mirror of
https://github.com/ziglang/zig.git
synced 2024-12-11 21:56:27 +00:00
f3edff439e
`getExternalExecutor` is moved from `std.zig.CrossTarget` to `std.zig.system.NativeTargetInfo.getExternalExecutor`. The function also now communicates a bit more information about *why* the host is unable to execute a binary. The CLI is updated to report this information in a useful manner. `getExternalExecutor` is also improved to detect such patterns as: * x86_64 is able to execute x86 binaries * aarch64 is able to execute arm binaries * etc. Added qemu-hexagon support to `getExternalExecutor`. `std.Target.canExecBinaries` of is removed; callers should use the more powerful `getExternalExecutor` instead. Now that `zig test` tries to run the resulting binary no matter what, this commit has a follow-up change to the build system and docgen to utilize the `getExternalExecutor` function and pass `--test-no-exec` in some cases to avoid getting the error. Additionally: * refactor: extract NativePaths and NativeTargetInfo into their own files named after the structs. * small improvement to langref to reduce the complexity of the `callconv` expression in a couple examples.
1939 lines
80 KiB
Zig
1939 lines
80 KiB
Zig
const std = @import("std");
|
|
const builtin = @import("builtin");
|
|
const io = std.io;
|
|
const fs = std.fs;
|
|
const process = std.process;
|
|
const ChildProcess = std.ChildProcess;
|
|
const Progress = std.Progress;
|
|
const print = std.debug.print;
|
|
const mem = std.mem;
|
|
const testing = std.testing;
|
|
const Allocator = std.mem.Allocator;
|
|
|
|
const max_doc_file_size = 10 * 1024 * 1024;
|
|
|
|
const exe_ext = @as(std.zig.CrossTarget, .{}).exeFileExt();
|
|
const obj_ext = builtin.object_format.fileExt(builtin.cpu.arch);
|
|
const tmp_dir_name = "docgen_tmp";
|
|
const test_out_path = tmp_dir_name ++ fs.path.sep_str ++ "test" ++ exe_ext;
|
|
|
|
pub fn main() !void {
|
|
var arena = std.heap.ArenaAllocator.init(std.heap.page_allocator);
|
|
defer arena.deinit();
|
|
|
|
const allocator = arena.allocator();
|
|
|
|
var args_it = process.args();
|
|
|
|
if (!args_it.skip()) @panic("expected self arg");
|
|
|
|
const zig_exe = try (args_it.next(allocator) orelse @panic("expected zig exe arg"));
|
|
defer allocator.free(zig_exe);
|
|
|
|
const in_file_name = try (args_it.next(allocator) orelse @panic("expected input arg"));
|
|
defer allocator.free(in_file_name);
|
|
|
|
const out_file_name = try (args_it.next(allocator) orelse @panic("expected output arg"));
|
|
defer allocator.free(out_file_name);
|
|
|
|
var do_code_tests = true;
|
|
if (args_it.next(allocator)) |arg| {
|
|
if (mem.eql(u8, try arg, "--skip-code-tests")) {
|
|
do_code_tests = false;
|
|
} else {
|
|
@panic("unrecognized arg");
|
|
}
|
|
}
|
|
|
|
var in_file = try fs.cwd().openFile(in_file_name, .{ .read = true });
|
|
defer in_file.close();
|
|
|
|
var out_file = try fs.cwd().createFile(out_file_name, .{});
|
|
defer out_file.close();
|
|
|
|
const input_file_bytes = try in_file.reader().readAllAlloc(allocator, max_doc_file_size);
|
|
|
|
var buffered_writer = io.bufferedWriter(out_file.writer());
|
|
|
|
var tokenizer = Tokenizer.init(in_file_name, input_file_bytes);
|
|
var toc = try genToc(allocator, &tokenizer);
|
|
|
|
try fs.cwd().makePath(tmp_dir_name);
|
|
defer fs.cwd().deleteTree(tmp_dir_name) catch {};
|
|
|
|
try genHtml(allocator, &tokenizer, &toc, buffered_writer.writer(), zig_exe, do_code_tests);
|
|
try buffered_writer.flush();
|
|
}
|
|
|
|
const Token = struct {
|
|
id: Id,
|
|
start: usize,
|
|
end: usize,
|
|
|
|
const Id = enum {
|
|
Invalid,
|
|
Content,
|
|
BracketOpen,
|
|
TagContent,
|
|
Separator,
|
|
BracketClose,
|
|
Eof,
|
|
};
|
|
};
|
|
|
|
const Tokenizer = struct {
|
|
buffer: []const u8,
|
|
index: usize,
|
|
state: State,
|
|
source_file_name: []const u8,
|
|
code_node_count: usize,
|
|
|
|
const State = enum {
|
|
Start,
|
|
LBracket,
|
|
Hash,
|
|
TagName,
|
|
Eof,
|
|
};
|
|
|
|
fn init(source_file_name: []const u8, buffer: []const u8) Tokenizer {
|
|
return Tokenizer{
|
|
.buffer = buffer,
|
|
.index = 0,
|
|
.state = State.Start,
|
|
.source_file_name = source_file_name,
|
|
.code_node_count = 0,
|
|
};
|
|
}
|
|
|
|
fn next(self: *Tokenizer) Token {
|
|
var result = Token{
|
|
.id = Token.Id.Eof,
|
|
.start = self.index,
|
|
.end = undefined,
|
|
};
|
|
while (self.index < self.buffer.len) : (self.index += 1) {
|
|
const c = self.buffer[self.index];
|
|
switch (self.state) {
|
|
State.Start => switch (c) {
|
|
'{' => {
|
|
self.state = State.LBracket;
|
|
},
|
|
else => {
|
|
result.id = Token.Id.Content;
|
|
},
|
|
},
|
|
State.LBracket => switch (c) {
|
|
'#' => {
|
|
if (result.id != Token.Id.Eof) {
|
|
self.index -= 1;
|
|
self.state = State.Start;
|
|
break;
|
|
} else {
|
|
result.id = Token.Id.BracketOpen;
|
|
self.index += 1;
|
|
self.state = State.TagName;
|
|
break;
|
|
}
|
|
},
|
|
else => {
|
|
result.id = Token.Id.Content;
|
|
self.state = State.Start;
|
|
},
|
|
},
|
|
State.TagName => switch (c) {
|
|
'|' => {
|
|
if (result.id != Token.Id.Eof) {
|
|
break;
|
|
} else {
|
|
result.id = Token.Id.Separator;
|
|
self.index += 1;
|
|
break;
|
|
}
|
|
},
|
|
'#' => {
|
|
self.state = State.Hash;
|
|
},
|
|
else => {
|
|
result.id = Token.Id.TagContent;
|
|
},
|
|
},
|
|
State.Hash => switch (c) {
|
|
'}' => {
|
|
if (result.id != Token.Id.Eof) {
|
|
self.index -= 1;
|
|
self.state = State.TagName;
|
|
break;
|
|
} else {
|
|
result.id = Token.Id.BracketClose;
|
|
self.index += 1;
|
|
self.state = State.Start;
|
|
break;
|
|
}
|
|
},
|
|
else => {
|
|
result.id = Token.Id.TagContent;
|
|
self.state = State.TagName;
|
|
},
|
|
},
|
|
State.Eof => unreachable,
|
|
}
|
|
} else {
|
|
switch (self.state) {
|
|
State.Start, State.LBracket, State.Eof => {},
|
|
else => {
|
|
result.id = Token.Id.Invalid;
|
|
},
|
|
}
|
|
self.state = State.Eof;
|
|
}
|
|
result.end = self.index;
|
|
return result;
|
|
}
|
|
|
|
const Location = struct {
|
|
line: usize,
|
|
column: usize,
|
|
line_start: usize,
|
|
line_end: usize,
|
|
};
|
|
|
|
fn getTokenLocation(self: *Tokenizer, token: Token) Location {
|
|
var loc = Location{
|
|
.line = 0,
|
|
.column = 0,
|
|
.line_start = 0,
|
|
.line_end = 0,
|
|
};
|
|
for (self.buffer) |c, i| {
|
|
if (i == token.start) {
|
|
loc.line_end = i;
|
|
while (loc.line_end < self.buffer.len and self.buffer[loc.line_end] != '\n') : (loc.line_end += 1) {}
|
|
return loc;
|
|
}
|
|
if (c == '\n') {
|
|
loc.line += 1;
|
|
loc.column = 0;
|
|
loc.line_start = i + 1;
|
|
} else {
|
|
loc.column += 1;
|
|
}
|
|
}
|
|
return loc;
|
|
}
|
|
};
|
|
|
|
fn parseError(tokenizer: *Tokenizer, token: Token, comptime fmt: []const u8, args: anytype) anyerror {
|
|
const loc = tokenizer.getTokenLocation(token);
|
|
const args_prefix = .{ tokenizer.source_file_name, loc.line + 1, loc.column + 1 };
|
|
print("{s}:{d}:{d}: error: " ++ fmt ++ "\n", args_prefix ++ args);
|
|
if (loc.line_start <= loc.line_end) {
|
|
print("{s}\n", .{tokenizer.buffer[loc.line_start..loc.line_end]});
|
|
{
|
|
var i: usize = 0;
|
|
while (i < loc.column) : (i += 1) {
|
|
print(" ", .{});
|
|
}
|
|
}
|
|
{
|
|
const caret_count = std.math.min(token.end, loc.line_end) - token.start;
|
|
var i: usize = 0;
|
|
while (i < caret_count) : (i += 1) {
|
|
print("~", .{});
|
|
}
|
|
}
|
|
print("\n", .{});
|
|
}
|
|
return error.ParseError;
|
|
}
|
|
|
|
fn assertToken(tokenizer: *Tokenizer, token: Token, id: Token.Id) !void {
|
|
if (token.id != id) {
|
|
return parseError(tokenizer, token, "expected {s}, found {s}", .{ @tagName(id), @tagName(token.id) });
|
|
}
|
|
}
|
|
|
|
