zig/lib/compiler/resinator/ico.zig
mlugg d0e74ffe52
compiler: rework comptime pointer representation and access
We've got a big one here! This commit reworks how we represent pointers
in the InternPool, and rewrites the logic for loading and storing from
them at comptime.

Firstly, the pointer representation. Previously, pointers were
represented in a highly structured manner: pointers to fields, array
elements, etc, were explicitly represented. This works well for simple
cases, but is quite difficult to handle in the cases of unusual
reinterpretations, pointer casts, offsets, etc. Therefore, pointers are
now represented in a more "flat" manner. For types without well-defined
layouts -- such as comptime-only types, automatic-layout aggregates, and
so on -- we still use this "hierarchical" structure. However, for types
with well-defined layouts, we use a byte offset associated with the
pointer. This allows the comptime pointer access logic to deal with
reinterpreted pointers far more gracefully, because the "base address"
of a pointer -- for instance a `field` -- is a single value which
pointer accesses cannot exceed since the parent has undefined layout.
This strategy is also more useful to most backends -- see the updated
logic in `codegen.zig` and `codegen/llvm.zig`. For backends which do
prefer a chain of field and elements accesses for lowering pointer
values, such as SPIR-V, there is a helpful function in `Value` which
creates a strategy to derive a pointer value using ideally only field
and element accesses. This is actually more correct than the previous
logic, since it correctly handles pointer casts which, after the dust
has settled, end up referring exactly to an aggregate field or array
element.

In terms of the pointer access code, it has been rewritten from the
ground up. The old logic had become rather a mess of special cases being
added whenever bugs were hit, and was still riddled with bugs. The new
logic was written to handle the "difficult" cases correctly, the most
notable of which is restructuring of a comptime-only array (for
instance, converting a `[3][2]comptime_int` to a `[2][3]comptime_int`.
Currently, the logic for loading and storing work somewhat differently,
but a future change will likely improve the loading logic to bring it
more in line with the store strategy. As far as I can tell, the rewrite
has fixed all bugs exposed by #19414.

As a part of this, the comptime bitcast logic has also been rewritten.
Previously, bitcasts simply worked by serializing the entire value into
an in-memory buffer, then deserializing it. This strategy has two key
weaknesses: pointers, and undefined values. Representations of these
values at comptime cannot be easily serialized/deserialized whilst
preserving data, which means many bitcasts would become runtime-known if
pointers were involved, or would turn `undefined` values into `0xAA`.
The new logic works by "flattening" the datastructure to be cast into a
sequence of bit-packed atomic values, and then "unflattening" it; using
serialization when necessary, but with special handling for `undefined`
values and for pointers which align in virtual memory. The resulting
code is definitely slower -- more on this later -- but it is correct.

The pointer access and bitcast logic required some helper functions and
types which are not generally useful elsewhere, so I opted to split them
into separate files `Sema/comptime_ptr_access.zig` and
`Sema/bitcast.zig`, with simple re-exports in `Sema.zig` for their small
public APIs.

Whilst working on this branch, I caught various unrelated bugs with
transitive Sema errors, and with the handling of `undefined` values.
These bugs have been fixed, and corresponding behavior test added.

In terms of performance, I do anticipate that this commit will regress
performance somewhat, because the new pointer access and bitcast logic
is necessarily more complex. I have not yet taken performance
measurements, but will do shortly, and post the results in this PR. If
the performance regression is severe, I will do work to to optimize the
new logic before merge.

