1885 lines
82 KiB
Python
1885 lines
82 KiB
Python
import t, c
|
||
from stdint import *
|
||
import memhub
|
||
import string
|
||
import stdio
|
||
import stdlib
|
||
import llvmlite
|
||
import w32.win32file as win32file
|
||
import w32.win32base as win32base
|
||
import w32.fileio as fileio
|
||
import viperlib
|
||
import sys
|
||
import ast
|
||
import lib.core.VLogger as VLogger
|
||
import lib.core.IncludesScanner as IncludesScanner
|
||
import lib.core.Handles.HandlesStruct as HandlesStruct
|
||
import lib.core.Handles.HandlesImports as HandlesImports
|
||
import lib.core.Handles.HandlesTranslator as HandlesTranslator
|
||
import lib.core.BuildPipeline as BuildPipeline
|
||
import lib.Projectrans.Config as Config
|
||
|
||
|
||
# ============================================================
|
||
# StubMerger - stub 声明合并模块
|
||
#
|
||
# 从 temp 目录加载所有 .stub.ll 文件，将声明内容合并到主 IR。
|
||
# 这是 Phase2 的核心功能：翻译主文件时，将所有依赖的 stub
|
||
# 声明合并到主 IR 前面，使链接器能解析跨模块符号引用。
|
||
#
|
||
# 合并策略：
|
||
# 1. 扫描 temp 目录，收集所有 {sha1}.stub.ll 文件
|
||
# 2. 读取每个 stub 文件内容
|
||
# 3. 过滤去重（避免重复声明）
|
||
# 4. 将声明插入到主 IR 的 target triple 之后
|
||
# ============================================================
|
||
|
||
# 全局 mbuddy 指针
|
||
_mbuddy: memhub.MemManager | t.CPtr
|
||
|
||
# 最大 includes SHA1 数（用于过滤）
|
||
MAX_INCLUDES_SHA1: t.CDefine = 256
|
||
# stub 读取缓冲区大小
|
||
STUB_READ_BUF_SIZE: t.CDefine = 1048576
|
||
# 最大 includes 条目数
|
||
MAX_INCLUDES: t.CDefine = 256
|
||
|
||
# 全局 SHA1 映射存储器（内存中，不依赖 _sha1_map.txt 文件）
|
||
# Phase1 扫描完成后填充，供所有函数使用
|
||
_sha1_store_arr: bytes # SHA1 数组（每个 17 字节，MAX_INCLUDES 个）
|
||
_sha1_store_mod: bytes # 模块名数组（每个 64 字节，MAX_INCLUDES 个）
|
||
_sha1_store_rel: bytes # 相对路径数组（每个 256 字节，MAX_INCLUDES 个）
|
||
_sha1_store_count: int # 条目数
|
||
|
||
# rel_path 最大长度
|
||
MAX_REL_PATH_LEN: t.CDefine = 256
|
||
|
||
|
||
# ============================================================
|
||
# _sliced_path - 构建切片路径（stdlib.malloc 分配，调用者负责 free）
|
||
#
|
||
# 根据 Config.Sha1SliceLevel 将文件分散到 SHA1 前缀子目录:
|
||
# level=0 → {temp_dir}/{sha1}.{ext}
|
||
# level=1 → {temp_dir}/b7/{sha1}.{ext}
|
||
# level=2 → {temp_dir}/b7/90/{sha1}.{ext}
|
||
#
|
||
# 自动调用 BuildPipeline.ensure_dir 创建子目录（level>0 时）。
|
||
#
|
||
# Args:
|
||
# temp_dir: 基础目录
|
||
# td_len: temp_dir 长度（避免重复 strlen）
|
||
# sha1: SHA1 字符串
|
||
# ext: 文件扩展名（如 "stub.ll"）
|
||
#
|
||
# Returns:
|
||
# 完整路径（stdlib.malloc 分配，需 stdlib.free），None 失败
|
||
# ============================================================
|
||
def _sliced_path(temp_dir: str, td_len: t.CSizeT, sha1: str, ext: str) -> str:
|
||
"""构建切片路径（stdlib.malloc 分配，调用者负责 free）
|
||
|
||
level>0 时自动调用 ensure_dir 创建子目录。
|
||
"""
|
||