fn eatToken(tokenizer: *Tokenizer, id: Token.Id) !Token {
|
|
const token = tokenizer.next();
|
|
try assertToken(tokenizer, token, id);
|
|
return token;
|
|
}
|
|
|
|
const HeaderOpen = struct {
|
|
name: []const u8,
|
|
url: []const u8,
|
|
n: usize,
|
|
};
|
|
|
|
const SeeAlsoItem = struct {
|
|
name: []const u8,
|
|
token: Token,
|
|
};
|
|
|
|
const ExpectedOutcome = enum {
|
|
Succeed,
|
|
Fail,
|
|
BuildFail,
|
|
};
|
|
|
|
const Code = struct {
|
|
id: Id,
|
|
name: []const u8,
|
|
source_token: Token,
|
|
just_check_syntax: bool,
|
|
mode: std.builtin.Mode,
|
|
link_objects: []const []const u8,
|
|
target_str: ?[]const u8,
|
|
link_libc: bool,
|
|
link_mode: ?std.builtin.LinkMode,
|
|
disable_cache: bool,
|
|
verbose_cimport: bool,
|
|
|
|
const Id = union(enum) {
|
|
Test,
|
|
TestError: []const u8,
|
|
TestSafety: []const u8,
|
|
Exe: ExpectedOutcome,
|
|
Obj: ?[]const u8,
|
|
Lib,
|
|
};
|
|
};
|
|
|
|
const Link = struct {
|
|
url: []const u8,
|
|
name: []const u8,
|
|
token: Token,
|
|
};
|
|
|
|
const SyntaxBlock = struct {
|
|
source_type: SourceType,
|
|
name: []const u8,
|
|
source_token: Token,
|
|
|
|
const SourceType = enum {
|
|
zig,
|
|
c,
|
|
peg,
|
|
javascript,
|
|
};
|
|
};
|
|
|
|
const Node = union(enum) {
|
|
Content: []const u8,
|
|
Nav,
|
|
Builtin: Token,
|
|
HeaderOpen: HeaderOpen,
|
|
SeeAlso: []const SeeAlsoItem,
|
|
Code: Code,
|
|
Link: Link,
|
|
InlineSyntax: Token,
|
|
Shell: Token,
|
|
SyntaxBlock: SyntaxBlock,
|
|
};
|
|
|
|
const Toc = struct {
|
|
nodes: []Node,
|
|
toc: []u8,
|
|
urls: std.StringHashMap(Token),
|
|
};
|
|
|
|
const Action = enum {
|
|
Open,
|
|
Close,
|
|
};
|
|
|
|
fn genToc(allocator: Allocator, tokenizer: *Tokenizer) !Toc {
|
|
var urls = std.StringHashMap(Token).init(allocator);
|
|
errdefer urls.deinit();
|
|
|
|
var header_stack_size: usize = 0;
|
|
var last_action = Action.Open;
|
|
var last_columns: ?u8 = null;
|
|
|
|
var toc_buf = std.ArrayList(u8).init(allocator);
|
|
defer toc_buf.deinit();
|
|
|
|
var toc = toc_buf.writer();
|
|
|
|
var nodes = std.ArrayList(Node).init(allocator);
|
|
defer nodes.deinit();
|
|
|
|
try toc.writeByte('\n');
|
|
|
|
while (true) {
|
|
const token = tokenizer.next();
|
|
switch (token.id) {
|
|
Token.Id.Eof => {
|
|
if (header_stack_size != 0) {
|
|
return parseError(tokenizer, token, "unbalanced headers", .{});
|
|
}
|
|
try toc.writeAll(" </ul>\n");
|
|
break;
|
|
},
|
|
Token.Id.Content => {
|
|
try nodes.append(Node{ .Content = tokenizer.buffer[token.start..token.end] });
|
|
},
|
|
Token.Id.BracketOpen => {
|
|
const tag_token = try eatToken(tokenizer, Token.Id.TagContent);
|
|
const tag_name = tokenizer.buffer[tag_token.start..tag_token.end];
|
|
|
|
if (mem.eql(u8, tag_name, "nav")) {
|
|
_ = try eatToken(tokenizer, Token.Id.BracketClose);
|
|
|
|
try nodes.append(Node.Nav);
|
|
} else if (mem.eql(u8, tag_name, "builtin")) {
|
|
_ = try eatToken(tokenizer, Token.Id.BracketClose);
|
|
try nodes.append(Node{ .Builtin = tag_token });
|
|
} else if (mem.eql(u8, tag_name, "header_open")) {
|
|
_ = try eatToken(tokenizer, Token.Id.Separator);
|
|
const content_token = try eatToken(tokenizer, Token.Id.TagContent);
|
|
const content = tokenizer.buffer[content_token.start..content_token.end];
|
|
var columns: ?u8 = null;
|
|
while (true) {
|
|
const bracket_tok = tokenizer.next();
|
|
switch (bracket_tok.id) {
|
|
.BracketClose => break,
|
|
.Separator => continue,
|
|
.TagContent => {
|
|
const param = tokenizer.buffer[bracket_tok.start..bracket_tok.end];
|
|
if (mem.eql(u8, param, "2col")) {
|
|
columns = 2;
|
|
} else {
|
|
return parseError(
|
|
tokenizer,
|
|
bracket_tok,
|
|
"unrecognized header_open param: {s}",
|
|
.{param},
|
|
);
|
|
}
|
|
},
|
|
else => return parseError(tokenizer, bracket_tok, "invalid header_open token", .{}),
|
|
}
|
|
}
|
|
|
|
header_stack_size += 1;
|
|
|
|
const urlized = try urlize(allocator, content);
|
|
try nodes.append(Node{
|
|
.HeaderOpen = HeaderOpen{
|
|
.name = content,
|
|
.url = urlized,
|
|
.n = header_stack_size + 1, // highest-level section headers start at h2
|
|
},
|
|
});
|
|
if (try urls.fetchPut(urlized, tag_token)) |kv| {
|
|
parseError(tokenizer, tag_token, "duplicate header url: #{s}", .{urlized}) catch {};
|
|
parseError(tokenizer, kv.value, "other tag here", .{}) catch {};
|
|
return error.ParseError;
|
|
}
|
|
if (last_action == Action.Open) {
|
|
try toc.writeByte('\n');
|
|
try toc.writeByteNTimes(' ', header_stack_size * 4);
|
|
if (last_columns) |n| {
|
|
try toc.print("<ul style=\"columns: {}\">\n", .{n});
|
|
} else {
|
|
try toc.writeAll("<ul>\n");
|
|
}
|
|
} else {
|
|
last_action = Action.Open;
|
|
}
|
|
last_columns = columns;
|
|
try toc.writeByteNTimes(' ', 4 + header_stack_size * 4);
|
|
try toc.print("<li><a id=\"toc-{s}\" href=\"#{s}\">{s}</a>", .{ urlized, urlized, content });
|
|
} else if (mem.eql(u8, tag_name, "header_close")) {
|
|
if (header_stack_size == 0) {
|
|
return parseError(tokenizer, tag_token, "unbalanced close header", .{});
|
|
}
|
|
header_stack_size -= 1;
|
|
_ = try eatToken(tokenizer, Token.Id.BracketClose);
|
|
|
|
if (last_action == Action.Close) {
|
|
try toc.writeByteNTimes(' ', 8 + header_stack_size * 4);
|
|
try toc.writeAll("</ul></li>\n");
|
|
} else {
|
|
try toc.writeAll("</li>\n");
|
|
last_action = Action.Close;
|
|
}
|
|
} else if (mem.eql(u8, tag_name, "see_also")) {
|
|
var list = std.ArrayList(SeeAlsoItem).init(allocator);
|
|
errdefer list.deinit();
|
|
|
|
while (true) {
|
|
const see_also_tok = tokenizer.next();
|
|
switch (see_also_tok.id) {
|
|
Token.Id.TagContent => {
|
|
const content = tokenizer.buffer[see_also_tok.start..see_also_tok.end];
|
|
try list.append(SeeAlsoItem{
|
|
.name = content,
|
|
.token = see_also_tok,
|
|
});
|
|
},
|
|
Token.Id.Separator => {},
|
|
Token.Id.BracketClose => {
|
|
try nodes.append(Node{ .SeeAlso = list.toOwnedSlice() });
|
|
break;
|
|
},
|
|
else => return parseError(tokenizer, see_also_tok, "invalid see_also token", .{}),
|
|
}
|
|
}
|
|
} else if (mem.eql(u8, tag_name, "link")) {
|
|
_ = try eatToken(tokenizer, Token.Id.Separator);
|
|
const name_tok = try eatToken(tokenizer, Token.Id.TagContent);
|
|
const name = tokenizer.buffer[name_tok.start..name_tok.end];
|
|
|
|
const url_name = blk: {
|
|
const tok = tokenizer.next();
|
|
switch (tok.id) {
|
|
Token.Id.BracketClose => break :blk name,
|
|
Token.Id.Separator => {
|
|
const explicit_text = try eatToken(tokenizer, Token.Id.TagContent);
|
|
_ = try eatToken(tokenizer, Token.Id.BracketClose);
|
|
break :blk tokenizer.buffer[explicit_text.start..explicit_text.end];
|
|
},
|
|
else => return parseError(tokenizer, tok, "invalid link token", .{}),
|
|
}
|
|
};
|
|
|
|
try nodes.append(Node{
|
|
.Link = Link{
|
|
.url = try urlize(allocator, url_name),
|
|
.name = name,
|
|
.token = name_tok,
|
|
},
|
|
});
|
|
} else if (mem.eql(u8, tag_name, "code_begin")) {
|
|
_ = try eatToken(tokenizer, Token.Id.Separator);
|
|
const code_kind_tok = try eatToken(tokenizer, Token.Id.TagContent);
|
|
var name: []const u8 = "test";
|
|
const maybe_sep = tokenizer.next();
|
|
switch (maybe_sep.id) {
|
|
Token.Id.Separator => {
|
|
const name_tok = try eatToken(tokenizer, Token.Id.TagContent);
|
|
name = tokenizer.buffer[name_tok.start..name_tok.end];
|
|
_ = try eatToken(tokenizer, Token.Id.BracketClose);
|
|
},
|
|
Token.Id.BracketClose => {},
|
|
else => return parseError(tokenizer, token, "invalid token", .{}),
|
|
}
|
|
const code_kind_str = tokenizer.buffer[code_kind_tok.start..code_kind_tok.end];
|
|
var code_kind_id: Code.Id = undefined;
|
|
var just_check_syntax = false;
|
|
if (mem.eql(u8, code_kind_str, "exe")) {
|
|
