| ... | @@ -316,56 +316,51 @@ pub fn fetchAndAddDependencies( | ... | @@ -316,56 +316,51 @@ pub fn fetchAndAddDependencies( |
| 316 | for (manifest.dependencies.keys(), 0..) |name, i| { | 316 | for (manifest.dependencies.keys(), 0..) |name, i| { |
| 317 | const dep = deps_list[i]; | 317 | const dep = deps_list[i]; |
| 318 | | 318 | |
| 319 | const sub = try fetchAndUnpack( | 319 | const sub_pkg = try getCachedPackage( |
| 320 | thread_pool, | 320 | http_client.allocator, |
| 321 | http_client, | | |
| 322 | global_cache_directory, | 321 | global_cache_directory, |
| 323 | dep, | 322 | dep, |
| 324 | report, | 323 | report, |
| 325 | all_modules, | 324 | all_modules, |
| 326 | root_prog_node, | 325 | root_prog_node, |
| 327 | name, | 326 | ) orelse m: { |
| 328 | ); | 327 | const mod = try fetchAndUnpack( |
| | 328 | thread_pool, |
| | 329 | http_client, |
| | 330 | directory, |
| | 331 | global_cache_directory, |
| | 332 | dep, |
| | 333 | report, |
| | 334 | all_modules, |
| | 335 | root_prog_node, |
| | 336 | name, |
| | 337 | ); |
| 329 | | 338 | |
| 330 | if (sub.mod) |mod| { | 339 | try mod.fetchAndAddDependencies( |
| 331 | if (!sub.found_existing) { | 340 | deps_pkg, |
| 332 | try mod.fetchAndAddDependencies( | 341 | arena, |
| 333 | deps_pkg, | 342 | thread_pool, |
| 334 | arena, | 343 | http_client, |
| 335 | thread_pool, | 344 | mod.root_src_directory, |
| 336 | http_client, | 345 | global_cache_directory, |
| 337 | mod.root_src_directory, | 346 | local_cache_directory, |
| 338 | global_cache_directory, | 347 | dependencies_source, |
| 339 | local_cache_directory, | 348 | error_bundle, |
| 340 | dependencies_source, | 349 | all_modules, |
| 341 | error_bundle, | 350 | root_prog_node, |
| 342 | all_modules, | 351 | dep.hash.?, |
| 343 | root_prog_node, | 352 | ); |
| 344 | dep.hash.?, | | |
| 345 | ); | | |
| 346 | } | | |
| 347 | | 353 | |
| 348 | try pkg.add(gpa, name, mod); | 354 | break :m mod; |
| 349 | if (deps_pkg.table.get(dep.hash.?)) |other_sub| { | 355 | }; |
| 350 | // This should be the same package (and hence module) since it's the same hash | 356 | |
| 351 | // TODO: dedup multiple versions of the same package | 357 | try pkg.add(gpa, name, sub_pkg); |
| 352 | assert(other_sub == mod); | 358 | if (deps_pkg.table.get(dep.hash.?)) |other_sub| { |
| 353 | } else { | 359 | // This should be the same package (and hence module) since it's the same hash |
| 354 | try deps_pkg.add(gpa, dep.hash.?, mod); | 360 | // TODO: dedup multiple versions of the same package |
| 355 | } | 361 | assert(other_sub == sub_pkg); |
| 356 | } else if (!sub.found_existing) { | 362 | } else { |
| 357 | const pkg_dir_sub_path = "p" ++ fs.path.sep_str ++ (dep.hash.?)[0..hex_multihash_len]; | 363 | try deps_pkg.add(gpa, dep.hash.?, sub_pkg); |
| 358 | const build_root = try global_cache_directory.join(arena, &.{pkg_dir_sub_path}); | | |
| 359 | try dependencies_source.writer().print( | | |
| 360 | \\ pub const {} = struct {{ | | |
| 361 | \\ pub const build_root = "{}"; | | |
| 362 | \\ pub const deps: []const struct {{ []const u8, []const u8 }} = &.{{}}; | | |
| 363 | \\ }}; | | |
| 364 | \\ | | |
| 365 | , .{ | | |
| 366 | std.zig.fmtId(dep.hash.?), | | |
| 367 | std.zig.fmtEscapes(build_root), | | |
| 368 | }); | | |
| 369 | } | 364 | } |
| 370 | } | 365 | } |
| 371 | | 366 | |
| ... | @@ -490,6 +485,316 @@ const Report = struct { | ... | @@ -490,6 +485,316 @@ const Report = struct { |
| 490 | } | 485 | } |
| 491 | }; | 486 | }; |
| 492 | | 487 | |
| | 488 | const FetchLocation = union(SourceType) { |
| | 489 | /// The absolute path to a file or directory. |
| | 490 | /// This may be a file that requires unpacking (such as a .tar.gz), |
| | 491 | /// or the path to the root directory of a package. |
| | 492 | file: []const u8, |
| | 493 | http_request: std.Uri, |
| | 494 | |
| | 495 | pub fn init(gpa: Allocator, uri: std.Uri, directory: Compilation.Directory, dep: Manifest.Dependency, report: Report) !FetchLocation { |