if temp_dir is None or sha1 is None or ext is None:
|
||
return None
|
||
sha1_len: t.CSizeT = string.strlen(sha1)
|
||
ext_len: t.CSizeT = string.strlen(ext)
|
||
|
||
# 子目录前缀缓冲区（最多 3 层 = 8 字节 + null = 16 足够）
|
||
SUBDIR_BUF_SIZE: t.CSizeT = 16
|
||
subdir_buf: bytes = stdlib.malloc(SUBDIR_BUF_SIZE)
|
||
if subdir_buf is None:
|
||
return None
|
||
subdir_buf[0] = '\0'
|
||
sub_len: t.CSizeT = 0
|
||
level: int = Config.Sha1SliceLevel
|
||
if level > 0:
|
||
li: int = 0
|
||
while li < level:
|
||
if li > 0:
|
||
subdir_buf[sub_len] = '/'
|
||
sub_len += 1
|
||
idx: int = li * 2
|
||
subdir_buf[sub_len] = sha1[idx]
|
||
subdir_buf[sub_len + 1] = sha1[idx + 1]
|
||
sub_len += 2
|
||
li += 1
|
||
subdir_buf[sub_len] = '\0'
|
||
|
||
# 确保子目录存在: {temp_dir}/{subdir}
|
||
# 不创建目录会导致 fileio.File(MODE.W) 静默失败（file.closed==True），
|
||
# 进而 BuildCombinedIR 读不到文件返回 0（"BuildCombinedIR 失败"）。
|
||
dir_path_len: t.CSizeT = td_len + sub_len + 2
|
||
dir_path: bytes = stdlib.malloc(dir_path_len)
|
||
if dir_path is not None:
|
||
viperlib.snprintf(dir_path, dir_path_len, "%s/%s", temp_dir, subdir_buf)
|
||
BuildPipeline.ensure_dir(dir_path)
|
||
stdlib.free(dir_path)
|
||
|
||
# 构建完整路径
|
||
path: str = None
|
||
path_len: t.CSizeT = 0
|
||
if sub_len > 0:
|
||
path_len = td_len + sub_len + sha1_len + ext_len + 4
|
||
path = stdlib.malloc(path_len)
|
||
if path is not None:
|
||
viperlib.snprintf(path, path_len, "%s/%s/%s.%s", temp_dir, subdir_buf, sha1, ext)
|
||
else:
|
||
path_len = td_len + sha1_len + ext_len + 3
|
||
path = stdlib.malloc(path_len)
|
||
if path is not None:
|
||
viperlib.snprintf(path, path_len, "%s/%s.%s", temp_dir, sha1, ext)
|
||
|
||
stdlib.free(subdir_buf)
|
||
return path
|
||
|
||
|
||
# ============================================================
|
||
# _collect_stub_sha1s - 递归扫描子目录，收集 .stub.ll 文件的 SHA1
|
||
#
|
||
# 根据 level 递归扫描子目录:
|
||
# level=0 → 扫描 {base_dir}/*.stub.ll，从文件名提取 SHA1
|
||
# level>0 → 扫描 {base_dir}/* 的子目录，对每个子目录递归调用
|
||
#
|
||
# Args:
|
||
# base_dir: 当前扫描目录
|
||
# bd_len: base_dir 长度
|
||
# level: 剩余递归层数
|
||
# out_arr: 输出数组（每个 SHA1 17 字节 = 16字符 + null）
|
||
# out_count: 当前已收集数量
|
||
# max_count: 最大数量
|
||
#
|
||
# Returns:
|
||
# 收集后的总数量
|
||
# ============================================================
|
||
def _collect_stub_sha1s(base_dir: str, bd_len: t.CSizeT, level: int,
|
||
out_arr: str, out_count: int, max_count: int) -> int:
|
||
"""递归扫描子目录，收集 .stub.ll 文件的 SHA1"""
|
||
if base_dir is None or out_arr is None:
|
||
return out_count
|
||
if out_count >= max_count:
|
||
return out_count
|
||
|
||
find_data: win32file.WIN32_FIND_DATAA | t.CPtr = stdlib.malloc(win32file.WIN32_FIND_DATAA.__sizeof__())
|
||