code_kind_id = Code.Id{ .Exe = ExpectedOutcome.Succeed };
|
|
} else if (mem.eql(u8, code_kind_str, "exe_err")) {
|
|
code_kind_id = Code.Id{ .Exe = ExpectedOutcome.Fail };
|
|
} else if (mem.eql(u8, code_kind_str, "exe_build_err")) {
|
|
code_kind_id = Code.Id{ .Exe = ExpectedOutcome.BuildFail };
|
|
} else if (mem.eql(u8, code_kind_str, "test")) {
|
|
code_kind_id = Code.Id.Test;
|
|
} else if (mem.eql(u8, code_kind_str, "test_err")) {
|
|
code_kind_id = Code.Id{ .TestError = name };
|
|
name = "test";
|
|
} else if (mem.eql(u8, code_kind_str, "test_safety")) {
|
|
code_kind_id = Code.Id{ .TestSafety = name };
|
|
name = "test";
|
|
} else if (mem.eql(u8, code_kind_str, "obj")) {
|
|
code_kind_id = Code.Id{ .Obj = null };
|
|
} else if (mem.eql(u8, code_kind_str, "obj_err")) {
|
|
code_kind_id = Code.Id{ .Obj = name };
|
|
name = "test";
|
|
} else if (mem.eql(u8, code_kind_str, "lib")) {
|
|
code_kind_id = Code.Id.Lib;
|
|
} else if (mem.eql(u8, code_kind_str, "syntax")) {
|
|
code_kind_id = Code.Id{ .Obj = null };
|
|
just_check_syntax = true;
|
|
} else {
|
|
return parseError(tokenizer, code_kind_tok, "unrecognized code kind: {s}", .{code_kind_str});
|
|
}
|
|
|
|
var mode: std.builtin.Mode = .Debug;
|
|
var link_objects = std.ArrayList([]const u8).init(allocator);
|
|
defer link_objects.deinit();
|
|
var target_str: ?[]const u8 = null;
|
|
var link_libc = false;
|
|
var link_mode: ?std.builtin.LinkMode = null;
|
|
var disable_cache = false;
|
|
var verbose_cimport = false;
|
|
|
|
const source_token = while (true) {
|
|
const content_tok = try eatToken(tokenizer, Token.Id.Content);
|
|
_ = try eatToken(tokenizer, Token.Id.BracketOpen);
|
|
const end_code_tag = try eatToken(tokenizer, Token.Id.TagContent);
|
|
const end_tag_name = tokenizer.buffer[end_code_tag.start..end_code_tag.end];
|
|
if (mem.eql(u8, end_tag_name, "code_release_fast")) {
|
|
mode = .ReleaseFast;
|
|
} else if (mem.eql(u8, end_tag_name, "code_release_safe")) {
|
|
mode = .ReleaseSafe;
|
|
} else if (mem.eql(u8, end_tag_name, "code_disable_cache")) {
|
|
disable_cache = true;
|
|
} else if (mem.eql(u8, end_tag_name, "code_verbose_cimport")) {
|
|
verbose_cimport = true;
|
|
} else if (mem.eql(u8, end_tag_name, "code_link_object")) {
|
|
_ = try eatToken(tokenizer, Token.Id.Separator);
|
|
const obj_tok = try eatToken(tokenizer, Token.Id.TagContent);
|
|
try link_objects.append(tokenizer.buffer[obj_tok.start..obj_tok.end]);
|
|
} else if (mem.eql(u8, end_tag_name, "target_windows")) {
|
|
target_str = "x86_64-windows";
|
|
} else if (mem.eql(u8, end_tag_name, "target_linux_x86_64")) {
|
|
target_str = "x86_64-linux";
|
|
} else if (mem.eql(u8, end_tag_name, "target_linux_riscv64")) {
|
|
target_str = "riscv64-linux";
|
|
} else if (mem.eql(u8, end_tag_name, "target_wasm")) {
|
|
target_str = "wasm32-freestanding";
|
|
} else if (mem.eql(u8, end_tag_name, "target_wasi")) {
|
|
target_str = "wasm32-wasi";
|
|
} else if (mem.eql(u8, end_tag_name, "link_libc")) {
|
|
link_libc = true;
|
|
} else if (mem.eql(u8, end_tag_name, "link_mode_dynamic")) {
|
|
link_mode = .Dynamic;
|
|
} else if (mem.eql(u8, end_tag_name, "code_end")) {
|
|
_ = try eatToken(tokenizer, Token.Id.BracketClose);
|
|
break content_tok;
|
|
} else {
|
|
return parseError(
|
|
tokenizer,
|
|
end_code_tag,
|
|
"invalid token inside code_begin: {s}",
|
|
.{end_tag_name},
|
|
);
|
|
}
|
|
_ = try eatToken(tokenizer, Token.Id.BracketClose);
|
|
} else unreachable; // TODO issue #707
|
|
try nodes.append(Node{
|
|
.Code = Code{
|
|
.id = code_kind_id,
|
|
.name = name,
|
|
.source_token = source_token,
|
|
.just_check_syntax = just_check_syntax,
|
|
.mode = mode,
|
|
.link_objects = link_objects.toOwnedSlice(),
|
|
.target_str = target_str,
|
|
.link_libc = link_libc,
|
|
.link_mode = link_mode,
|
|
.disable_cache = disable_cache,
|
|
.verbose_cimport = verbose_cimport,
|
|
},
|
|
});
|
|
tokenizer.code_node_count += 1;
|
|
} else if (mem.eql(u8, tag_name, "syntax")) {
|
|
_ = try eatToken(tokenizer, Token.Id.BracketClose);
|
|
const content_tok = try eatToken(tokenizer, Token.Id.Content);
|
|
_ = try eatToken(tokenizer, Token.Id.BracketOpen);
|
|
const end_syntax_tag = try eatToken(tokenizer, Token.Id.TagContent);
|
|
const end_tag_name = tokenizer.buffer[end_syntax_tag.start..end_syntax_tag.end];
|
|
if (!mem.eql(u8, end_tag_name, "endsyntax")) {
|
|
return parseError(
|
|
tokenizer,
|
|
end_syntax_tag,
|
|
"invalid token inside syntax: {s}",
|
|
.{end_tag_name},
|
|
);
|
|
}
|
|
_ = try eatToken(tokenizer, Token.Id.BracketClose);
|
|
try nodes.append(Node{ .InlineSyntax = content_tok });
|
|
} else if (mem.eql(u8, tag_name, "shell_samp")) {
|
|
_ = try eatToken(tokenizer, Token.Id.BracketClose);
|
|
const content_tok = try eatToken(tokenizer, Token.Id.Content);
|
|
_ = try eatToken(tokenizer, Token.Id.BracketOpen);
|
|
const end_syntax_tag = try eatToken(tokenizer, Token.Id.TagContent);
|
|
const end_tag_name = tokenizer.buffer[end_syntax_tag.start..end_syntax_tag.end];
|
|
if (!mem.eql(u8, end_tag_name, "end_shell_samp")) {
|
|
return parseError(
|
|
tokenizer,
|
|
end_syntax_tag,
|
|
"invalid token inside syntax: {s}",
|
|
.{end_tag_name},
|
|
);
|
|
}
|
|
_ = try eatToken(tokenizer, Token.Id.BracketClose);
|
|
try nodes.append(Node{ .Shell = content_tok });
|
|
} else if (mem.eql(u8, tag_name, "syntax_block")) {
|
|
_ = try eatToken(tokenizer, Token.Id.Separator);
|
|
const source_type_tok = try eatToken(tokenizer, Token.Id.TagContent);
|
|
var name: []const u8 = "sample_code";
|
|
const maybe_sep = tokenizer.next();
|
|
switch (maybe_sep.id) {
|
|
Token.Id.Separator => {
|
|
const name_tok = try eatToken(tokenizer, Token.Id.TagContent);
|
|
name = tokenizer.buffer[name_tok.start..name_tok.end];
|
|
_ = try eatToken(tokenizer, Token.Id.BracketClose);
|
|
},
|
|
Token.Id.BracketClose => {},
|
|
else => return parseError(tokenizer, token, "invalid token", .{}),
|
|
}
|
|
const source_type_str = tokenizer.buffer[source_type_tok.start..source_type_tok.end];
|
|
var source_type: SyntaxBlock.SourceType = undefined;
|
|
if (mem.eql(u8, source_type_str, "zig")) {
|
|
source_type = SyntaxBlock.SourceType.zig;
|
|
} else if (mem.eql(u8, source_type_str, "c")) {
|
|
source_type = SyntaxBlock.SourceType.c;
|
|
} else if (mem.eql(u8, source_type_str, "peg")) {
|
|
source_type = SyntaxBlock.SourceType.peg;
|
|
} else if (mem.eql(u8, source_type_str, "javascript")) {
|
|
source_type = SyntaxBlock.SourceType.javascript;
|
|
} else {
|
|
return parseError(tokenizer, source_type_tok, "unrecognized code kind: {s}", .{source_type_str});
|
|
}
|
|
const source_token = while (true) {
|
|
const content_tok = try eatToken(tokenizer, Token.Id.Content);
|
|
_ = try eatToken(tokenizer, Token.Id.BracketOpen);
|
|
const end_code_tag = try eatToken(tokenizer, Token.Id.TagContent);
|
|
const end_tag_name = tokenizer.buffer[end_code_tag.start..end_code_tag.end];
|
|
if (mem.eql(u8, end_tag_name, "end_syntax_block")) {
|
|
_ = try eatToken(tokenizer, Token.Id.BracketClose);
|
|
break content_tok;
|
|
} else {
|
|
return parseError(
|
|
tokenizer,
|
|
end_code_tag,
|
|
"invalid token inside code_begin: {s}",
|
|
.{end_tag_name},
|
|
);
|
|
}
|
|
_ = try eatToken(tokenizer, Token.Id.BracketClose);
|
|
} else unreachable; // TODO issue #707
|
|
try nodes.append(Node{ .SyntaxBlock = SyntaxBlock{ .source_type = source_type, .name = name, .source_token = source_token } });
|
|
} else {
|
|
return parseError(tokenizer, tag_token, "unrecognized tag name: {s}", .{tag_name});
|
|
}
|
|
},
|
|
else => return parseError(tokenizer, token, "invalid token", .{}),
|
|
}
|
|
}
|
|
|
|
return Toc{
|
|