| | 496 | const source_type = getPackageSourceType(uri) catch |
| | 497 | return report.fail(dep.location_tok, "Unknown scheme: {s}", .{uri.scheme}); |
| | 498 | |
| | 499 | return switch (source_type) { |
| | 500 | .file => f: { |
| | 501 | const path = if (builtin.os.tag == .windows) p: { |
| | 502 | var uri_str = std.ArrayList(u8).init(gpa); |
| | 503 | defer uri_str.deinit(); |
| | 504 | try uri.format("+/", .{}, uri_str.writer()); |
| | 505 | const uri_str_z = try gpa.dupeZ(u8, uri_str.items); |
| | 506 | defer gpa.free(uri_str_z); |
| | 507 | |
| | 508 | var buf: [std.os.windows.MAX_PATH:0]u8 = undefined; |
| | 509 | var buf_len: std.os.windows.DWORD = std.os.windows.MAX_PATH; |
| | 510 | const result = std.os.windows.shlwapi.PathCreateFromUrlA(uri_str_z, &buf, &buf_len, 0); |
| | 511 | |
| | 512 | if (result != std.os.windows.S_OK) return report.fail(dep.location_tok, "Invalid URI", .{}); |
| | 513 | |
| | 514 | break :p try gpa.dupe(u8, buf[0..buf_len]); |
| | 515 | } else try std.Uri.unescapeString(gpa, uri.path); |
| | 516 | defer gpa.free(path); |
| | 517 | |
| | 518 | const new_path = try fs.path.resolve(gpa, &.{ directory.path.?, path }); |
| | 519 | |
| | 520 | break :f .{ .file = new_path }; |
| | 521 | }, |
| | 522 | .http_request => r: { |
| | 523 | break :r .{ .http_request = uri }; |
| | 524 | }, |
| | 525 | }; |
| | 526 | } |
| | 527 | |
| | 528 | pub fn deinit(f: *FetchLocation, gpa: Allocator) void { |
| | 529 | switch (f.*) { |
| | 530 | .file => |path| gpa.free(path), |
| | 531 | .http_request => {}, |
| | 532 | } |
| | 533 | f.* = undefined; |
| | 534 | } |
| | 535 | |
| | 536 | const SourceType = enum { |
| | 537 | file, |
| | 538 | http_request, |
| | 539 | }; |
| | 540 | |
| | 541 | fn getPackageSourceType(uri: std.Uri) error{UnknownScheme}!SourceType { |
| | 542 | const package_source_map = std.ComptimeStringMap( |
| | 543 | SourceType, |
| | 544 | .{ |
| | 545 | .{ "file", .file }, |
| | 546 | .{ "http", .http_request }, |
| | 547 | .{ "https", .http_request }, |
| | 548 | }, |
| | 549 | ); |
| | 550 | return package_source_map.get(uri.scheme) orelse error.UnknownScheme; |
| | 551 | } |
| | 552 | |
| | 553 | pub fn isDirectory(path: []const u8, root_dir: Compilation.Directory) !bool { |
| | 554 | return if (mem.endsWith(u8, path, std.fs.path.sep_str)) |
| | 555 | true |
| | 556 | else if (std.fs.path.extension(path).len > 0) |
| | 557 | false |
| | 558 | else d: { |
| | 559 | // It's common to write directories without a trailing '/'. |
| | 560 | // This is some special casing logic to detect directories if |
| | 561 | // the file type cannot be determined from the extension. |
| | 562 | var dir = root_dir.handle.openDir(path, .{}) catch |err| switch (err) { |
| | 563 | error.NotDir => break :d false, |
| | 564 | else => break :d err, |
| | 565 | }; |
| | 566 | defer dir.close(); |
| | 567 | break :d true; |
| | 568 | }; |
| | 569 | } |
| | 570 | |
| | 571 | pub fn fetch( |
| | 572 | f: FetchLocation, |
| | 573 | gpa: Allocator, |
| | 574 | root_dir: Compilation.Directory, |
| | 575 | http_client: *std.http.Client, |
| | 576 | dep: Manifest.Dependency, |
| | 577 | report: Report, |
| | 578 | ) !ReadableResource { |
| | 579 | switch (f) { |
| | 580 | .file => |file| { |
| | 581 | const is_dir = isDirectory(file, root_dir) catch |
| | 582 | return report.fail(dep.location_tok, "File not found: {s}", .{file}); |
| | 583 | |
| | 584 | return if (is_dir) |
| | 585 | .{ |
| | 586 | .path = try gpa.dupe(u8, file), |
| | 587 | .resource = .{ .directory = try fs.openIterableDirAbsolute(file, .{}) }, |
| | 588 | } |
| | 589 | else |
| | 590 | .{ |
| | 591 | .path = try gpa.dupe(u8, file), |
| | 592 | .resource = .{ .file = try fs.openFileAbsolute(file, .{}) }, |
| | 593 | }; |
| | 594 | }, |
| | 595 | .http_request => |uri| { |
| | 596 | var h = std.http.Headers{ .allocator = gpa }; |
| | 597 | defer h.deinit(); |
| | 598 | |
| | 599 | var req = try http_client.request(.GET, uri, h, .{}); |
| | 600 | |
| | 601 | try req.start(.{}); |
