Lightpanda: the headless browser designed for AI and automation
`Download`: bring back `filename` parsing + inform about `Content-Type`
Halil Durak53bacde3fa7e
diff --git a/src/Notification.zig b/src/Notification.zigindex 063303868..3a6f0cd9f 100644--- a/src/Notification.zig+++ b/src/Notification.zig@@ -329,6 +329,8 @@ pub const DownloadWillBegin = struct {guid: []const u8,url: []const u8,suggested_filename: []const u8,+ // The response's Content-Type, if it had one.+ content_type: ?[]const u8,};// Emitted by Frame as a download is written to disk: once when it startsdiff --git a/src/browser/Download.zig b/src/browser/Download.zigindex ef2d59dc1..34e0a1618 100644--- a/src/browser/Download.zig+++ b/src/browser/Download.zig@@ -23,9 +23,12 @@ const Frame = @import("Frame.zig");const URL = @import("URL.zig");const HttpClient = @import("../network/HttpClient.zig");+const sys_url = @import("../sys/url.zig");const uuidv4 = @import("../id.zig").uuidv4;+const latin1ToUtf8 = @import("../string.zig").latin1ToUtf8;const log = lp.log;+const Allocator = std.mem.Allocator;const Download = @This();// UUIDv4.@@ -43,24 +46,35 @@ total: ?usize = null,frame: *Frame,/// Prefer `from*` variants to initialize.-fn init(frame: *Frame, url: []const u8, status: u16, filename: []const u8, maybe_total: ?usize) !Download {+/// `maybe_suggestion` is the caller's sanitized pick; `null` falls back to the `guid`.+fn init(frame: *Frame, transfer: *HttpClient.Transfer, maybe_suggestion: ?[]const u8) !Download {const session = frame._session;+ // The final URL, after any redirects.+ const url = transfer.req.url;// `guid` is the CDP "Global Unique Identifier" which helps us find// where to dispatch events on.var guid: [36]u8 = undefined;uuidv4(&guid);- const succeeded = isSuccessStatus(status);+ // The suggestion may point into the response headers, which don't+ // outlive the transfer. Content-Disposition filenames can be in a+ // legacy encoding (e.g. Shift_JIS) but must yield valid UTF-8.+ const name = maybe_suggestion orelse &guid;+ const filename = if (std.unicode.utf8ValidateSlice(name))+ try frame.page_arena.dupe(u8, name)+ else+ try latin1ToUtf8(frame.page_arena, name);++ const succeeded = isSuccessStatus(transfer.responseStatus() orelse 0);const maybe_file: ?std.Io.File = if (succeeded == false) null else switch (session.download_behavior) {- inline .allow, .allow_and_name => blk: {+ inline .allow, .allow_and_name => |behavior| blk: {const download_path = session.download_path orelse {log.warn(.frame, "download without downloadPath", .{ .url = url });return error.WithoutDownloadPath;};// allowAndName stores the file under its guid.- //const on_disk_name = if (comptime behavior == .allow_and_name) @as([]const u8, guid) else filename;- const on_disk_name = filename;+ const on_disk_name: []const u8 = if (comptime behavior == .allow_and_name) &guid else filename;const cwd = std.Io.Dir.cwd();try cwd.createDirPath(lp.io, download_path);@@ -79,6 +93,7 @@ fn init(frame: *Frame, url: []const u8, status: u16, filename: []const u8, maybe.guid = &guid,.url = url,.suggested_filename = filename,+ .content_type = transfer.contentType(),});}@@ -89,7 +104,7 @@ fn init(frame: *Frame, url: []const u8, status: u16, filename: []const u8, maybe.received = 0,// Like Chrome, a failed download reports 0/0 bytes, whatever the// response advertised.- .total = if (succeeded) maybe_total else null,+ .total = if (succeeded) transfer.getContentLength() else null,.frame = frame,};}@@ -101,15 +116,7 @@ pub fn fromTransfer(frame: *Frame, transfer: *HttpClient.Transfer) !Download {return error.Denied;}- // Find `Content-Disposition`.- const disposition: HttpClient.Header = blk: {- for (transfer.responseHeaders()) |header| {- if (std.mem.eql(u8, header.name, "content-disposition")) {- break :blk header;- }- }- return error.MissingHeader;- };+ const disposition = contentDisposition(transfer) orelse return error.MissingHeader;if (std.ascii.eqlIgnoreCase(disposition.firstValue(), "attachment") == false) {return error.MissingHeader;}@@ -120,9 +127,15 @@ pub fn fromTransfer(frame: *Frame, transfer: *HttpClient.Transfer) !Download {return error.UnexpectedHttpStatus;}- // TODO: Find filename correctly.- const filename = "dummy.txt";- return init(frame, frame.url, status, filename, transfer.getContentLength());+ const maybe_suggestion = dispositionFilename(disposition) orelse (try urlBasename(frame.page_arena, frame.url));+ return init(frame, transfer, maybe_suggestion) catch |err| switch (err) {+ error.OutOfMemory => return err,+ // As before, a download we can't write renders the response instead.+ else => {+ log.err(.frame, "download init", .{ .err = err, .url = frame.url });+ return error.Unwritable;+ },+ };}/// Try to initialize an asynchronous `Download` from a href. In order to@@ -139,8 +152,9 @@ pub fn fromHref(frame: *Frame, href: []const u8, filename_hint: []const u8) !voiconst download = try frame._factory.create(Download{// Let `init` initialize this at `headerCallback`..guid = undefined,- // Pass-in hint, `headerCallback` will change this, though.- .filename = filename_hint,+ // Pass-in hint, `headerCallback` will change this, though. Duped,+ // since the attribute can change before the response arrives.