.nodes = nodes.toOwnedSlice(),
|
|
.toc = toc_buf.toOwnedSlice(),
|
|
.urls = urls,
|
|
};
|
|
}
|
|
|
|
fn urlize(allocator: Allocator, input: []const u8) ![]u8 {
|
|
var buf = std.ArrayList(u8).init(allocator);
|
|
defer buf.deinit();
|
|
|
|
const out = buf.writer();
|
|
for (input) |c| {
|
|
switch (c) {
|
|
'a'...'z', 'A'...'Z', '_', '-', '0'...'9' => {
|
|
try out.writeByte(c);
|
|
},
|
|
' ' => {
|
|
try out.writeByte('-');
|
|
},
|
|
else => {},
|
|
}
|
|
}
|
|
return buf.toOwnedSlice();
|
|
}
|
|
|
|
fn escapeHtml(allocator: Allocator, input: []const u8) ![]u8 {
|
|
var buf = std.ArrayList(u8).init(allocator);
|
|
defer buf.deinit();
|
|
|
|
const out = buf.writer();
|
|
try writeEscaped(out, input);
|
|
return buf.toOwnedSlice();
|
|
}
|
|
|
|
fn writeEscaped(out: anytype, input: []const u8) !void {
|
|
for (input) |c| {
|
|
try switch (c) {
|
|
'&' => out.writeAll("&"),
|
|
'<' => out.writeAll("<"),
|
|
'>' => out.writeAll(">"),
|
|
'"' => out.writeAll("""),
|
|
else => out.writeByte(c),
|
|
};
|
|
}
|
|
}
|
|
|
|
//#define VT_RED "\x1b[31;1m"
|
|
//#define VT_GREEN "\x1b[32;1m"
|
|
//#define VT_CYAN "\x1b[36;1m"
|
|
//#define VT_WHITE "\x1b[37;1m"
|
|
//#define VT_BOLD "\x1b[0;1m"
|
|
//#define VT_RESET "\x1b[0m"
|
|
|
|
const TermState = enum {
|
|
Start,
|
|
Escape,
|
|
LBracket,
|
|
Number,
|
|
AfterNumber,
|
|
Arg,
|
|
ArgNumber,
|
|
ExpectEnd,
|
|
};
|
|
|
|
test "term color" {
|
|
const input_bytes = "A\x1b[32;1mgreen\x1b[0mB";
|
|
const result = try termColor(std.testing.allocator, input_bytes);
|
|
defer std.testing.allocator.free(result);
|
|
try testing.expectEqualSlices(u8, "A<span class=\"t32_1\">green</span>B", result);
|
|
}
|
|
|
|
fn termColor(allocator: Allocator, input: []const u8) ![]u8 {
|
|
var buf = std.ArrayList(u8).init(allocator);
|
|
defer buf.deinit();
|
|
|
|
var out = buf.writer();
|
|
var number_start_index: usize = undefined;
|
|
var first_number: usize = undefined;
|
|
var second_number: usize = undefined;
|
|
var i: usize = 0;
|
|
var state = TermState.Start;
|
|
var open_span_count: usize = 0;
|
|
while (i < input.len) : (i += 1) {
|
|
const c = input[i];
|
|
switch (state) {
|
|
TermState.Start => switch (c) {
|
|
'\x1b' => state = TermState.Escape,
|
|
else => try out.writeByte(c),
|
|
},
|
|
TermState.Escape => switch (c) {
|
|
'[' => state = TermState.LBracket,
|
|
else => return error.UnsupportedEscape,
|
|
},
|
|
TermState.LBracket => switch (c) {
|
|
'0'...'9' => {
|
|
number_start_index = i;
|
|
state = TermState.Number;
|
|
},
|
|
else => return error.UnsupportedEscape,
|
|
},
|
|
TermState.Number => switch (c) {
|
|
'0'...'9' => {},
|
|
else => {
|
|
first_number = std.fmt.parseInt(usize, input[number_start_index..i], 10) catch unreachable;
|
|
second_number = 0;
|
|
state = TermState.AfterNumber;
|
|
i -= 1;
|
|
},
|
|
},
|
|
|
|
TermState.AfterNumber => switch (c) {
|
|
';' => state = TermState.Arg,
|
|
else => {
|
|
state = TermState.ExpectEnd;
|
|
i -= 1;
|
|
},
|
|
},
|
|
TermState.Arg => switch (c) {
|
|
'0'...'9' => {
|
|
number_start_index = i;
|
|
state = TermState.ArgNumber;
|
|
},
|
|
else => return error.UnsupportedEscape,
|
|
},
|
|
TermState.ArgNumber => switch (c) {
|
|
'0'...'9' => {},
|
|
else => {
|
|
second_number = std.fmt.parseInt(usize, input[number_start_index..i], 10) catch unreachable;
|
|
state = TermState.ExpectEnd;
|
|
i -= 1;
|
|
},
|
|
},
|
|
TermState.ExpectEnd => switch (c) {
|
|
'm' => {
|
|
state = TermState.Start;
|
|
while (open_span_count != 0) : (open_span_count -= 1) {
|
|
try out.writeAll("</span>");
|
|
}
|
|
if (first_number != 0 or second_number != 0) {
|
|
try out.print("<span class=\"t{d}_{d}\">", .{ first_number, second_number });
|
|
open_span_count += 1;
|
|
}
|
|
},
|
|
else => return error.UnsupportedEscape,
|
|
},
|
|
}
|
|
}
|
|
return buf.toOwnedSlice();
|
|
}
|
|
|
|
const builtin_types = [_][]const u8{
|
|
"f16", "f32", "f64", "f128", "c_longdouble", "c_short",
|
|
"c_ushort", "c_int", "c_uint", "c_long", "c_ulong", "c_longlong",
|
|
"c_ulonglong", "c_char", "c_void", "void", "bool", "isize",
|
|
"usize", "noreturn", "type", "anyerror", "comptime_int", "comptime_float",
|
|
};
|
|
|
|
fn isType(name: []const u8) bool {
|
|
for (builtin_types) |t| {
|
|
if (mem.eql(u8, t, name))
|
|
return true;
|
|
}
|
|
return false;
|
|
}
|
|
|
|
const start_line = "<span class=\"line\">";
|
|
const end_line = "</span>";
|
|
|
|
fn writeEscapedLines(out: anytype, text: []const u8) !void {
|
|
for (text) |char| {
|
|
if (char == '\n') {
|
|
try out.writeAll(end_line);
|
|
try out.writeAll("\n");
|
|
try out.writeAll(start_line);
|
|
} else {
|
|
try writeEscaped(out, &[_]u8{char});
|
|
}
|
|
}
|
|
}
|
|
|
|
fn tokenizeAndPrintRaw(
|
|
allocator: Allocator,
|
|
docgen_tokenizer: *Tokenizer,
|
|
out: anytype,
|
|
source_token: Token,
|
|
raw_src: []const u8,
|
|
) !void {
|
|
const src_non_terminated = mem.trim(u8, raw_src, " \n");
|
|
const src = try allocator.dupeZ(u8, src_non_terminated);
|
|
|
|
try out.writeAll("<code>" ++ start_line);
|
|
var tokenizer = std.zig.Tokenizer.init(src);
|
|
var index: usize = 0;
|
|
var next_tok_is_fn = false;
|
|
while (true) {
|
|
const prev_tok_was_fn = next_tok_is_fn;
|
|
next_tok_is_fn = false;
|
|
|
|
const token = tokenizer.next();
|
|
if (mem.indexOf(u8, src[index..token.loc.start], "//")) |comment_start_off| {
|
|
// render one comment
|
|
const comment_start = index + comment_start_off;
|
|
const comment_end_off = mem.indexOf(u8, src[comment_start..token.loc.start], "\n");
|
|
const comment_end = if (comment_end_off) |o| comment_start + o else token.loc.start;
|
|
|
|
try writeEscapedLines(out, src[index..comment_start]);
|
|
try out.writeAll("<span class=\"tok-comment\">");
|
|
try writeEscaped(out, src[comment_start..comment_end]);
|
|
try out.writeAll("</span>");
|
|
index = comment_end;
|
|
tokenizer.index = index;
|
|
continue;
|
|
}
|
|
|
|
try writeEscapedLines(out, src[index..token.loc.start]);
|
|
switch (token.tag) {
|
|
.eof => break,
|
|
|
|
.keyword_addrspace,
|
|
.keyword_align,
|
|
.keyword_and,
|
|
.keyword_asm,
|
|
.keyword_async,
|
|
.keyword_await,
|
|
.keyword_break,
|
|
.keyword_catch,
|
|
.keyword_comptime,
|
|
.keyword_const,
|
|
.keyword_continue,
|
|
.keyword_defer,
|
|
.keyword_else,
|
|
.keyword_enum,
|
|
.keyword_errdefer,
|
|
.keyword_error,
|
|
.keyword_export,
|
|
.keyword_extern,
|
|
.keyword_for,
|
|
.keyword_if,
|
|
.keyword_inline,
|
|
.keyword_noalias,
|
|
.keyword_noinline,
|
|
.keyword_nosuspend,
|
|
.keyword_opaque,
|
|
.keyword_or,
|
|
.keyword_orelse,
|
|
.keyword_packed,
|
|
.keyword_anyframe,
|
|
.keyword_pub,
|
|
.keyword_resume,
|
|
.keyword_return,
|
|
.keyword_linksection,
|
|
.keyword_callconv,
|
|
.keyword_struct,
|
|
.keyword_suspend,
|
|
.keyword_switch,
|
|
.keyword_test,
|
|
.keyword_threadlocal,
|
|
.keyword_try,
|
|
.keyword_union,
|
|
.keyword_unreachable,
|
|
.keyword_usingnamespace,
|
|
.keyword_var,
|
|
.keyword_volatile,
|
|
.keyword_allowzero,
|
|
.keyword_while,
|
|
.keyword_anytype,
|
|
=> {
|
|
try out.writeAll("<span class=\"tok-kw\">");
|
|
try writeEscaped(out, src[token.loc.start..token.loc.end]);
|
|
try out.writeAll("</span>");
|
|
},
|
|
|
|
.keyword_fn => {
|
|
try out.writeAll("<span class=\"tok-kw\">");
|
|
try writeEscaped(out, src[token.loc.start..token.loc.end]);
|
|
try out.writeAll("</span>");
|
|
next_tok_is_fn = true;
|
|
},
|
|
|
|
.string_literal,
|
|
.char_literal,
|
|
=> {
|
|
try out.writeAll("<span class=\"tok-str\">");
|
|
try writeEscaped(out, src[token.loc.start..token.loc.end]);
|
|
try out.writeAll("</span>");
|
|
},
|
|
|
|
.multiline_string_literal_line => {
|
|