| | 602 | try req.wait(); |
| | 603 | |
| | 604 | if (req.response.status != .ok) { |
| | 605 | return report.fail(dep.location_tok, "Expected response status '200 OK' got '{} {s}'", .{ |
| | 606 | @intFromEnum(req.response.status), |
| | 607 | req.response.status.phrase() orelse "", |
| | 608 | }); |
| | 609 | } |
| | 610 | |
| | 611 | return .{ |
| | 612 | .path = try gpa.dupe(u8, uri.path), |
| | 613 | .resource = .{ .http_request = req }, |
| | 614 | }; |
| | 615 | }, |
| | 616 | } |
| | 617 | } |
| | 618 | }; |
| | 619 | |
| | 620 | const ReadableResource = struct { |
| | 621 | path: []const u8, |
| | 622 | resource: union(enum) { |
| | 623 | file: fs.File, |
| | 624 | directory: fs.IterableDir, |
| | 625 | http_request: std.http.Client.Request, |
| | 626 | }, |
| | 627 | |
| | 628 | /// Unpack the package into the global cache directory. |
| | 629 | /// If `ps` does not require unpacking (for example, if it is a directory), then no caching is performed. |
| | 630 | /// In either case, the hash is computed and returned along with the path to the package. |
| | 631 | pub fn unpack( |
| | 632 | rr: *ReadableResource, |
| | 633 | allocator: Allocator, |
| | 634 | thread_pool: *ThreadPool, |
| | 635 | global_cache_directory: Compilation.Directory, |
| | 636 | dep: Manifest.Dependency, |
| | 637 | report: Report, |
| | 638 | pkg_prog_node: *std.Progress.Node, |
| | 639 | ) !PackageLocation { |
| | 640 | switch (rr.resource) { |
| | 641 | .directory => |dir| { |
| | 642 | const actual_hash = try computePackageHash(thread_pool, dir); |
| | 643 | return .{ |
| | 644 | .hash = actual_hash, |
| | 645 | .dir_path = try allocator.dupe(u8, rr.path), |
| | 646 | }; |
| | 647 | }, |
| | 648 | inline .file, .http_request => |*r| { |
| | 649 | const s = fs.path.sep_str; |
| | 650 | const rand_int = std.crypto.random.int(u64); |
| | 651 | const tmp_dir_sub_path = "tmp" ++ s ++ Manifest.hex64(rand_int); |
| | 652 | |
| | 653 | const actual_hash = h: { |
| | 654 | var tmp_directory: Compilation.Directory = d: { |
| | 655 | const path = try global_cache_directory.join(allocator, &.{tmp_dir_sub_path}); |
| | 656 | errdefer allocator.free(path); |
| | 657 | |
| | 658 | const iterable_dir = try global_cache_directory.handle.makeOpenPathIterable(tmp_dir_sub_path, .{}); |
| | 659 | errdefer iterable_dir.close(); |
| | 660 | |
| | 661 | break :d .{ |
| | 662 | .path = path, |
| | 663 | .handle = iterable_dir.dir, |
| | 664 | }; |
| | 665 | }; |
| | 666 | defer tmp_directory.closeAndFree(allocator); |
| | 667 | |
| | 668 | const opt_content_length = try rr.getSize(); |
| | 669 | |
| | 670 | var prog_reader: ProgressReader(@TypeOf(r.reader())) = .{ |
| | 671 | .child_reader = r.reader(), |
| | 672 | .prog_node = pkg_prog_node, |
| | 673 | .unit = if (opt_content_length) |content_length| unit: { |
| | 674 | const kib = content_length / 1024; |
| | 675 | const mib = kib / 1024; |
| | 676 | if (mib > 0) { |
| | 677 | pkg_prog_node.setEstimatedTotalItems(@intCast(mib)); |
| | 678 | pkg_prog_node.setUnit("MiB"); |
| | 679 | break :unit .mib; |
| | 680 | } else { |
| | 681 | pkg_prog_node.setEstimatedTotalItems(@intCast(@max(1, kib))); |
| | 682 | pkg_prog_node.setUnit("KiB"); |
| | 683 | break :unit .kib; |
| | 684 | } |
| | 685 | } else .any, |
| | 686 | }; |
| | 687 | pkg_prog_node.context.refresh(); |
| | 688 | |
| | 689 | switch (try rr.getFileType(dep, report)) { |
| | 690 | .@"tar.gz" => try unpackTarball(allocator, prog_reader, tmp_directory.handle, std.compress.gzip), |
| | 691 | // I have not checked what buffer sizes the xz decompression implementation uses |
| | 692 | // by default, so the same logic applies for buffering the reader as for gzip. |
| | 693 | .@"tar.xz" => try unpackTarball(allocator, prog_reader, tmp_directory.handle, std.compress.xz), |
| | 694 | } |
| | 695 | |
| | 696 | // Unpack completed - stop showing amount as progress |
| | 697 | pkg_prog_node.setEstimatedTotalItems(0); |
| | 698 | pkg_prog_node.setCompletedItems(0); |