+ .filename = try frame.page_arena.dupe(u8, filename_hint),.frame = frame,});// `transfer.deinit()` fires no callbacks, so this is still ours on error.@@ -227,9 +241,20 @@ fn headerCallback(transfer: *HttpClient.Transfer) !HttpClient.Transfer.HeaderRes// The final URL, after any redirects.const final_url = transfer.req.url;+ // setDownloadBehavior can change while the fetch is in-flight.+ if (frame._session.download_behavior == .deny) {+ log.info(.frame, "download denied", .{ .url = final_url });+ return .abort;+ }+const status = transfer.responseStatus() orelse return .abort;- download.* = init(frame, final_url, status, "dummy.txt", transfer.getContentLength()) catch |err| {+ // A Content-Disposition filename wins over the `download` attribute.+ const disposition_name = if (contentDisposition(transfer)) |disposition| dispositionFilename(disposition) else null;+ const maybe_suggestion = disposition_name orelse+ sanitizeFilename(download.filename) orelse (try urlBasename(frame.page_arena, final_url));++ download.* = init(frame, transfer, maybe_suggestion) catch |err| {log.err(.frame, "download init", .{ .err = err });return .abort;};@@ -299,3 +324,93 @@ fn fromRaw(ptr: *anyopaque) struct { *Download, bool } {inline fn isSuccessStatus(status: u16) bool {return status >= 200 and status <= 203;}++fn contentDisposition(transfer: *HttpClient.Transfer) ?HttpClient.Header {+ for (transfer.responseHeaders()) |header| {+ if (std.mem.eql(u8, header.name, "content-disposition")) {+ return header;+ }+ }+ return null;+}++// Extracts the filename from a Content-Disposition header, handling the quoted,+// unquoted, and RFC 5987 (filename*=charset''value) forms. Path components are+// stripped so the result is always a bare basename.+fn dispositionFilename(disposition: HttpClient.Header) ?[]const u8 {+ // Prefer the extended filename*= form when present, per RFC 6266.+ if (disposition.param("filename*")) |ext| {+ // charset'lang'value — take everything after the second quote.+ if (std.mem.findScalar(u8, ext, '\'')) |first| {+ if (std.mem.findScalarPos(u8, ext, first + 1, '\'')) |second| {+ return sanitizeFilename(ext[second + 1 ..]);+ }+ }+ return sanitizeFilename(ext);+ }+ if (disposition.param("filename")) |name| {+ return sanitizeFilename(name);+ }+ return null;+}++// Strips any directory components, guarding against path traversal. Content-+// Disposition can carry Windows separators, so backslashes are stripped too,+// regardless of the host platform.+fn sanitizeFilename(name: []const u8) ?[]const u8 {+ var out = std.fs.path.basename(name);+ if (std.mem.findScalarLast(u8, out, '\\')) |i| {+ out = out[i + 1 ..];+ }+ if (out.len == 0 or std.mem.eql(u8, out, ".") or std.mem.eql(u8, out, "..")) {+ return null;+ }+ return out;+}++// Derives a filename from a URL's last path segment. The path is taken from a+// real URL parse (rust-url) so query/fragment and percent-encoding are handled+// the same way the rest of the browser handles URLs. The result is duped into+// `arena`, since the parsed URL is freed before this returns.+fn urlBasename(arena: Allocator, url: []const u8) !?[]const u8 {+ var err: i32 = 0;+ const u = sys_url.url_parse(url.ptr, url.len, &err) orelse return null;+ defer sys_url.url_free(u);++ var ptr: [*]const u8 = undefined;+ var len: usize = undefined;+ sys_url.url_get_path(u, &ptr, &len);++ const name = sanitizeFilename(ptr[0..len]) orelse return null;+ return try arena.dupe(u8, name);+}++const testing = @import("../testing.zig");++fn dispositionHeader(value: []const u8) HttpClient.Header {+ return .{ .name = "content-disposition", .value = value };+}++test "Download: dispositionFilename" {+ try testing.expectEqualSlices(u8, "report.csv", dispositionFilename(dispositionHeader("attachment; filename=\"report.csv\"")).?);+ try testing.expectEqualSlices(u8, "report.csv", dispositionFilename(dispositionHeader("attachment; filename=report.csv")).?);+ try testing.expectEqualSlices(u8, "r e.csv", dispositionFilename(dispositionHeader("attachment; filename=\"r e.csv\"")).?);+ // RFC 5987 extended form is preferred over the plain filename when present+ // (the value is taken verbatim after the charset'lang' prefix).+ try testing.expectEqualSlices(u8, "extended.txt", dispositionFilename(dispositionHeader("attachment; filename=\"fallback.txt\"; filename*=UTF-8''extended.txt")).?);+ try testing.expect(dispositionFilename(dispositionHeader("attachment")) == null);+ // Path components are stripped to guard against traversal.+ try testing.expectEqualSlices(u8, "evil.sh", dispositionFilename(dispositionHeader("attachment; filename=\"../../evil.sh\"")).?);+ try testing.expectEqualSlices(u8, "evil.sh", dispositionFilename(dispositionHeader("attachment; filename=\"..\\..\\evil.sh\"")).?);+ try testing.expect(dispositionFilename(dispositionHeader("attachment; filename=\"..\"")) == null);+}++test "Download: urlBasename" {+ var arena = std.heap.ArenaAllocator.init(testing.allocator);+ defer arena.deinit();+ const a = arena.allocator();++ try testing.expectEqualSlices(u8, "report.csv", (try urlBasename(a, "http://x.com/a/b/report.csv")).?);+ try testing.expectEqualSlices(u8, "report.csv", (try urlBasename(a, "http://x.com/report.csv?v=1#x")).?);+ try testing.expect((try urlBasename(a, "http://x.com/")) == null);+}