if (src[token.loc.end - 1] == '\n') {
|
|
try out.writeAll("<span class=\"tok-str\">");
|
|
try writeEscaped(out, src[token.loc.start .. token.loc.end - 1]);
|
|
try out.writeAll("</span>" ++ end_line ++ "\n" ++ start_line);
|
|
} else {
|
|
try out.writeAll("<span class=\"tok-str\">");
|
|
try writeEscaped(out, src[token.loc.start..token.loc.end]);
|
|
try out.writeAll("</span>");
|
|
}
|
|
},
|
|
|
|
.builtin => {
|
|
try out.writeAll("<span class=\"tok-builtin\">");
|
|
try writeEscaped(out, src[token.loc.start..token.loc.end]);
|
|
try out.writeAll("</span>");
|
|
},
|
|
|
|
.doc_comment,
|
|
.container_doc_comment,
|
|
=> {
|
|
try out.writeAll("<span class=\"tok-comment\">");
|
|
try writeEscaped(out, src[token.loc.start..token.loc.end]);
|
|
try out.writeAll("</span>");
|
|
},
|
|
|
|
.identifier => {
|
|
const tok_bytes = src[token.loc.start..token.loc.end];
|
|
if (mem.eql(u8, tok_bytes, "undefined") or
|
|
mem.eql(u8, tok_bytes, "null") or
|
|
mem.eql(u8, tok_bytes, "true") or
|
|
mem.eql(u8, tok_bytes, "false"))
|
|
{
|
|
try out.writeAll("<span class=\"tok-null\">");
|
|
try writeEscaped(out, tok_bytes);
|
|
try out.writeAll("</span>");
|
|
} else if (prev_tok_was_fn) {
|
|
try out.writeAll("<span class=\"tok-fn\">");
|
|
try writeEscaped(out, tok_bytes);
|
|
try out.writeAll("</span>");
|
|
} else {
|
|
const is_int = blk: {
|
|
if (src[token.loc.start] != 'i' and src[token.loc.start] != 'u')
|
|
break :blk false;
|
|
var i = token.loc.start + 1;
|
|
if (i == token.loc.end)
|
|
break :blk false;
|
|
while (i != token.loc.end) : (i += 1) {
|
|
if (src[i] < '0' or src[i] > '9')
|
|
break :blk false;
|
|
}
|
|
break :blk true;
|
|
};
|
|
if (is_int or isType(tok_bytes)) {
|
|
try out.writeAll("<span class=\"tok-type\">");
|
|
try writeEscaped(out, tok_bytes);
|
|
try out.writeAll("</span>");
|
|
} else {
|
|
try writeEscaped(out, tok_bytes);
|
|
}
|
|
}
|
|
},
|
|
|
|
.integer_literal,
|
|
.float_literal,
|
|
=> {
|
|
try out.writeAll("<span class=\"tok-number\">");
|
|
try writeEscaped(out, src[token.loc.start..token.loc.end]);
|
|
try out.writeAll("</span>");
|
|
},
|
|
|
|
.bang,
|
|
.pipe,
|
|
.pipe_pipe,
|
|
.pipe_equal,
|
|
.equal,
|
|
.equal_equal,
|
|
.equal_angle_bracket_right,
|
|
.bang_equal,
|
|
.l_paren,
|
|
.r_paren,
|
|
.semicolon,
|
|
.percent,
|
|
.percent_equal,
|
|
.l_brace,
|
|
.r_brace,
|
|
.l_bracket,
|
|
.r_bracket,
|
|
.period,
|
|
.period_asterisk,
|
|
.ellipsis2,
|
|
.ellipsis3,
|
|
.caret,
|
|
.caret_equal,
|
|
.plus,
|
|
.plus_plus,
|
|
.plus_equal,
|
|
.plus_percent,
|
|
.plus_percent_equal,
|
|
.plus_pipe,
|
|
.plus_pipe_equal,
|
|
.minus,
|
|
.minus_equal,
|
|
.minus_percent,
|
|
.minus_percent_equal,
|
|
.minus_pipe,
|
|
.minus_pipe_equal,
|
|
.asterisk,
|
|
.asterisk_equal,
|
|
.asterisk_asterisk,
|
|
.asterisk_percent,
|
|
.asterisk_percent_equal,
|
|
.asterisk_pipe,
|
|
.asterisk_pipe_equal,
|
|
.arrow,
|
|
.colon,
|
|
.slash,
|
|
.slash_equal,
|
|
.comma,
|
|
.ampersand,
|
|
.ampersand_equal,
|
|
.question_mark,
|
|
.angle_bracket_left,
|
|
.angle_bracket_left_equal,
|
|
.angle_bracket_angle_bracket_left,
|
|
.angle_bracket_angle_bracket_left_equal,
|
|
.angle_bracket_angle_bracket_left_pipe,
|
|
.angle_bracket_angle_bracket_left_pipe_equal,
|
|
.angle_bracket_right,
|
|
.angle_bracket_right_equal,
|
|
.angle_bracket_angle_bracket_right,
|
|
.angle_bracket_angle_bracket_right_equal,
|
|
.tilde,
|
|
=> try writeEscaped(out, src[token.loc.start..token.loc.end]),
|
|
|
|
.invalid, .invalid_periodasterisks => return parseError(
|
|
docgen_tokenizer,
|
|
source_token,
|
|
"syntax error",
|
|
.{},
|
|
),
|
|
}
|
|
index = token.loc.end;
|
|
}
|
|
try out.writeAll(end_line ++ "</code>");
|
|
}
|
|
|
|
fn tokenizeAndPrint(
|
|
allocator: Allocator,
|
|
docgen_tokenizer: *Tokenizer,
|
|
out: anytype,
|
|
source_token: Token,
|
|
) !void {
|
|
const raw_src = docgen_tokenizer.buffer[source_token.start..source_token.end];
|
|
return tokenizeAndPrintRaw(allocator, docgen_tokenizer, out, source_token, raw_src);
|
|
}
|
|
|
|
fn printSourceBlock(allocator: Allocator, docgen_tokenizer: *Tokenizer, out: anytype, syntax_block: SyntaxBlock) !void {
|
|
const source_type = @tagName(syntax_block.source_type);
|
|
|
|
try out.print("<figure><figcaption class=\"{s}-cap\"><cite class=\"file\">{s}</cite></figcaption><pre>", .{ source_type, syntax_block.name });
|
|
switch (syntax_block.source_type) {
|
|
.zig => try tokenizeAndPrint(allocator, docgen_tokenizer, out, syntax_block.source_token),
|
|
else => {
|
|
const raw_source = docgen_tokenizer.buffer[syntax_block.source_token.start..syntax_block.source_token.end];
|
|
const trimmed_raw_source = mem.trim(u8, raw_source, " \n");
|
|
|
|
try out.writeAll("<code>" ++ start_line);
|
|
try writeEscapedLines(out, trimmed_raw_source);
|
|
try out.writeAll(end_line ++ "</code>");
|
|
},
|
|
}
|
|
try out.writeAll("</pre></figure>");
|
|
}
|
|
|
|
fn printShell(out: anytype, shell_content: []const u8) !void {
|
|
const trimmed_shell_content = mem.trim(u8, shell_content, " \n");
|
|
try out.writeAll("<figure><figcaption class=\"shell-cap\">Shell</figcaption><pre><samp>");
|
|
var cmd_cont: bool = false;
|
|
var iter = std.mem.split(u8, trimmed_shell_content, "\n");
|
|
while (iter.next()) |orig_line| {
|
|
const line = mem.trimRight(u8, orig_line, " ");
|
|
if (!cmd_cont and line.len > 1 and mem.eql(u8, line[0..2], "$ ") and line[line.len - 1] != '\\') {
|
|
try out.print(start_line ++ "$ <kbd>{s}</kbd>" ++ end_line ++ "\n", .{std.mem.trimLeft(u8, line[1..], " ")});
|
|
} else if (!cmd_cont and line.len > 1 and mem.eql(u8, line[0..2], "$ ") and line[line.len - 1] == '\\') {
|
|
try out.print(start_line ++ "$ <kbd>{s}" ++ end_line ++ "\n", .{std.mem.trimLeft(u8, line[1..], " ")});
|
|
cmd_cont = true;
|
|
} else if (line.len > 0 and line[line.len - 1] != '\\' and cmd_cont) {
|
|
try out.print(start_line ++ "{s}</kbd>" ++ end_line ++ "\n", .{line});
|
|
cmd_cont = false;
|
|
} else {
|
|
try out.print(start_line ++ "{s}" ++ end_line ++ "\n", .{line});
|
|
}
|
|
}
|
|
|
|
try out.writeAll("</samp></pre></figure>");
|
|
}
|
|
|
|
fn genHtml(
|
|
allocator: Allocator,
|
|
tokenizer: *Tokenizer,
|
|
toc: *Toc,
|
|
out: anytype,
|
|
zig_exe: []const u8,
|
|
do_code_tests: bool,
|
|
) !void {
|
|
var progress = Progress{};
|
|
const root_node = try progress.start("Generating docgen examples", toc.nodes.len);
|
|
defer root_node.end();
|
|
|
|
var env_map = try process.getEnvMap(allocator);
|
|
try env_map.put("ZIG_DEBUG_COLOR", "1");
|
|
|
|
const host = try std.zig.system.NativeTargetInfo.detect(allocator, .{});
|
|
const builtin_code = try getBuiltinCode(allocator, &env_map, zig_exe);
|
|
|
|
for (toc.nodes) |node| {
|
|
defer root_node.completeOne();
|
|
switch (node) {
|
|
.Content => |data| {
|
|
try out.writeAll(data);
|
|
},
|
|
.Link => |info| {
|
|
if (!toc.urls.contains(info.url)) {
|
|
return parseError(tokenizer, info.token, "url not found: {s}", .{info.url});
|
|
}
|
|
try out.print("<a href=\"#{s}\">{s}</a>", .{ info.url, info.name });
|
|
},
|
|
.Nav => {
|
|
try out.writeAll(toc.toc);
|
|
},
|
|
.Builtin => |tok| {
|
|
try out.writeAll("<figure><figcaption class=\"zig-cap\"><cite>@import(\"builtin\")</cite></figcaption><pre>");
|
|
try tokenizeAndPrintRaw(allocator, tokenizer, out, tok, builtin_code);
|
|
try out.writeAll("</pre></figure>");
|
|
},
|
|
.HeaderOpen => |info| {
|
|
try out.print(
|
|
"<h{d} id=\"{s}\"><a href=\"#toc-{s}\">{s}</a> <a class=\"hdr\" href=\"#{s}\">§</a></h{d}>\n",
|
|
.{ info.n, info.url, info.url, info.name, info.url, info.n },
|
|
);
|
|
},
|
|
.SeeAlso => |items| {
|
|
try out.writeAll("<p>See also:</p><ul>\n");
|