| | 699 | pkg_prog_node.context.refresh(); |
| | 700 | |
| | 701 | // TODO: delete files not included in the package prior to computing the package hash. |
| | 702 | // for example, if the ini file has directives to include/not include certain files, |
| | 703 | // apply those rules directly to the filesystem right here. This ensures that files |
| | 704 | // not protected by the hash are not present on the file system. |
| | 705 | |
| | 706 | break :h try computePackageHash(thread_pool, .{ .dir = tmp_directory.handle }); |
| | 707 | }; |
| | 708 | |
| | 709 | const pkg_dir_sub_path = "p" ++ s ++ Manifest.hexDigest(actual_hash); |
| | 710 | const unpacked_path = try global_cache_directory.join(allocator, &.{pkg_dir_sub_path}); |
| | 711 | errdefer allocator.free(unpacked_path); |
| | 712 | |
| | 713 | const relative_unpacked_path = try fs.path.relative(allocator, global_cache_directory.path.?, unpacked_path); |
| | 714 | defer allocator.free(relative_unpacked_path); |
| | 715 | try renameTmpIntoCache(global_cache_directory.handle, tmp_dir_sub_path, relative_unpacked_path); |
| | 716 | |
| | 717 | return .{ |
| | 718 | .hash = actual_hash, |
| | 719 | .dir_path = unpacked_path, |
| | 720 | }; |
| | 721 | }, |
| | 722 | } |
| | 723 | } |
| | 724 | |
| | 725 | const FileType = enum { |
| | 726 | @"tar.gz", |
| | 727 | @"tar.xz", |
| | 728 | }; |
| | 729 | |
| | 730 | pub fn getSize(rr: ReadableResource) !?u64 { |
| | 731 | switch (rr.resource) { |
| | 732 | // TODO: Handle case of chunked content-length |
| | 733 | .http_request => |req| return req.response.content_length, |
| | 734 | .file => |f| return (try f.metadata()).size(), |
| | 735 | .directory => unreachable, |
| | 736 | } |
| | 737 | } |
| | 738 | |
| | 739 | pub fn getFileType(rr: ReadableResource, dep: Manifest.Dependency, report: Report) !FileType { |
| | 740 | switch (rr.resource) { |
| | 741 | .file => { |
| | 742 | return if (mem.endsWith(u8, rr.path, ".tar.gz")) |
| | 743 | .@"tar.gz" |
| | 744 | else if (mem.endsWith(u8, rr.path, ".tar.xz")) |
| | 745 | .@"tar.xz" |
| | 746 | else |
| | 747 | return report.fail(dep.location_tok, "Unknown file type", .{}); |
| | 748 | }, |
| | 749 | .directory => return error.IsDir, |
| | 750 | .http_request => |req| { |
| | 751 | const content_type = req.response.headers.getFirstValue("Content-Type") orelse |
| | 752 | return report.fail(dep.location_tok, "Missing 'Content-Type' header", .{}); |
| | 753 | |
| | 754 | // If the response has a different content type than the URI indicates, override |
| | 755 | // the previously assumed file type. |
| | 756 | return if (ascii.eqlIgnoreCase(content_type, "application/gzip") or |
| | 757 | ascii.eqlIgnoreCase(content_type, "application/x-gzip") or |
| | 758 | ascii.eqlIgnoreCase(content_type, "application/tar+gzip")) |
| | 759 | .@"tar.gz" |
| | 760 | else if (ascii.eqlIgnoreCase(content_type, "application/x-xz")) |
| | 761 | .@"tar.xz" |
| | 762 | else if (ascii.eqlIgnoreCase(content_type, "application/octet-stream")) ty: { |
| | 763 | // support gitlab tarball urls such as https://gitlab.com/<namespace>/<project>/-/archive/<sha>/<project>-<sha>.tar.gz |
| | 764 | // whose content-disposition header is: 'attachment; filename="<project>-<sha>.tar.gz"' |
| | 765 | const content_disposition = req.response.headers.getFirstValue("Content-Disposition") orelse |
| | 766 | return report.fail(dep.location_tok, "Missing 'Content-Disposition' header for Content-Type=application/octet-stream", .{}); |
| | 767 | if (mem.startsWith(u8, content_disposition, "attachment;") and |
| | 768 | mem.endsWith(u8, content_disposition, ".tar.gz\"")) |
| | 769 | { |
| | 770 | break :ty .@"tar.gz"; |
| | 771 | } else return report.fail(dep.location_tok, "Unsupported 'Content-Disposition' header value: '{s}' for Content-Type=application/octet-stream", .{content_disposition}); |
| | 772 | } else return report.fail(dep.location_tok, "Unrecognized value for 'Content-Type' header: {s}", .{content_type}); |
| | 773 | }, |