if find_data is None:
|
||
return out_count
|
||
string.memset(find_data, 0, win32file.WIN32_FIND_DATAA.__sizeof__())
|
||
|
||
if level <= 0:
|
||
# 扫描 {base_dir}/*.stub.ll
|
||
pattern: bytes = stdlib.malloc(bd_len + 16)
|
||
if pattern is None:
|
||
stdlib.free(find_data)
|
||
return out_count
|
||
viperlib.snprintf(pattern, bd_len + 16, "%s/*.stub.ll", base_dir)
|
||
handle: win32base.HANDLE = win32file.FindFirstFileA(pattern, find_data)
|
||
stdlib.free(pattern)
|
||
if handle != win32base.INVALID_HANDLE_VALUE:
|
||
while 1:
|
||
if out_count >= max_count:
|
||
break
|
||
fname: str = find_data.cFileName
|
||
if fname is not None:
|
||
if fname[0] != '\0':
|
||
# 从文件名提取 SHA1（前 16 字符）
|
||
string.strncpy(out_arr + out_count * 17, fname, 16)
|
||
out_arr[out_count * 17 + 16] = '\0'
|
||
out_count += 1
|
||
if win32file.FindNextFileA(handle, find_data) == 0:
|
||
break
|
||
win32file.FindClose(handle)
|
||
else:
|
||
# 扫描 {base_dir}/* 的子目录
|
||
pattern = stdlib.malloc(bd_len + 4)
|
||
if pattern is None:
|
||
stdlib.free(find_data)
|
||
return out_count
|
||
viperlib.snprintf(pattern, bd_len + 4, "%s/*", base_dir)
|
||
handle = win32file.FindFirstFileA(pattern, find_data)
|
||
stdlib.free(pattern)
|
||
if handle != win32base.INVALID_HANDLE_VALUE:
|
||
while 1:
|
||
# 检查是否是目录（FILE_ATTRIBUTE_DIRECTORY = 0x10 = 16）
|
||
attrs: t.CUInt32T = find_data.dwFileAttributes
|
||
if (attrs & 16) != 0:
|
||
fname = find_data.cFileName
|
||
if fname is not None:
|
||
# 跳过 "." 和 ".."
|
||
is_dot: int = 0
|
||
if fname[0] == '.':
|
||
if fname[1] == '\0':
|
||
is_dot = 1
|
||
elif fname[1] == '.' and fname[2] == '\0':
|
||
is_dot = 1
|
||
if is_dot == 0:
|
||
# 构建子目录路径并递归扫描
|
||
fname_len: t.CSizeT = string.strlen(fname)
|
||
sub_dir: str = stdlib.malloc(bd_len + fname_len + 2)
|
||
if sub_dir is not None:
|
||
viperlib.snprintf(sub_dir, bd_len + fname_len + 2, "%s/%s", base_dir, fname)
|
||
sub_dir_len: t.CSizeT = string.strlen(sub_dir)
|
||
out_count = _collect_stub_sha1s(sub_dir, sub_dir_len, level - 1, out_arr, out_count, max_count)
|
||
stdlib.free(sub_dir)
|
||
if win32file.FindNextFileA(handle, find_data) == 0:
|
||
break
|
||
win32file.FindClose(handle)
|
||
|
||
stdlib.free(find_data)
|
||
return out_count
|
||
|
||
|
||
# ============================================================
|
||
# _load_includes_sha1_set - 从全局内存存储器收集 includes 的 SHA1
|
||
#
|
||
# 数据来源是 PopulateSha1MapStore 填充的全局存储器（不读取 _sha1_map.txt）。
|
||
# 用于过滤 stub 文件，避免加载 TransPyV 自身源文件（lib/*）的 stub 导致链接失败。
|
||
# ============================================================
|
||
def _load_includes_sha1_set(pool: memhub.MemBuddy | t.CPtr,
|
||
temp_dir: str,
|
||
sha1_set: str) -> int:
|
||
"""从全局内存存储器将 includes 的 SHA1 写入 sha1_set
|
||
|
||
pool/temp_dir 参数保留以兼容现有调用者，但不再使用。
|
||
数据来源是 PopulateSha1MapStore 填充的全局存储器。
|
||
sha1_set 大小为 MAX_INCLUDES_SHA1 * 17（每个 SHA1 16字符+null）
|
||
返回找到的 includes SHA1 数量，-1 表示错误"""
|
||
if sha1_set is None:
|
||
return -1
|
||
if _sha1_store_arr is None:
|
||
return -1
|
||
count: int = 0
|
||
for i in range(_sha1_store_count):
|
||
if count >= MAX_INCLUDES_SHA1:
|
||
break
|
||
idx: t.CSizeT = t.CSizeT(i) * 17
|
||
string.strcpy(sha1_set + t.CSizeT(count) * 17, _sha1_store_arr + idx)
|
||
count += 1
|
||
return count
|
||
|
||
|
||
# ============================================================
|
||
# _is_in_sha1_set - 检查 SHA1 是否在集合中
|
||
# ============================================================
|
||
def _is_in_sha1_set(sha1: str, sha1_set: str, set_count: int) -> int:
|
||
"""检查 sha1 是否在集合中"""
|
||
if sha1 is None or sha1_set is None:
|
||
return 0
|
||
for i in range(set_count):
|
||
if string.strcmp(sha1_set + i * 17, sha1) == 0:
|
||
return 1
|
||
return 0
|
||
|
||
|
||
# ============================================================
|
||
# WriteIncludesSha1Map - 将 includes 扫描结果写入 _sha1_map.txt（人类可读输出）
|
||
#
|
||
# _sha1_map.txt 仅作为人类可读的调试输出，程序内部不读取此文件。
|
||
# 机器分析使用 PopulateSha1MapStore 填充的全局内存存储器。
|
||
#
|
||
# Args:
|
||
# mb: 内存池
|
||
# temp_dir: 临时目录（_sha1_map.txt 写入位置）
|
||
# scan_result: IncludesScanner 扫描结果
|
||
#
|
||
# Returns:
|
||
# 0 成功，非 0 失败
|
||
# ============================================================
|
||
def WriteIncludesSha1Map(mb: memhub.MemBuddy | t.CPtr, temp_dir: str,
|
||
scan_result: IncludesScanner.ScanResult | t.CPtr,
|
||
filter_set: t.CChar | t.CPtr,
|
||
filter_count: int) -> int:
|
||
"""将 includes 扫描结果写入 _sha1_map.txt（人类可读输出，程序内部使用内存存储器）
|
||
|
||
若 filter_set 不为 None 且 filter_count > 0，只写入 filter_set 中的条目。
|
||
"""
|
||
if temp_dir is None or scan_result is None:
|
||
return 1
|
||
|
||
# 构造路径 temp_dir/_sha1_map.txt
|
||
dir_len: t.CSizeT = string.strlen(temp_dir)
|
||
map_path: bytes = stdlib.malloc(dir_len + 32)
|
||
if map_path is None:
|
||
return 1
|
||
viperlib.snprintf(map_path, dir_len + 32, "%s/_sha1_map.txt", temp_dir)
|
||
|
||
# 打开文件写入（CREATE_ALWAYS）
|
||
f: fileio.File | t.CPtr = fileio.File(map_path, fileio.MODE.W)
|
||
if f.closed:
|
||
return 1
|
||
|
||
# 写入每个 include 条目: {sha1}:includes/{rel_path}\n
|
||
entry_size_w: t.CSizeT = IncludesScanner.FileEntry.__sizeof__()
|
||
line_buf: bytes = stdlib.malloc(512)
|
||
if line_buf is None:
|
||
f.close()
|
||
return 1
|
||
|
||
written_count: int = 0
|
||
for i in range(scan_result.Count):
|
||
ea_w: t.CUInt64T = t.CUInt64T(scan_result.Entries) + i * entry_size_w
|
||