Resolves: #19452
Resolves: #19460
2024-04-17 13:41:25 +01:00

312 lines
13 KiB
Zig

//! https://devblogs.microsoft.com/oldnewthing/20120720-00/?p=7083
//! https://learn.microsoft.com/en-us/previous-versions/ms997538(v=msdn.10)
//! https://learn.microsoft.com/en-us/windows/win32/menurc/newheader
//! https://learn.microsoft.com/en-us/windows/win32/menurc/resdir
//! https://learn.microsoft.com/en-us/windows/win32/menurc/localheader
const std = @import("std");
const builtin = @import("builtin");
const native_endian = builtin.cpu.arch.endian();
pub const ReadError = std.mem.Allocator.Error || error{ InvalidHeader, InvalidImageType, ImpossibleDataSize, UnexpectedEOF, ReadError };
pub fn read(allocator: std.mem.Allocator, reader: anytype, max_size: u64) ReadError!IconDir {
// Some Reader implementations have an empty ReadError error set which would
// cause 'unreachable else' if we tried to use an else in the switch, so we
// need to detect this case and not try to translate to ReadError
const empty_reader_errorset = @typeInfo(@TypeOf(reader).Error).ErrorSet == null or @typeInfo(@TypeOf(reader).Error).ErrorSet.?.len == 0;
if (empty_reader_errorset) {
return readAnyError(allocator, reader, max_size) catch |err| switch (err) {
error.EndOfStream => error.UnexpectedEOF,
else => |e| return e,
};
} else {
return readAnyError(allocator, reader, max_size) catch |err| switch (err) {
error.OutOfMemory,
error.InvalidHeader,
error.InvalidImageType,
error.ImpossibleDataSize,
=> |e| return e,
error.EndOfStream => error.UnexpectedEOF,
// The remaining errors are dependent on the `reader`, so
// we just translate them all to generic ReadError
else => error.ReadError,
};
}
}
// TODO: This seems like a somewhat strange pattern, could be a better way
// to do this. Maybe it makes more sense to handle the translation
// at the call site instead of having a helper function here.
pub fn readAnyError(allocator: std.mem.Allocator, reader: anytype, max_size: u64) !IconDir {
const reserved = try reader.readInt(u16, .little);
if (reserved != 0) {
return error.InvalidHeader;
}
const image_type = reader.readEnum(ImageType, .little) catch |err| switch (err) {
error.InvalidValue => return error.InvalidImageType,
else => |e| return e,
};
const num_images = try reader.readInt(u16, .little);
// To avoid over-allocation in the case of a file that says it has way more
// entries than it actually does, we use an ArrayList with a conservatively
// limited initial capacity instead of allocating the entire slice at once.
const initial_capacity = @min(num_images, 8);
var entries = try std.ArrayList(Entry).initCapacity(allocator, initial_capacity);
errdefer entries.deinit();
var i: usize = 0;
while (i < num_images) : (i += 1) {
var entry: Entry = undefined;
entry.width = try reader.readByte();
entry.height = try reader.readByte();
entry.num_colors = try reader.readByte();
entry.reserved = try reader.readByte();
switch (image_type) {
.icon => {
entry.type_specific_data = .{ .icon = .{
.color_planes = try reader.readInt(u16, .little),
.bits_per_pixel = try reader.readInt(u16, .little),
} };
},
.cursor => {
entry.type_specific_data = .{ .cursor = .{
.hotspot_x = try reader.readInt(u16, .little),
.hotspot_y = try reader.readInt(u16, .little),
} };
},
}
entry.data_size_in_bytes = try reader.readInt(u32, .little);
entry.data_offset_from_start_of_file = try reader.readInt(u32, .little);
// Validate that the offset/data size is feasible
if (@as(u64, entry.data_offset_from_start_of_file) + entry.data_size_in_bytes > max_size) {
return error.ImpossibleDataSize;
}
// and that the data size is large enough for at least the header of an image
// Note: This avoids needing to deal with a miscompilation from the Win32 RC