|
for (items) |item| {
|
|
const url = try urlize(allocator, item.name);
|
|
if (!toc.urls.contains(url)) {
|
|
return parseError(tokenizer, item.token, "url not found: {s}", .{url});
|
|
}
|
|
try out.print("<li><a href=\"#{s}\">{s}</a></li>\n", .{ url, item.name });
|
|
}
|
|
try out.writeAll("</ul>\n");
|
|
},
|
|
.InlineSyntax => |content_tok| {
|
|
try tokenizeAndPrint(allocator, tokenizer, out, content_tok);
|
|
},
|
|
.Shell => |content_tok| {
|
|
const raw_shell_content = tokenizer.buffer[content_tok.start..content_tok.end];
|
|
try printShell(out, raw_shell_content);
|
|
},
|
|
.SyntaxBlock => |syntax_block| {
|
|
try printSourceBlock(allocator, tokenizer, out, syntax_block);
|
|
},
|
|
.Code => |code| {
|
|
const name_plus_ext = try std.fmt.allocPrint(allocator, "{s}.zig", .{code.name});
|
|
const syntax_block = SyntaxBlock{
|
|
.source_type = .zig,
|
|
.name = name_plus_ext,
|
|
.source_token = code.source_token,
|
|
};
|
|
|
|
try printSourceBlock(allocator, tokenizer, out, syntax_block);
|
|
|
|
if (!do_code_tests) {
|
|
continue;
|
|
}
|
|
|
|
const raw_source = tokenizer.buffer[code.source_token.start..code.source_token.end];
|
|
const trimmed_raw_source = mem.trim(u8, raw_source, " \n");
|
|
const tmp_source_file_name = try fs.path.join(
|
|
allocator,
|
|
&[_][]const u8{ tmp_dir_name, name_plus_ext },
|
|
);
|
|
try fs.cwd().writeFile(tmp_source_file_name, trimmed_raw_source);
|
|
|
|
var shell_buffer = std.ArrayList(u8).init(allocator);
|
|
defer shell_buffer.deinit();
|
|
var shell_out = shell_buffer.writer();
|
|
|
|
switch (code.id) {
|
|
Code.Id.Exe => |expected_outcome| code_block: {
|
|
var build_args = std.ArrayList([]const u8).init(allocator);
|
|
defer build_args.deinit();
|
|
try build_args.appendSlice(&[_][]const u8{
|
|
zig_exe, "build-exe",
|
|
"--name", code.name,
|
|
"--color", "on",
|
|
"--enable-cache", tmp_source_file_name,
|
|
});
|
|
|
|
try shell_out.print("$ zig build-exe {s} ", .{name_plus_ext});
|
|
|
|
switch (code.mode) {
|
|
.Debug => {},
|
|
else => {
|
|
try build_args.appendSlice(&[_][]const u8{ "-O", @tagName(code.mode) });
|
|
try shell_out.print("-O {s} ", .{@tagName(code.mode)});
|
|
},
|
|
}
|
|
for (code.link_objects) |link_object| {
|
|
const name_with_ext = try std.fmt.allocPrint(allocator, "{s}{s}", .{ link_object, obj_ext });
|
|
const full_path_object = try fs.path.join(
|
|
allocator,
|
|
&[_][]const u8{ tmp_dir_name, name_with_ext },
|
|
);
|
|
try build_args.append(full_path_object);
|
|
try shell_out.print("{s} ", .{name_with_ext});
|
|
}
|
|
if (code.link_libc) {
|
|
try build_args.append("-lc");
|
|
try shell_out.print("-lc ", .{});
|
|
}
|
|
const target = try std.zig.CrossTarget.parse(.{
|
|
.arch_os_abi = code.target_str orelse "native",
|
|
});
|
|
if (code.target_str) |triple| {
|
|
try build_args.appendSlice(&[_][]const u8{ "-target", triple });
|
|
try shell_out.print("-target {s} ", .{triple});
|
|
}
|
|
if (code.verbose_cimport) {
|
|
try build_args.append("--verbose-cimport");
|
|
try shell_out.print("--verbose-cimport ", .{});
|
|
}
|
|
|
|
try shell_out.print("\n", .{});
|
|
|
|
if (expected_outcome == .BuildFail) {
|
|
const result = try ChildProcess.exec(.{
|
|
.allocator = allocator,
|
|
.argv = build_args.items,
|
|
.env_map = &env_map,
|
|
.max_output_bytes = max_doc_file_size,
|
|
});
|
|
switch (result.term) {
|
|
.Exited => |exit_code| {
|
|
if (exit_code == 0) {
|
|
progress.log("", .{});
|
|
print("{s}\nThe following command incorrectly succeeded:\n", .{result.stderr});
|
|
dumpArgs(build_args.items);
|
|
return parseError(tokenizer, code.source_token, "example incorrectly compiled", .{});
|
|
}
|
|
},
|
|
else => {
|
|
progress.log("", .{});
|
|
print("{s}\nThe following command crashed:\n", .{result.stderr});
|
|
dumpArgs(build_args.items);
|
|
return parseError(tokenizer, code.source_token, "example compile crashed", .{});
|
|
},
|
|
}
|
|
const escaped_stderr = try escapeHtml(allocator, result.stderr);
|
|
const colored_stderr = try termColor(allocator, escaped_stderr);
|
|
try shell_out.writeAll(colored_stderr);
|
|
break :code_block;
|
|
}
|
|
const exec_result = exec(allocator, &env_map, build_args.items) catch
|
|
return parseError(tokenizer, code.source_token, "example failed to compile", .{});
|
|
|
|
if (code.verbose_cimport) {
|
|
const escaped_build_stderr = try escapeHtml(allocator, exec_result.stderr);
|
|
try shell_out.writeAll(escaped_build_stderr);
|
|
}
|
|
|
|
if (code.target_str) |triple| {
|
|
if (mem.startsWith(u8, triple, "wasm32") or
|
|
mem.startsWith(u8, triple, "riscv64-linux") or
|
|
(mem.startsWith(u8, triple, "x86_64-linux") and
|
|
builtin.os.tag != .linux or builtin.cpu.arch != .x86_64))
|
|
{
|
|
// skip execution
|
|
break :code_block;
|
|
}
|
|
}
|
|
|
|
const path_to_exe_dir = mem.trim(u8, exec_result.stdout, " \r\n");
|
|
const path_to_exe_basename = try std.fmt.allocPrint(allocator, "{s}{s}", .{
|
|
code.name,
|
|
target.exeFileExt(),
|
|
});
|
|
const path_to_exe = try fs.path.join(allocator, &[_][]const u8{
|
|
path_to_exe_dir,
|
|
path_to_exe_basename,
|
|
});
|
|
const run_args = &[_][]const u8{path_to_exe};
|
|
|
|
var exited_with_signal = false;
|
|
|
|
const result = if (expected_outcome == ExpectedOutcome.Fail) blk: {
|
|
const result = try ChildProcess.exec(.{
|
|
.allocator = allocator,
|
|
.argv = run_args,
|
|
.env_map = &env_map,
|
|
.max_output_bytes = max_doc_file_size,
|
|
});
|
|
switch (result.term) {
|
|
.Exited => |exit_code| {
|
|
if (exit_code == 0) {
|
|
progress.log("", .{});
|
|
print("{s}\nThe following command incorrectly succeeded:\n", .{result.stderr});
|
|
dumpArgs(run_args);
|
|
return parseError(tokenizer, code.source_token, "example incorrectly compiled", .{});
|
|
}
|
|
},
|
|
.Signal => exited_with_signal = true,
|
|
else => {},
|
|
}
|
|
break :blk result;
|
|
} else blk: {
|
|
break :blk exec(allocator, &env_map, run_args) catch return parseError(tokenizer, code.source_token, "example crashed", .{});
|
|
};
|
|
|
|
const escaped_stderr = try escapeHtml(allocator, result.stderr);
|
|
const escaped_stdout = try escapeHtml(allocator, result.stdout);
|
|
|
|
const colored_stderr = try termColor(allocator, escaped_stderr);
|
|
const colored_stdout = try termColor(allocator, escaped_stdout);
|
|
|
|
try shell_out.print("\n$ ./{s}\n{s}{s}", .{ code.name, colored_stdout, colored_stderr });
|
|
if (exited_with_signal) {
|
|
try shell_out.print("(process terminated by signal)", .{});
|
|
}
|
|
try shell_out.writeAll("\n");
|
|
},
|
|
Code.Id.Test => {
|
|
var test_args = std.ArrayList([]const u8).init(allocator);
|
|
defer test_args.deinit();
|
|
|
|
try test_args.appendSlice(&[_][]const u8{
|
|
zig_exe, "test", tmp_source_file_name,
|
|
});
|
|
try shell_out.print("$ zig test {s}.zig ", .{code.name});
|
|
|
|
switch (code.mode) {
|
|
.Debug => {},
|
|
else => {
|
|
try test_args.appendSlice(&[_][]const u8{
|
|
"-O", @tagName(code.mode),
|
|
});
|
|
try shell_out.print("-O {s} ", .{@tagName(code.mode)});
|
|
},
|
|
}
|
|
if (code.link_libc) {
|
|
try test_args.append("-lc");
|
|
try shell_out.print("-lc ", .{});
|
|
}
|
|
if (code.target_str) |triple| {
|
|
try test_args.appendSlice(&[_][]const u8{ "-target", triple });
|
|
try shell_out.print("-target {s} ", .{triple});
|
|
|
|
const cross_target = try std.zig.CrossTarget.parse(.{
|
|
.arch_os_abi = triple,
|
|
});
|
|
const target_info = try std.zig.system.NativeTargetInfo.detect(
|
|
allocator,
|
|
cross_target,
|
|
);
|
|
switch (host.getExternalExecutor(target_info, .{
|
|
.link_libc = code.link_libc,
|
|
})) {
|
|
.native => {},
|
|
else => {
|
|
try test_args.appendSlice(&[_][]const u8{"--test-no-exec"});