ent_w: IncludesScanner.FileEntry | t.CPtr = (IncludesScanner.FileEntry | t.CPtr)(t.CVoid(ea_w, t.CPtr))
|
||
if ent_w is None or ent_w.Sha1 is None or ent_w.RelPath is None:
|
||
continue
|
||
# 若提供过滤集合，只写入集合中的条目
|
||
if filter_set is not None and filter_count > 0:
|
||
if _is_in_sha1_set(ent_w.Sha1, filter_set, filter_count) == 0:
|
||
continue
|
||
viperlib.snprintf(line_buf, 512, "%s:includes/%s\n", ent_w.Sha1, ent_w.RelPath)
|
||
line_len_w: t.CSizeT = string.strlen(line_buf)
|
||
f.write(line_buf, line_len_w)
|
||
written_count += 1
|
||
|
||
f.close()
|
||
return 0
|
||
|
||
|
||
# ============================================================
|
||
# PopulateSha1MapStore - 从扫描结果填充全局 SHA1 映射存储器
|
||
#
|
||
# 遍历 ScanResult.Entries，为每个条目构建 sha1→module_name 映射。
|
||
# 模块名通过 _PathToModuleName(includes/rel_path) 计算。
|
||
# 内存用 stdlib.malloc 分配（全局存储器，不随 mbuddy 释放）。
|
||
#
|
||
# Args:
|
||
# scan_result: IncludesScanner 扫描结果
|
||
#
|
||
# Returns:
|
||
# 填充的条目数，-1 表示失败
|
||
# ============================================================
|
||
def PopulateSha1MapStore(scan_result: IncludesScanner.ScanResult | t.CPtr) -> int:
|
||
"""从扫描结果填充全局 SHA1 映射存储器
|
||
|
||
遍历 ScanResult.Entries，为每个条目构建 sha1→module_name 映射。
|
||
模块名通过 _PathToModuleName(includes/rel_path) 计算。
|
||
内存用 stdlib.malloc 分配（全局存储器，不随 mbuddy 释放）。
|
||
"""
|
||
global _sha1_store_arr, _sha1_store_mod, _sha1_store_rel, _sha1_store_count
|
||
if scan_result is None:
|
||
return -1
|
||
# 释放旧存储器
|
||
if _sha1_store_arr is not None:
|
||
stdlib.free(_sha1_store_arr)
|
||
if _sha1_store_mod is not None:
|
||
stdlib.free(_sha1_store_mod)
|
||
if _sha1_store_rel is not None:
|
||
stdlib.free(_sha1_store_rel)
|
||
_sha1_store_arr = stdlib.malloc(MAX_INCLUDES * 17)
|
||
_sha1_store_mod = stdlib.malloc(MAX_INCLUDES * 64)
|
||
_sha1_store_rel = stdlib.malloc(MAX_INCLUDES * MAX_REL_PATH_LEN)
|
||
if _sha1_store_arr is None or _sha1_store_mod is None or _sha1_store_rel is None:
|
||
return -1
|
||
string.memset(_sha1_store_arr, 0, MAX_INCLUDES * 17)
|
||
string.memset(_sha1_store_mod, 0, MAX_INCLUDES * 64)
|
||
string.memset(_sha1_store_rel, 0, MAX_INCLUDES * MAX_REL_PATH_LEN)
|
||
_sha1_store_count = 0
|
||
entry_size_p: t.CSizeT = IncludesScanner.FileEntry.__sizeof__()
|
||
for i in range(scan_result.Count):
|
||
if _sha1_store_count >= MAX_INCLUDES:
|
||
break
|
||
ea: t.CUInt64T = t.CUInt64T(scan_result.Entries) + i * entry_size_p
|
||
ent: IncludesScanner.FileEntry | t.CPtr = (IncludesScanner.FileEntry | t.CPtr)(t.CVoid(ea, t.CPtr))
|
||
if ent is None or ent.Sha1 is None or ent.RelPath is None:
|
||
continue
|
||
# 构建完整路径 includes/{RelPath}（_PathToModuleName 需要）
|
||
rp_len: t.CSizeT = string.strlen(ent.RelPath)