// compiler when the data size of an image is specified as zero but there
// is data to-be-read at the offset. The Win32 RC compiler will output
// an ICON/CURSOR resource with a bogus size in its header but with no actual
// data bytes in it, leading to an invalid .res. Similarly, if, for example,
// there is valid PNG data at the image's offset, but the size is specified
// as fewer bytes than the PNG header, then the Win32 RC compiler will still
// treat it as a PNG (e.g. unconditionally set num_planes to 1) but the data
// of the resource will only be 1 byte so treating it as a PNG doesn't make
// sense (especially not when you have to read past the data size to determine
// that it's a PNG).
if (entry.data_size_in_bytes < 16) {
return error.ImpossibleDataSize;
}
try entries.append(entry);
}
return .{
.image_type = image_type,
.entries = try entries.toOwnedSlice(),
.allocator = allocator,
};
}
pub const ImageType = enum(u16) {
icon = 1,
cursor = 2,
};
pub const IconDir = struct {
image_type: ImageType,
/// Note: entries.len will always fit into a u16, since the field containing the
/// number of images in an ico file is a u16.
entries: []Entry,
allocator: std.mem.Allocator,
pub fn deinit(self: IconDir) void {
self.allocator.free(self.entries);
}
pub const res_header_byte_len = 6;
pub fn getResDataSize(self: IconDir) u32 {
// maxInt(u16) * Entry.res_byte_len = 917,490 which is well within the u32 range.
// Note: self.entries.len is limited to maxInt(u16)
return @intCast(IconDir.res_header_byte_len + self.entries.len * Entry.res_byte_len);
}
pub fn writeResData(self: IconDir, writer: anytype, first_image_id: u16) !void {
try writer.writeInt(u16, 0, .little);
try writer.writeInt(u16, @intFromEnum(self.image_type), .little);
// We know that entries.len must fit into a u16
try writer.writeInt(u16, @as(u16, @intCast(self.entries.len)), .little);
var image_id = first_image_id;
for (self.entries) |entry| {
try entry.writeResData(writer, image_id);
image_id += 1;
}
}
};
pub const Entry = struct {
// Icons are limited to u8 sizes, cursors can have u16,
// so we store as u16 and truncate when needed.
width: u16,
height: u16,
num_colors: u8,
/// This should always be zero, but whatever value it is gets
/// carried over so we need to store it
reserved: u8,
type_specific_data: union(ImageType) {
icon: struct {
color_planes: u16,
bits_per_pixel: u16,
},
cursor: struct {
hotspot_x: u16,
hotspot_y: u16,
},
},
data_size_in_bytes: u32,
data_offset_from_start_of_file: u32,
pub const res_byte_len = 14;
pub fn writeResData(self: Entry, writer: anytype, id: u16) !void {
switch (self.type_specific_data) {
.icon => |icon_data| {
try writer.writeInt(u8, @as(u8, @truncate(self.width)), .little);
try writer.writeInt(u8, @as(u8, @truncate(self.height)), .little);
try writer.writeInt(u8, self.num_colors, .little);
try writer.writeInt(u8, self.reserved, .little);
try writer.writeInt(u16, icon_data.color_planes, .little);
try writer.writeInt(u16, icon_data.bits_per_pixel, .little);
try writer.writeInt(u32, self.data_size_in_bytes, .little);
},
.cursor => |cursor_data| {
try writer.writeInt(u16, self.width, .little);
try writer.writeInt(u16, self.height, .little);
try writer.writeInt(u16, cursor_data.hotspot_x, .little);
try writer.writeInt(u16, cursor_data.hotspot_y, .little);
try writer.writeInt(u32, self.data_size_in_bytes + 4, .little);
},
}
try writer.writeInt(u16, id, .little);
}
};
test "icon" {
const data = "\x00\x00\x01\x00\x01\x00\x10\x10\x00\x00\x01\x00\x10\x00\x10\x00\x00\x00\x16\x00\x00\x00" ++ [_]u8{0} ** 16;
var fbs = std.io.fixedBufferStream(data);
const icon = try read(std.testing.allocator, fbs.reader(), data.len);
defer icon.deinit();
try std.testing.expectEqual(ImageType.icon, icon.image_type);
