a non-UTF-8 status text and headers as strings" {
+ var ctx = try testing.context();
+ defer ctx.deinit();
+ const bc = try ctx.loadBrowserContext(.{ .id = "BID-RS", .session_id = "SID-RS" });
+ try enableResponseStage(&ctx);
+
+ var probe: ResponseStageProbe = .{};
+ defer probe.deinit();
+ try probe.start(bc, "http://127.0.0.1:9582/latin1_response");
+ try ResponseStageProbe.tickUntilPaused(bc, 1);
+
+ // Latin-1 is transcoded to UTF-8, not written as an array of numbers.
+ try ctx.expectSentEvent("Fetch.requestPaused", .{
+ .requestId = "INT-0000000001",
+ .responseStatusCode = 200,
+ .responseStatusText = "Caf\xC3\xA9",
+ }, .{ .session_id = "SID-RS" });
+ var latin1_value: ?[]const u8 = null;
+ var latin1_name = false;
+ for ((try sentPausedParams(&ctx, "INT-0000000001")).get("responseHeaders").?.array.items) |header| {
+ const name = header.object.get("name").?;
+ const value = header.object.get("value").?;
+ try testing.expect(name == .string and value == .string);
+ if (std.mem.eql(u8, name.string, "x-latin1")) latin1_value = value.string;
+ if (std.mem.eql(u8, name.string, "x-caf\xC3\xA9")) latin1_name = true;
+ }
+ try testing.expectEqualSlices(u8, "La\xC2\xA0Paz", latin1_value.?);
+ try testing.expect(latin1_name);
+
+ try ctx.processMessage(.{ .id = 2, .method = "Fetch.continueResponse", .params = .{ .requestId = "INT-0000000001" } });
+ try ctx.expectSentResult(null, .{ .id = 2 });
+ try probe.tickUntilSettled(bc);
+ try testing.expectEqual(null, probe.err);
+ try testing.expectEqualSlices(u8, "latin1", probe.body.items);
+}
+
test "cdp.Fetch: response stage failRequest fails the request without delivering it" {
var ctx = try testing.context();
defer ctx.deinit();
diff --git a/src/server/cdp/domains/network.zig b/src/server/cdp/domains/network.zig
index 4407d3986..830c602b1 100644
--- a/src/server/cdp/domains/network.zig
+++ b/src/server/cdp/domains/network.zig
@@ -697,7 +697,7 @@ const ResponseWriter = struct {
try jws.objectField("statusText");
// The server's own phrase, not a canonical one: Chrome reports the
// wire, empty included.
- try jws.write(transfer.statusText() orelse "");
+ try jws.write(SafeString.wrap(transfer.statusText() orelse ""));
}
{
@@ -1582,6 +1582,34 @@ test "cdp.Network: getResponseBody base64-encodes a non-UTF-8 body" {
try ctx.expectSentResult(.{ .body = "/2Fi/g==", .base64Encoded = true }, .{ .id = 2 });
}
+test "cdp.Network: responseReceived reports a non-UTF-8 status text and headers as strings" {
+ var ctx = try testing.context();
+ defer ctx.deinit();
+
+ const bc = try ctx.loadBrowserContext(.{ .id = "BID-L1", .session_id = "SID-L1" });
+
+ try ctx.processMessage(.{ .id = 1, .method = "Network.enable" });
+ try ctx.expectSentResult(null, .{ .id = 1 });
+
+ const url = "http://127.0.0.1:9582/latin1_response";
+ const page = try bc.session.createPage();
+ try page.navigate(url, .{});
+ try testing.waitForPage(bc);
+
+ // Latin-1 is transcoded to UTF-8, not written as an array of numbers.
+ try ctx.expectSentEvent("Network.responseReceived", .{
+ .response = .{
+ .url = url,
+ .status = 200,
+ .statusText = "Caf\xC3\xA9",
+ .headers = .{
+ .@"x-latin1" = "La\xC2\xA0Paz",
+ .@"x-caf\xC3\xA9" = "1",
+ },
+ },
+ }, .{ .session_id = "SID-L1" });
+}
+
test "cdp.Network: enable maxTotalBufferSize evicts oldest bodies first" {
var ctx = try testing.context();
defer ctx.deinit();
diff --git a/src/testing.zig b/src/testing.zig
index aef335dae..6631cb3f4 100644
--- a/src/testing.zig
+++ b/src/testing.zig
@@ -1163,6 +1163,18 @@ fn testHTTPHandler(req: *std.http.Server.Request) !void {
});
}
+ // Latin-1 bytes in the reason phrase, a header name and a header value.
+ if (std.mem.eql(u8, path, "/latin1_response")) {
+ return req.respond("latin1", .{
+ .reason = "Caf\xE9",
+ .extra_headers = &.{
+ .{ .name = "Content-Type", .value = "text/plain" },
+ .{ .name = "X-Latin1", .value = "La\xA0Paz" },
+ .{ .name = "X-Caf\xE9", .value = "1" },
+ },
+ });
+ }
+
// Latin-1 sheets for css/stylesheet_charset.html: undecoded, the 0xE9
// swallows the bytes after it and the rule that follows is lost.
if (std.mem.eql(u8, path, "/styles/latin1-header.css")) {