|
||
full_path_buf: bytes = stdlib.malloc(rp_len + 10)
|
||
if full_path_buf is None:
|
||
continue
|
||
viperlib.snprintf(full_path_buf, rp_len + 10, "includes/%s", ent.RelPath)
|
||
mod_name: str = _PathToModuleName(full_path_buf)
|
||
stdlib.free(full_path_buf)
|
||
if mod_name is not None:
|
||
idx: t.CSizeT = t.CSizeT(_sha1_store_count) * 17
|
||
string.strcpy(_sha1_store_arr + idx, ent.Sha1)
|
||
idx2: t.CSizeT = t.CSizeT(_sha1_store_count) * 64
|
||
mn_len: t.CSizeT = string.strlen(mod_name)
|
||
if mn_len < 64:
|
||
string.strcpy(_sha1_store_mod + idx2, mod_name)
|
||
else:
|
||
string.strncpy(_sha1_store_mod + idx2, mod_name, 63)
|
||
_sha1_store_mod[idx2 + 63] = '\0'
|
||
stdlib.free(mod_name)
|
||
# 保存 rel_path（供 Phase2 3.5 编译缺失 includes 使用）
|
||
idx3: t.CSizeT = t.CSizeT(_sha1_store_count) * MAX_REL_PATH_LEN
|
||
if rp_len < MAX_REL_PATH_LEN:
|
||
string.strcpy(_sha1_store_rel + idx3, ent.RelPath)
|
||
else:
|
||
string.strncpy(_sha1_store_rel + idx3, ent.RelPath, MAX_REL_PATH_LEN - 1)
|
||
_sha1_store_rel[idx3 + MAX_REL_PATH_LEN - 1] = '\0'
|
||
_sha1_store_count += 1
|
||
return _sha1_store_count
|
||
|
||
|
||
# ============================================================
|
||
# AppendToSha1MapStore - 追加条目到全局 SHA1 映射存储器
|
||
#
|
||
# Phase1 调用 PopulateSha1MapStore 填充 includes 文件，
|
||
# Phase2 扫描 App 源文件后调用此函数追加 App 模块条目，
|
||
# 确保编译 App 自身时跨模块 CDefine 查找能命中 App 模块
|
||
# （如 StubMerger.MAX_INCLUDES 在编译 Phase2 时通过
|
||
# _build_source_path_from_sha1 查找 StubMerger 的源路径）。
|
||
#
|
||
# 若存储器未初始化（PopulateSha1MapStore 未调用），自动初始化。
|
||
# 若 SHA1 已存在（去重），跳过。
|
||
#
|
||
# Args:
|
||
# sha1: SHA1 字符串（16 字符 + null）
|
||
# rel_path: 相对路径（如 "lib/core/StubMerger.py"，
|
||
# 不含 includes/ 前缀，与 _sha1_map.txt 格式一致）
|
||
#
|
||
# Returns:
|
||
# 0 成功（含去重跳过），1 失败
|
||
# ============================================================
|
||
def AppendToSha1MapStore(sha1: str, rel_path: str) -> int:
|
||
"""追加条目到全局 SHA1 映射存储器（Phase2 用）"""
|
||
global _sha1_store_arr, _sha1_store_mod, _sha1_store_rel, _sha1_store_count
|
||
if sha1 is None or rel_path is None:
|
||
return 1
|
||
# 存储器未初始化：自动初始化（Phase2 可能先于 Phase1 调用）
|
||
if _sha1_store_arr is None:
|
||
_sha1_store_arr = stdlib.malloc(MAX_INCLUDES * 17)
|
||
_sha1_store_mod = stdlib.malloc(MAX_INCLUDES * 64)
|
||
_sha1_store_rel = stdlib.malloc(MAX_INCLUDES * MAX_REL_PATH_LEN)
|
||
if _sha1_store_arr is None or _sha1_store_mod is None or _sha1_store_rel is None:
|
||
return 1
|
||
string.memset(_sha1_store_arr, 0, MAX_INCLUDES * 17)
|
||
string.memset(_sha1_store_mod, 0, MAX_INCLUDES * 64)
|
||
string.memset(_sha1_store_rel, 0, MAX_INCLUDES * MAX_REL_PATH_LEN)
|
||
_sha1_store_count = 0