| | 774 | } |
| | 775 | } |
| | 776 | |
| | 777 | pub fn deinit(rr: *ReadableResource, gpa: Allocator) void { |
| | 778 | gpa.free(rr.path); |
| | 779 | switch (rr.resource) { |
| | 780 | .file => |file| file.close(), |
| | 781 | .directory => |*dir| dir.close(), |
| | 782 | .http_request => |*req| req.deinit(), |
| | 783 | } |
| | 784 | rr.* = undefined; |
| | 785 | } |
| | 786 | }; |
| | 787 | |
| | 788 | pub const PackageLocation = struct { |
| | 789 | hash: [Manifest.Hash.digest_length]u8, |
| | 790 | dir_path: []const u8, |
| | 791 | |
| | 792 | pub fn deinit(pl: *PackageLocation, allocator: Allocator) void { |
| | 793 | allocator.free(pl.dir_path); |
| | 794 | pl.* = undefined; |
| | 795 | } |
| | 796 | }; |
| | 797 | |
| 493 | const hex_multihash_len = 2 * Manifest.multihash_len; | 798 | const hex_multihash_len = 2 * Manifest.multihash_len; |
| 494 | const MultiHashHexDigest = [hex_multihash_len]u8; | 799 | const MultiHashHexDigest = [hex_multihash_len]u8; |
| 495 | /// This is to avoid creating multiple modules for the same build.zig file. | 800 | /// This is to avoid creating multiple modules for the same build.zig file. |
| ... | @@ -542,29 +847,24 @@ fn ProgressReader(comptime ReaderType: type) type { | ... | @@ -542,29 +847,24 @@ fn ProgressReader(comptime ReaderType: type) type { |
| 542 | }; | 847 | }; |
| 543 | } | 848 | } |
| 544 | | 849 | |
| 545 | fn fetchAndUnpack( | 850 | fn getCachedPackage( |
| 546 | thread_pool: *ThreadPool, | 851 | gpa: Allocator, |
| 547 | http_client: *std.http.Client, | | |
| 548 | global_cache_directory: Compilation.Directory, | 852 | global_cache_directory: Compilation.Directory, |
| 549 | dep: Manifest.Dependency, | 853 | dep: Manifest.Dependency, |
| 550 | report: Report, | 854 | report: Report, |
| 551 | all_modules: *AllModules, | 855 | all_modules: *AllModules, |
| 552 | root_prog_node: *std.Progress.Node, | 856 | root_prog_node: *std.Progress.Node, |
| 553 | /// This does not have to be any form of canonical or fully-qualified name: it | 857 | ) !?*Package { |
| 554 | /// is only intended to be human-readable for progress reporting. | 858 | _ = report; |
| 555 | name_for_prog: []const u8, | | |
| 556 | ) !struct { mod: ?*Package, found_existing: bool } { | | |
| 557 | const gpa = http_client.allocator; | | |
| 558 | const s = fs.path.sep_str; | 859 | const s = fs.path.sep_str; |
| 559 | | | |
| 560 | // Check if the expected_hash is already present in the global package | 860 | // Check if the expected_hash is already present in the global package |
| 561 | // cache, and thereby avoid both fetching and unpacking. | 861 | // cache, and thereby avoid both fetching and unpacking. |
| 562 | if (dep.hash) |h| cached: { | 862 | if (dep.hash) |h| { |
| 563 | const hex_digest = h[0..hex_multihash_len]; | 863 | const hex_digest = h[0..hex_multihash_len]; |
| 564 | const pkg_dir_sub_path = "p" ++ s ++ hex_digest; | 864 | const pkg_dir_sub_path = "p" ++ s ++ hex_digest; |
| 565 | | 865 | |
| 566 | var pkg_dir = global_cache_directory.handle.openDir(pkg_dir_sub_path, .{}) catch |err| switch (err) { | 866 | var pkg_dir = global_cache_directory.handle.openDir(pkg_dir_sub_path, .{}) catch |err| switch (err) { |
| 567 | error.FileNotFound => break :cached, | 867 | error.FileNotFound => return null, |
| 568 | else => |e| return e, | 868 | else => |e| return e, |
| 569 | }; | 869 | }; |
| 570 | errdefer pkg_dir.close(); | 870 | errdefer pkg_dir.close(); |
| ... | @@ -574,16 +874,7 @@ fn fetchAndUnpack( | ... | @@ -574,16 +874,7 @@ fn fetchAndUnpack( |
| 574 | const gop = try all_modules.getOrPut(gpa, hex_digest.*); | 874 | const gop = try all_modules.getOrPut(gpa, hex_digest.*); |
| 575 | if (gop.found_existing) { | 875 | if (gop.found_existing) { |
| 576 | if (gop.value_ptr.*) |mod| { | 876 | if (gop.value_ptr.*) |mod| { |
| 577 | return switch (mod) { | 877 | return mod; |
| 578 | .zig_pkg => |pkg| .{ | | |
| 579 | .mod = pkg, | | |
| 580 | .found_existing = true, | | |
| 581 | }, | | |