|
|
try shell_out.writeAll("--test-no-exec");
|
|
},
|
|
}
|
|
}
|
|
const result = exec(allocator, &env_map, test_args.items) catch
|
|
return parseError(tokenizer, code.source_token, "test failed", .{});
|
|
const escaped_stderr = try escapeHtml(allocator, result.stderr);
|
|
const escaped_stdout = try escapeHtml(allocator, result.stdout);
|
|
try shell_out.print("\n{s}{s}\n", .{ escaped_stderr, escaped_stdout });
|
|
},
|
|
Code.Id.TestError => |error_match| {
|
|
var test_args = std.ArrayList([]const u8).init(allocator);
|
|
defer test_args.deinit();
|
|
|
|
try test_args.appendSlice(&[_][]const u8{
|
|
zig_exe,
|
|
"test",
|
|
"--color",
|
|
"on",
|
|
tmp_source_file_name,
|
|
});
|
|
try shell_out.print("$ zig test {s}.zig ", .{code.name});
|
|
|
|
switch (code.mode) {
|
|
.Debug => {},
|
|
else => {
|
|
try test_args.appendSlice(&[_][]const u8{ "-O", @tagName(code.mode) });
|
|
try shell_out.print("-O {s} ", .{@tagName(code.mode)});
|
|
},
|
|
}
|
|
const result = try ChildProcess.exec(.{
|
|
.allocator = allocator,
|
|
.argv = test_args.items,
|
|
.env_map = &env_map,
|
|
.max_output_bytes = max_doc_file_size,
|
|
});
|
|
switch (result.term) {
|
|
.Exited => |exit_code| {
|
|
if (exit_code == 0) {
|
|
progress.log("", .{});
|
|
print("{s}\nThe following command incorrectly succeeded:\n", .{result.stderr});
|
|
dumpArgs(test_args.items);
|
|
return parseError(tokenizer, code.source_token, "example incorrectly compiled", .{});
|
|
}
|
|
},
|
|
else => {
|
|
progress.log("", .{});
|
|
print("{s}\nThe following command crashed:\n", .{result.stderr});
|
|
dumpArgs(test_args.items);
|
|
return parseError(tokenizer, code.source_token, "example compile crashed", .{});
|
|
},
|
|
}
|
|
if (mem.indexOf(u8, result.stderr, error_match) == null) {
|
|
progress.log("", .{});
|
|
print("{s}\nExpected to find '{s}' in stderr\n", .{ result.stderr, error_match });
|
|
return parseError(tokenizer, code.source_token, "example did not have expected compile error", .{});
|
|
}
|
|
const escaped_stderr = try escapeHtml(allocator, result.stderr);
|
|
const colored_stderr = try termColor(allocator, escaped_stderr);
|
|
try shell_out.print("\n{s}\n", .{colored_stderr});
|
|
},
|
|
|
|
Code.Id.TestSafety => |error_match| {
|
|
var test_args = std.ArrayList([]const u8).init(allocator);
|
|
defer test_args.deinit();
|
|
|
|
try test_args.appendSlice(&[_][]const u8{
|
|
zig_exe, "test", tmp_source_file_name,
|
|
});
|
|
var mode_arg: []const u8 = "";
|
|
switch (code.mode) {
|
|
.Debug => {},
|
|
.ReleaseSafe => {
|
|
try test_args.append("-OReleaseSafe");
|
|
mode_arg = "-OReleaseSafe";
|
|
},
|
|
.ReleaseFast => {
|
|
try test_args.append("-OReleaseFast");
|
|
mode_arg = "-OReleaseFast";
|
|
},
|
|
.ReleaseSmall => {
|
|
try test_args.append("-OReleaseSmall");
|
|
mode_arg = "-OReleaseSmall";
|
|
},
|
|
}
|
|
|
|
const result = try ChildProcess.exec(.{
|
|
.allocator = allocator,
|
|
.argv = test_args.items,
|
|
.env_map = &env_map,
|
|
.max_output_bytes = max_doc_file_size,
|
|
});
|
|
switch (result.term) {
|
|
.Exited => |exit_code| {
|
|
if (exit_code == 0) {
|
|
progress.log("", .{});
|
|
print("{s}\nThe following command incorrectly succeeded:\n", .{result.stderr});
|
|
dumpArgs(test_args.items);
|
|
return parseError(tokenizer, code.source_token, "example test incorrectly succeeded", .{});
|
|
}
|
|
},
|
|
else => {
|
|
progress.log("", .{});
|
|
print("{s}\nThe following command crashed:\n", .{result.stderr});
|
|
dumpArgs(test_args.items);
|
|
return parseError(tokenizer, code.source_token, "example compile crashed", .{});
|
|
},
|
|
}
|
|
if (mem.indexOf(u8, result.stderr, error_match) == null) {
|
|
progress.log("", .{});
|
|
print("{s}\nExpected to find '{s}' in stderr\n", .{ result.stderr, error_match });
|
|
return parseError(tokenizer, code.source_token, "example did not have expected runtime safety error message", .{});
|
|
}
|
|
const escaped_stderr = try escapeHtml(allocator, result.stderr);
|
|
const colored_stderr = try termColor(allocator, escaped_stderr);
|
|
try shell_out.print("$ zig test {s}.zig {s}\n{s}\n", .{
|
|
code.name,
|
|
mode_arg,
|
|
colored_stderr,
|
|
});
|
|
},
|
|
Code.Id.Obj => |maybe_error_match| {
|
|
const name_plus_obj_ext = try std.fmt.allocPrint(allocator, "{s}{s}", .{ code.name, obj_ext });
|
|
var build_args = std.ArrayList([]const u8).init(allocator);
|
|
defer build_args.deinit();
|
|
|
|
try build_args.appendSlice(&[_][]const u8{
|
|
zig_exe,
|
|
"build-obj",
|
|
tmp_source_file_name,
|
|
"--color",
|
|
"on",
|
|
"--name",
|
|
code.name,
|
|
try std.fmt.allocPrint(allocator, "-femit-bin={s}{c}{s}", .{
|
|
tmp_dir_name, fs.path.sep, name_plus_obj_ext,
|
|
}),
|
|
});
|
|
|
|
try shell_out.print("$ zig build-obj {s}.zig ", .{code.name});
|
|
|
|
switch (code.mode) {
|
|
.Debug => {},
|
|
else => {
|
|
try build_args.appendSlice(&[_][]const u8{ "-O", @tagName(code.mode) });
|
|
try shell_out.print("-O {s} ", .{@tagName(code.mode)});
|
|
},
|
|
}
|
|
|
|
if (code.target_str) |triple| {
|
|
try build_args.appendSlice(&[_][]const u8{ "-target", triple });
|
|
try shell_out.print("-target {s} ", .{triple});
|
|
}
|
|
|
|
if (maybe_error_match) |error_match| {
|
|
const result = try ChildProcess.exec(.{
|
|
.allocator = allocator,
|
|
.argv = build_args.items,
|
|
.env_map = &env_map,
|
|
.max_output_bytes = max_doc_file_size,
|
|
});
|
|
switch (result.term) {
|
|
.Exited => |exit_code| {
|
|
if (exit_code == 0) {
|
|
progress.log("", .{});
|
|
print("{s}\nThe following command incorrectly succeeded:\n", .{result.stderr});
|
|
dumpArgs(build_args.items);
|
|
return parseError(tokenizer, code.source_token, "example build incorrectly succeeded", .{});
|
|
}
|
|
},
|
|
else => {
|
|
progress.log("", .{});
|
|
print("{s}\nThe following command crashed:\n", .{result.stderr});
|
|
dumpArgs(build_args.items);
|
|
return parseError(tokenizer, code.source_token, "example compile crashed", .{});
|
|
},
|
|
}
|
|
if (mem.indexOf(u8, result.stderr, error_match) == null) {
|
|
progress.log("", .{});
|
|
print("{s}\nExpected to find '{s}' in stderr\n", .{ result.stderr, error_match });
|
|
return parseError(tokenizer, code.source_token, "example did not have expected compile error message", .{});
|
|
}
|
|
const escaped_stderr = try escapeHtml(allocator, result.stderr);
|
|
const colored_stderr = try termColor(allocator, escaped_stderr);
|
|
try shell_out.print("\n{s} ", .{colored_stderr});
|
|
} else {
|
|
_ = exec(allocator, &env_map, build_args.items) catch return parseError(tokenizer, code.source_token, "example failed to compile", .{});
|
|
}
|
|
try shell_out.writeAll("\n");
|
|
},
|
|
Code.Id.Lib => {
|
|
const bin_basename = try std.zig.binNameAlloc(allocator, .{
|
|
.root_name = code.name,
|
|
.target = builtin.target,
|
|
.output_mode = .Lib,
|
|
});
|
|
|
|
var test_args = std.ArrayList([]const u8).init(allocator);
|
|
defer test_args.deinit();
|
|
|
|
try test_args.appendSlice(&[_][]const u8{
|
|
zig_exe,
|
|
"build-lib",
|
|
tmp_source_file_name,
|
|
try std.fmt.allocPrint(allocator, "-femit-bin={s}{s}{s}", .{
|
|
tmp_dir_name, fs.path.sep_str, bin_basename,
|
|
}),
|
|
});
|
|
try shell_out.print("$ zig build-lib {s}.zig ", .{code.name});
|
|
|
|
switch (code.mode) {
|
|
.Debug => {},
|
|
else => {
|
|
try test_args.appendSlice(&[_][]const u8{ "-O", @tagName(code.mode) });
|
|
try shell_out.print("-O {s} ", .{@tagName(code.mode)});
|
|
},
|
|
}
|
|
if (code.target_str) |triple| {
|
|
try test_args.appendSlice(&[_][]const u8{ "-target", triple });
|
|
try shell_out.print("-target {s} ", .{triple});
|
|
}
|
|
if (code.link_mode) |link_mode| {
|
|
switch (link_mode) {