try std.testing.expectEqual(@as(usize, 1), icon.entries.len);
}
test "icon too many images" {
// Note that with verifying that all data sizes are within the file bounds and >= 16,
// it's not possible to hit EOF when looking for more RESDIR structures, since they are
// themselves 16 bytes long, so we'll always hit ImpossibleDataSize instead.
const data = "\x00\x00\x01\x00\x02\x00\x10\x10\x00\x00\x01\x00\x10\x00\x10\x00\x00\x00\x16\x00\x00\x00" ++ [_]u8{0} ** 16;
var fbs = std.io.fixedBufferStream(data);
try std.testing.expectError(error.ImpossibleDataSize, read(std.testing.allocator, fbs.reader(), data.len));
}
test "icon data size past EOF" {
const data = "\x00\x00\x01\x00\x01\x00\x10\x10\x00\x00\x01\x00\x10\x00\x10\x01\x00\x00\x16\x00\x00\x00" ++ [_]u8{0} ** 16;
var fbs = std.io.fixedBufferStream(data);
try std.testing.expectError(error.ImpossibleDataSize, read(std.testing.allocator, fbs.reader(), data.len));
}
test "icon data offset past EOF" {
const data = "\x00\x00\x01\x00\x01\x00\x10\x10\x00\x00\x01\x00\x10\x00\x10\x00\x00\x00\x17\x00\x00\x00" ++ [_]u8{0} ** 16;
var fbs = std.io.fixedBufferStream(data);
try std.testing.expectError(error.ImpossibleDataSize, read(std.testing.allocator, fbs.reader(), data.len));
}
test "icon data size too small" {
const data = "\x00\x00\x01\x00\x01\x00\x10\x10\x00\x00\x01\x00\x10\x00\x0F\x00\x00\x00\x16\x00\x00\x00";
var fbs = std.io.fixedBufferStream(data);
try std.testing.expectError(error.ImpossibleDataSize, read(std.testing.allocator, fbs.reader(), data.len));
}
pub const ImageFormat = enum(u2) {
dib,
png,
riff,
const riff_header = std.mem.readInt(u32, "RIFF", native_endian);
const png_signature = std.mem.readInt(u64, "\x89PNG\r\n\x1a\n", native_endian);
const ihdr_code = std.mem.readInt(u32, "IHDR", native_endian);
const acon_form_type = std.mem.readInt(u32, "ACON", native_endian);
pub fn detect(header_bytes: *const [16]u8) ImageFormat {
if (std.mem.readInt(u32, header_bytes[0..4], native_endian) == riff_header) return .riff;
if (std.mem.readInt(u64, header_bytes[0..8], native_endian) == png_signature) return .png;
return .dib;
}
pub fn validate(format: ImageFormat, header_bytes: *const [16]u8) bool {
return switch (format) {
.png => std.mem.readInt(u32, header_bytes[12..16], native_endian) == ihdr_code,
.riff => std.mem.readInt(u32, header_bytes[8..12], native_endian) == acon_form_type,
.dib => true,
};
}
};
/// Contains only the fields of BITMAPINFOHEADER (WinGDI.h) that are both:
/// - relevant to what we need, and
/// - are shared between all versions of BITMAPINFOHEADER (V4, V5).
pub const BitmapHeader = extern struct {
bcSize: u32,
bcWidth: i32,
bcHeight: i32,
bcPlanes: u16,
bcBitCount: u16,
pub fn version(self: *const BitmapHeader) Version {
return Version.get(self.bcSize);
}
/// https://en.wikipedia.org/wiki/BMP_file_format#DIB_header_(bitmap_information_header)
pub const Version = enum(u3) {
unknown,
@"win2.0", // Windows 2.0 or later
@"nt3.1", // Windows NT, 3.1x or later
@"nt4.0", // Windows NT 4.0, 95 or later
@"nt5.0", // Windows NT 5.0, 98 or later
pub fn get(header_size: u32) Version {
return switch (header_size) {
len(.@"win2.0") => .@"win2.0",
len(.@"nt3.1") => .@"nt3.1",
len(.@"nt4.0") => .@"nt4.0",
len(.@"nt5.0") => .@"nt5.0",
else => .unknown,
};
}
pub fn len(comptime v: Version) comptime_int {
return switch (v) {
.@"win2.0" => 12,
.@"nt3.1" => 40,
.@"nt4.0" => 108,
.@"nt5.0" => 124,
.unknown => unreachable,
};
}
pub fn nameForErrorDisplay(v: Version) []const u8 {
return switch (v) {
.unknown => "unknown",
.@"win2.0" => "Windows 2.0 (BITMAPCOREHEADER)",
.@"nt3.1" => "Windows NT, 3.1x (BITMAPINFOHEADER)",
.@"nt4.0" => "Windows NT 4.0, 95 (BITMAPV4HEADER)",
.@"nt5.0" => "Windows NT 5.0, 98 (BITMAPV5HEADER)",
};
}
};
};