| 582 | .non_zig_pkg => .{ | | |
| 583 | .mod = null, | | |
| 584 | .found_existing = true, | | |
| 585 | }, | | |
| 586 | }; | | |
| 587 | } | 878 | } |
| 588 | } | 879 | } |
| 589 | | 880 | |
| ... | @@ -615,121 +906,60 @@ fn fetchAndUnpack( | ... | @@ -615,121 +906,60 @@ fn fetchAndUnpack( |
| 615 | .root_src_path = owned_src_path, | 906 | .root_src_path = owned_src_path, |
| 616 | }; | 907 | }; |
| 617 | | 908 | |
| 618 | gop.value_ptr.* = .{ .zig_pkg = ptr }; | 909 | gop.value_ptr.* = ptr; |
| 619 | return .{ | 910 | return ptr; |
| 620 | .mod = ptr, | | |
| 621 | .found_existing = false, | | |
| 622 | }; | | |
| 623 | } | 911 | } |
| 624 | | 912 | |
| | 913 | return null; |
| | 914 | } |
| | 915 | |
| | 916 | fn fetchAndUnpack( |
| | 917 | thread_pool: *ThreadPool, |
| | 918 | http_client: *std.http.Client, |
| | 919 | directory: Compilation.Directory, |
| | 920 | global_cache_directory: Compilation.Directory, |
| | 921 | dep: Manifest.Dependency, |
| | 922 | report: Report, |
| | 923 | all_modules: *AllModules, |
| | 924 | root_prog_node: *std.Progress.Node, |
| | 925 | /// This does not have to be any form of canonical or fully-qualified name: it |
| | 926 | /// is only intended to be human-readable for progress reporting. |
| | 927 | name_for_prog: []const u8, |
| | 928 | ) !*Package { |
| | 929 | const gpa = http_client.allocator; |
| | 930 | |
| 625 | var pkg_prog_node = root_prog_node.start(name_for_prog, 0); | 931 | var pkg_prog_node = root_prog_node.start(name_for_prog, 0); |
| 626 | defer pkg_prog_node.end(); | 932 | defer pkg_prog_node.end(); |
| 627 | pkg_prog_node.activate(); | 933 | pkg_prog_node.activate(); |
| 628 | pkg_prog_node.context.refresh(); | 934 | pkg_prog_node.context.refresh(); |
| 629 | | 935 | |
| 630 | const uri = try std.Uri.parse(dep.url); | 936 | const uri = switch (dep.location) { |
| 631 | | 937 | .url => |url| std.Uri.parse(url) catch |err| switch (err) { |
| 632 | const rand_int = std.crypto.random.int(u64); | 938 | error.UnexpectedCharacter => return report.fail(dep.location_tok, "failed to parse dependency location as URI.", .{}), |
| 633 | const tmp_dir_sub_path = "tmp" ++ s ++ Manifest.hex64(rand_int); | 939 | else => return err, |
| 634 | | 940 | }, |
| 635 | const actual_hash = a: { | 941 | .path => |path| std.Uri{ |
| 636 | var tmp_directory: Compilation.Directory = d: { | 942 | .scheme = "file", |
| 637 | const path = try global_cache_directory.join(gpa, &.{tmp_dir_sub_path}); | 943 | .user = null, |
| 638 | errdefer gpa.free(path); | 944 | .password = null, |
| 639 | | 945 | .host = null, |
| 640 | const iterable_dir = try global_cache_directory.handle.makeOpenPathIterable(tmp_dir_sub_path, .{}); | 946 | .port = null, |
| 641 | errdefer iterable_dir.close(); | 947 | .path = path, |
| 642 | | 948 | .query = null, |
| 643 | break :d .{ | 949 | .fragment = null, |
| 644 | .path = path, | 950 | }, |
| 645 | .handle = iterable_dir.dir, | 951 | }; |
| 646 | }; | | |
| 647 | }; | | |
| 648 | defer tmp_directory.closeAndFree(gpa); | | |
| 649 | | | |
| 650 | var h = std.http.Headers{ .allocator = gpa }; | | |
| 651 | defer h.deinit(); | | |
| 652 | | | |
| 653 | var req = try http_client.request(.GET, uri, h, .{}); | | |
| 654 | defer req.deinit(); | | |
| 655 | | | |
| 656 | try req.start(.{}); | | |
| 657 | try req.wait(); | | |
| 658 | | | |
| 659 | if (req.response.status != .ok) { | | |
| 660 | return report.fail(dep.url_tok, "Expected response status '200 OK' got '{} {s}'", .{ | | |
| 661 | @intFromEnum(req.response.status), | | |
| 662 | req.response.status.phrase() orelse "", | | |
| 663 | }); | | |
| 664 | } | | |
| 665 | | | |
| 666 | const content_type = req.response.headers.getFirstValue("Content-Type") orelse | | |
| 667 | return report.fail(dep.url_tok, "Missing 'Content-Type' header", .{}); | | |
| 668 | | | |
| 669 | var prog_reader: ProgressReader(std.http.Client.Request.Reader) = .{ | | |
| 670 | .child_reader = req.reader(), | | |
| 671 | .prog_node = &pkg_prog_node, | | |