|
|
.Static => {
|
|
try test_args.append("-static");
|
|
try shell_out.print("-static ", .{});
|
|
},
|
|
.Dynamic => {
|
|
try test_args.append("-dynamic");
|
|
try shell_out.print("-dynamic ", .{});
|
|
},
|
|
}
|
|
}
|
|
const result = exec(allocator, &env_map, test_args.items) catch return parseError(tokenizer, code.source_token, "test failed", .{});
|
|
const escaped_stderr = try escapeHtml(allocator, result.stderr);
|
|
const escaped_stdout = try escapeHtml(allocator, result.stdout);
|
|
try shell_out.print("\n{s}{s}\n", .{ escaped_stderr, escaped_stdout });
|
|
},
|
|
}
|
|
|
|
if (!code.just_check_syntax) {
|
|
try printShell(out, shell_buffer.items);
|
|
}
|
|
},
|
|
}
|
|
}
|
|
}
|
|
|
|
fn exec(allocator: Allocator, env_map: *std.BufMap, args: []const []const u8) !ChildProcess.ExecResult {
|
|
const result = try ChildProcess.exec(.{
|
|
.allocator = allocator,
|
|
.argv = args,
|
|
.env_map = env_map,
|
|
.max_output_bytes = max_doc_file_size,
|
|
});
|
|
switch (result.term) {
|
|
.Exited => |exit_code| {
|
|
if (exit_code != 0) {
|
|
print("{s}\nThe following command exited with code {}:\n", .{ result.stderr, exit_code });
|
|
dumpArgs(args);
|
|
return error.ChildExitError;
|
|
}
|
|
},
|
|
else => {
|
|
print("{s}\nThe following command crashed:\n", .{result.stderr});
|
|
dumpArgs(args);
|
|
return error.ChildCrashed;
|
|
},
|
|
}
|
|
return result;
|
|
}
|
|
|
|
fn getBuiltinCode(allocator: Allocator, env_map: *std.BufMap, zig_exe: []const u8) ![]const u8 {
|
|
const result = try exec(allocator, env_map, &[_][]const u8{ zig_exe, "build-obj", "--show-builtin" });
|
|
return result.stdout;
|
|
}
|
|
|
|
fn dumpArgs(args: []const []const u8) void {
|
|
for (args) |arg|
|
|
print("{s} ", .{arg})
|
|
else
|
|
print("\n", .{});
|
|
}
|
|
|
|
test "shell parsed" {
|
|
const test_allocator = std.testing.allocator;
|
|
|
|
{
|
|
const shell_out =
|
|
\\$ zig build test.zig
|
|
;
|
|
const expected =
|
|
\\<figure><figcaption class="shell-cap">Shell</figcaption><pre><samp><span class="line">$ <kbd>zig build test.zig</kbd></span>
|
|
\\</samp></pre></figure>
|
|
;
|
|
|
|
var buffer = std.ArrayList(u8).init(test_allocator);
|
|
defer buffer.deinit();
|
|
|
|
try printShell(buffer.writer(), shell_out);
|
|
std.log.emerg("{s}", .{buffer.items});
|
|
try testing.expectEqualSlices(u8, expected, buffer.items);
|
|
}
|
|
{
|
|
const shell_out =
|
|
\\$ zig build test.zig
|
|
\\build output
|
|
;
|
|
const expected =
|
|
\\<figure><figcaption class="shell-cap">Shell</figcaption><pre><samp><span class="line">$ <kbd>zig build test.zig</kbd></span>
|
|
\\<span class="line">build output</span>
|
|
\\</samp></pre></figure>
|
|
;
|
|
|
|
var buffer = std.ArrayList(u8).init(test_allocator);
|
|
defer buffer.deinit();
|
|
|
|
try printShell(buffer.writer(), shell_out);
|
|
try testing.expectEqualSlices(u8, expected, buffer.items);
|
|
}
|
|
{
|
|
const shell_out =
|
|
\\$ zig build test.zig
|
|
\\build output
|
|
\\$ ./test
|
|
;
|
|
const expected =
|
|
\\<figure><figcaption class="shell-cap">Shell</figcaption><pre><samp><span class="line">$ <kbd>zig build test.zig</kbd></span>
|
|
\\<span class="line">build output</span>
|
|
\\<span class="line">$ <kbd>./test</kbd></span>
|
|
\\</samp></pre></figure>
|
|
;
|
|
|
|
var buffer = std.ArrayList(u8).init(test_allocator);
|
|
defer buffer.deinit();
|
|
|
|
try printShell(buffer.writer(), shell_out);
|
|
try testing.expectEqualSlices(u8, expected, buffer.items);
|
|
}
|
|
{
|
|
const shell_out =
|
|
\\$ zig build test.zig
|
|
\\
|
|
\\$ ./test
|
|
\\output
|
|
;
|
|
const expected =
|
|
\\<figure><figcaption class="shell-cap">Shell</figcaption><pre><samp><span class="line">$ <kbd>zig build test.zig</kbd></span>
|
|
\\<span class="line"></span>
|
|
\\<span class="line">$ <kbd>./test</kbd></span>
|
|
\\<span class="line">output</span>
|
|
\\</samp></pre></figure>
|
|
;
|
|
|
|
var buffer = std.ArrayList(u8).init(test_allocator);
|
|
defer buffer.deinit();
|
|
|
|
try printShell(buffer.writer(), shell_out);
|
|
try testing.expectEqualSlices(u8, expected, buffer.items);
|
|
}
|
|
{
|
|
const shell_out =
|
|
\\$ zig build test.zig
|
|
\\$ ./test
|
|
\\output
|
|
;
|
|
const expected =
|
|
\\<figure><figcaption class="shell-cap">Shell</figcaption><pre><samp><span class="line">$ <kbd>zig build test.zig</kbd></span>
|
|
\\<span class="line">$ <kbd>./test</kbd></span>
|
|
\\<span class="line">output</span>
|
|
\\</samp></pre></figure>
|
|
;
|
|
|
|
var buffer = std.ArrayList(u8).init(test_allocator);
|
|
defer buffer.deinit();
|
|
|
|
try printShell(buffer.writer(), shell_out);
|
|
try testing.expectEqualSlices(u8, expected, buffer.items);
|
|
}
|
|
{
|
|
const shell_out =
|
|
\\$ zig build test.zig \
|
|
\\ --build-option
|
|
\\build output
|
|
\\$ ./test
|
|
\\output
|
|
;
|
|
const expected =
|
|
\\<figure><figcaption class="shell-cap">Shell</figcaption><pre><samp><span class="line">$ <kbd>zig build test.zig \</span>
|
|
\\<span class="line"> --build-option</kbd></span>
|
|
\\<span class="line">build output</span>
|
|
\\<span class="line">$ <kbd>./test</kbd></span>
|
|
\\<span class="line">output</span>
|
|
\\</samp></pre></figure>
|
|
;
|
|
|
|
var buffer = std.ArrayList(u8).init(test_allocator);
|
|
defer buffer.deinit();
|
|
|
|
try printShell(buffer.writer(), shell_out);
|
|
try testing.expectEqualSlices(u8, expected, buffer.items);
|
|
}
|
|
{
|
|
// intentional space after "--build-option1 \"
|
|
const shell_out =
|
|
\\$ zig build test.zig \
|
|
\\ --build-option1 \
|
|
\\ --build-option2
|
|
\\$ ./test
|
|
;
|
|
const expected =
|
|
\\<figure><figcaption class="shell-cap">Shell</figcaption><pre><samp><span class="line">$ <kbd>zig build test.zig \</span>
|
|
\\<span class="line"> --build-option1 \</span>
|
|
\\<span class="line"> --build-option2</kbd></span>
|
|
\\<span class="line">$ <kbd>./test</kbd></span>
|
|
\\</samp></pre></figure>
|
|
;
|
|
|
|
var buffer = std.ArrayList(u8).init(test_allocator);
|
|
defer buffer.deinit();
|
|
|
|
try printShell(buffer.writer(), shell_out);
|
|
try testing.expectEqualSlices(u8, expected, buffer.items);
|
|
}
|
|
{
|
|
const shell_out =
|
|
\\$ zig build test.zig \
|
|
\\$ ./test
|
|
;
|
|
const expected =
|
|
\\<figure><figcaption class="shell-cap">Shell</figcaption><pre><samp><span class="line">$ <kbd>zig build test.zig \</span>
|
|
\\<span class="line">$ ./test</kbd></span>
|
|
\\</samp></pre></figure>
|
|
;
|
|
|
|
var buffer = std.ArrayList(u8).init(test_allocator);
|
|
defer buffer.deinit();
|
|
|
|
try printShell(buffer.writer(), shell_out);
|
|
try testing.expectEqualSlices(u8, expected, buffer.items);
|
|
}
|
|
{
|
|
const shell_out =
|
|
\\$ zig build test.zig
|
|
\\$ ./test
|
|
\\$1
|
|
;
|
|
const expected =
|
|
\\<figure><figcaption class="shell-cap">Shell</figcaption><pre><samp><span class="line">$ <kbd>zig build test.zig</kbd></span>
|
|
\\<span class="line">$ <kbd>./test</kbd></span>
|
|
\\<span class="line">$1</span>
|
|
\\</samp></pre></figure>
|
|
;
|
|
|
|
var buffer = std.ArrayList(u8).init(test_allocator);
|
|
defer buffer.deinit();
|
|
|
|
try printShell(buffer.writer(), shell_out);
|
|
try testing.expectEqualSlices(u8, expected, buffer.items);
|
|
}
|
|
{
|
|
const shell_out =
|
|
\\$zig build test.zig
|
|
;
|
|
const expected =
|
|
\\<figure><figcaption class="shell-cap">Shell</figcaption><pre><samp><span class="line">$zig build test.zig</span>
|
|
\\</samp></pre></figure>
|
|
;
|
|
|
|
var buffer = std.ArrayList(u8).init(test_allocator);
|
|
defer buffer.deinit();
|
|
|
|
try printShell(buffer.writer(), shell_out);
|
|
try testing.expectEqualSlices(u8, expected, buffer.items);
|
|
}
|
|
}
|