| 672 | .unit = if (req.response.content_length) |content_length| unit: { | | |
| 673 | const kib = content_length / 1024; | | |
| 674 | const mib = kib / 1024; | | |
| 675 | if (mib > 0) { | | |
| 676 | pkg_prog_node.setEstimatedTotalItems(@intCast(mib)); | | |
| 677 | pkg_prog_node.setUnit("MiB"); | | |
| 678 | break :unit .mib; | | |
| 679 | } else { | | |
| 680 | pkg_prog_node.setEstimatedTotalItems(@intCast(@max(1, kib))); | | |
| 681 | pkg_prog_node.setUnit("KiB"); | | |
| 682 | break :unit .kib; | | |
| 683 | } | | |
| 684 | } else .any, | | |
| 685 | }; | | |
| 686 | pkg_prog_node.context.refresh(); | | |
| 687 | | | |
| 688 | if (ascii.eqlIgnoreCase(content_type, "application/gzip") or | | |
| 689 | ascii.eqlIgnoreCase(content_type, "application/x-gzip") or | | |
| 690 | ascii.eqlIgnoreCase(content_type, "application/tar+gzip")) | | |
| 691 | { | | |
| 692 | // I observed the gzip stream to read 1 byte at a time, so I am using a | | |
| 693 | // buffered reader on the front of it. | | |
| 694 | try unpackTarball(gpa, prog_reader.reader(), tmp_directory.handle, std.compress.gzip); | | |
| 695 | } else if (ascii.eqlIgnoreCase(content_type, "application/x-xz")) { | | |
| 696 | // I have not checked what buffer sizes the xz decompression implementation uses | | |
| 697 | // by default, so the same logic applies for buffering the reader as for gzip. | | |
| 698 | try unpackTarball(gpa, prog_reader.reader(), tmp_directory.handle, std.compress.xz); | | |
| 699 | } else if (ascii.eqlIgnoreCase(content_type, "application/octet-stream")) { | | |
| 700 | // support gitlab tarball urls such as https://gitlab.com/<namespace>/<project>/-/archive/<sha>/<project>-<sha>.tar.gz | | |
| 701 | // whose content-disposition header is: 'attachment; filename="<project>-<sha>.tar.gz"' | | |
| 702 | const content_disposition = req.response.headers.getFirstValue("Content-Disposition") orelse | | |
| 703 | return report.fail(dep.url_tok, "Missing 'Content-Disposition' header for Content-Type=application/octet-stream", .{}); | | |
| 704 | if (isTarAttachment(content_disposition)) { | | |
| 705 | try unpackTarball(gpa, prog_reader.reader(), tmp_directory.handle, std.compress.gzip); | | |
| 706 | } else return report.fail(dep.url_tok, "Unsupported 'Content-Disposition' header value: '{s}' for Content-Type=application/octet-stream", .{content_disposition}); | | |
| 707 | } else { | | |
| 708 | return report.fail(dep.url_tok, "Unsupported 'Content-Type' header value: '{s}'", .{content_type}); | | |
| 709 | } | | |
| 710 | | | |
| 711 | // Download completed - stop showing downloaded amount as progress | | |
| 712 | pkg_prog_node.setEstimatedTotalItems(0); | | |
| 713 | pkg_prog_node.setCompletedItems(0); | | |
| 714 | pkg_prog_node.context.refresh(); | | |
| 715 | | | |
| 716 | // TODO: delete files not included in the package prior to computing the package hash. | | |
| 717 | // for example, if the ini file has directives to include/not include certain files, | | |
| 718 | // apply those rules directly to the filesystem right here. This ensures that files | | |
| 719 | // not protected by the hash are not present on the file system. | | |
| 720 | | 952 | |
| 721 | // TODO: raise an error for files that have illegal paths on some operating systems. | 953 | var fetch_location = try FetchLocation.init(gpa, uri, directory, dep, report); |
| 722 | // For example, on Linux a path with a backslash should raise an error here. | 954 | defer fetch_location.deinit(gpa); |
| 723 | // Of course, if the ignore rules above omit the file from the package, then everything | | |
| 724 | // is fine and no error should be raised. | | |
| 725 | | 955 | |
| 726 | break :a try computePackageHash(thread_pool, .{ .dir = tmp_directory.handle }); | 956 | var readable_resource = try fetch_location.fetch(gpa, directory, http_client, dep, report); |
| 727 | }; | 957 | defer readable_resource.deinit(gpa); |
| 728 | | 958 | |
| 729 | const pkg_dir_sub_path = "p" ++ s ++ Manifest.hexDigest(actual_hash); | 959 | var package_location = try readable_resource.unpack(gpa, thread_pool, global_cache_directory, dep, report, &pkg_prog_node); |
| 730 | try renameTmpIntoCache(global_cache_directory.handle, tmp_dir_sub_path, pkg_dir_sub_path); | 960 | defer package_location.deinit(gpa); |
| 731 | | 961 | |
| 732 | const actual_hex = Manifest.hexDigest(actual_hash); | 962 | const actual_hex = Manifest.hexDigest(package_location.hash); |
| 733 | if (dep.hash) |h| { | 963 | if (dep.hash) |h| { |
| 734 | if (!mem.eql(u8, h, &actual_hex)) { | 964 | if (!mem.eql(u8, h, &actual_hex)) { |
| 735 | return report.fail(dep.hash_tok, "hash mismatch: expected: {s}, found: {s}", .{ | 965 | return report.fail(dep.hash_tok, "hash mismatch: expected: {s}, found: {s}", .{ |
| ... | @@ -743,9 +973,9 @@ fn fetchAndUnpack( | ... | @@ -743,9 +973,9 @@ fn fetchAndUnpack( |
| 743 | const eb = report.error_bundle; | 973 | const eb = report.error_bundle; |
| 744 | const notes_len = 1; | 974 | const notes_len = 1; |
| 745 | try Report.addErrorMessage(report.ast.*, file_path, eb, notes_len, .{ | 975 | try Report.addErrorMessage(report.ast.*, file_path, eb, notes_len, .{ |
| 746 | .tok = dep.url_tok, | 976 | .tok = dep.location_tok, |
| 747 | .off = 0, | 977 | .off = 0, |
| 748 | .msg = "url field is missing corresponding hash field", | 978 | .msg = "dependency is missing hash field", |
| 749 | }); | 979 | }); |
| 750 | const notes_start = try eb.reserveNotes(notes_len); | 980 | const notes_start = try eb.reserveNotes(notes_len); |
| 751 | eb.extra.items[notes_start] = @intFromEnum(try eb.addErrorMessage(.{ | 981 | eb.extra.items[notes_start] = @intFromEnum(try eb.addErrorMessage(.{ |
| ... | @@ -754,35 +984,24 @@ fn fetchAndUnpack( | ... | @@ -754,35 +984,24 @@ fn fetchAndUnpack( |
| 754 | return error.PackageFetchFailed; | 984 | return error.PackageFetchFailed; |
| 755 | } | 985 | } |
| 756 | | 986 | |
| 757 | const build_zig_path = try std.fs.path.join(gpa, &.{ pkg_dir_sub_path, build_zig_basename }); | 987 | const gop = try all_modules.getOrPut(gpa, actual_hex); |
| 758 | defer gpa.free(build_zig_path); | | |
| 759 | | | |
| 760 | global_cache_directory.handle.access(build_zig_path, .{}) catch |err| switch (err) { | | |
| 761 | error.FileNotFound => { | | |
| 762 | try all_modules.put(gpa, actual_hex, .non_zig_pkg); | | |
| 763 | return .{ | | |
| 764 | .mod = null, | | |
| 765 | .found_existing = false, | | |
| 766 | }; | | |
| 767 | }, | | |
| 768 | else => return err, | | |
| 769 | }; | | |
| 770 | | 988 | |
| 771 | const mod = try createWithDir(gpa, global_cache_directory, pkg_dir_sub_path, build_zig_basename); | 989 | if (gop.found_existing and gop.value_ptr.* != null) { |
| 772 | try all_modules.put(gpa, actual_hex, .{ .zig_pkg = mod }); | 990 | return gop.value_ptr.*.?; |
| 773 | return .{ | 991 | } else { |
| 774 | .mod = mod, | 992 | const module = try create(gpa, package_location.dir_path, build_zig_basename); |
| 775 | .found_existing = false, | 993 | gop.value_ptr.* = module; |
| 776 | }; | 994 | return module; |
| | 995 | } |
| 777 | } | 996 | } |
| 778 | | 997 | |
| 779 | fn unpackTarball( | 998 | fn unpackTarball( |
| 780 | gpa: Allocator, | 999 | gpa: Allocator, |
| 781 | req_reader: anytype, | 1000 | reader: anytype, |
| 782 | out_dir: fs.Dir, | 1001 | out_dir: fs.Dir, |
| 783 | comptime compression: type, | 1002 | comptime compression: type, |
| 784 | ) !void { | 1003 | ) !void { |
| 785 | var br = std.io.bufferedReaderSize(std.crypto.tls.max_ciphertext_record_len, req_reader); | 1004 | var br = std.io.bufferedReaderSize(std.crypto.tls.max_ciphertext_record_len, reader); |
| 786 | | 1005 | |
| 787 | var decompress = try compression.decompress(gpa, br.reader()); | 1006 | var decompress = try compression.decompress(gpa, br.reader()); |
| 788 | defer decompress.deinit(); | 1007 | defer decompress.deinit(); |
Adam Goertz <adambgoertz@gmail.com> 2023-07-12 02:45:51+00:00
Andrew Kelley <andrew@ziglang.org> 2023-09-29 00:32:43-07:00
lib