|
||
if _sha1_store_count >= MAX_INCLUDES:
|
||
return 1
|
||
# 去重：若 SHA1 已存在则跳过
|
||
for i in range(_sha1_store_count):
|
||
sidx_d: t.CSizeT = t.CSizeT(i) * 17
|
||
if string.strcmp(_sha1_store_arr + sidx_d, sha1) == 0:
|
||
return 0
|
||
# 模块名：rel_path "lib/core/StubMerger.py" → "lib.core.StubMerger"
|
||
full_path_buf: bytes = stdlib.malloc(string.strlen(rel_path) + 10)
|
||
if full_path_buf is None:
|
||
return 1
|
||
viperlib.snprintf(full_path_buf, string.strlen(rel_path) + 10, "includes/%s", rel_path)
|
||
mod_name: str = _PathToModuleName(full_path_buf)
|
||
stdlib.free(full_path_buf)
|
||
if mod_name is None:
|
||
return 1
|
||
# 写入 SHA1
|
||
idx_a: t.CSizeT = t.CSizeT(_sha1_store_count) * 17
|
||
string.strcpy(_sha1_store_arr + idx_a, sha1)
|
||
# 写入模块名
|
||
idx_m: t.CSizeT = t.CSizeT(_sha1_store_count) * 64
|
||
mn_len: t.CSizeT = string.strlen(mod_name)
|
||
if mn_len < 64:
|
||
string.strcpy(_sha1_store_mod + idx_m, mod_name)
|
||
else:
|
||
string.strncpy(_sha1_store_mod + idx_m, mod_name, 63)
|
||
_sha1_store_mod[idx_m + 63] = '\0'
|
||
stdlib.free(mod_name)
|
||
# 写入相对路径
|
||
idx_r: t.CSizeT = t.CSizeT(_sha1_store_count) * MAX_REL_PATH_LEN
|
||
rp_len: t.CSizeT = string.strlen(rel_path)
|
||
if rp_len < MAX_REL_PATH_LEN:
|
||
string.strcpy(_sha1_store_rel + idx_r, rel_path)
|
||
else:
|
||
string.strncpy(_sha1_store_rel + idx_r, rel_path, MAX_REL_PATH_LEN - 1)
|
||
_sha1_store_rel[idx_r + MAX_REL_PATH_LEN - 1] = '\0'
|
||
_sha1_store_count += 1
|
||
return 0
|
||
|
||
|
||
# ============================================================
|
||
# GetSha1StoreCount - 获取全局存储器条目数
|
||
# ============================================================
|
||
def GetSha1StoreCount() -> int:
|
||
"""返回全局 SHA1 存储器的条目数"""
|
||
return _sha1_store_count
|
||
|
||
|
||
def GetSha1StoreArrPtr() -> bytes | t.CPtr:
|
||
"""返回 SHA1 数组指针（每个条目 17 字节）"""
|
||
return _sha1_store_arr
|
||
|
||
|
||
def GetSha1StoreModArrPtr() -> bytes | t.CPtr:
|
||
"""返回模块名数组指针（每个条目 64 字节）"""
|
||
return _sha1_store_mod
|
||
|
||
|
||
def GetSha1StoreRelArrPtr() -> bytes | t.CPtr:
|
||
"""返回相对路径数组指针（每个条目 MAX_REL_PATH_LEN 字节）"""
|
||
return _sha1_store_rel
|
||
|
||
|
||
# ============================================================
|
||
# GetSha1StoreArr - 获取全局存储器数组指针（直接遍历用）
|
||
#
|
||
# 返回 sha1_arr、mod_arr、rel_arr 三个数组的指针和条目数。
|
||
# 调用者直接遍历数组：sha1_arr[i*17]、mod_arr[i*64]、rel_arr[i*MAX_REL_PATH_LEN]
|
||
# 内存归全局存储器所有，调用者不要 free
|
||
# ============================================================
|
||
def GetSha1StoreArr(out_sha1_arr: t.CPtr, out_mod_arr: t.CPtr, out_rel_arr: t.CPtr) -> int:
|
||
"""获取全局存储器数组指针
|
||
|
||
out_sha1_arr/out_mod_arr/out_rel_arr 是调用者提供的 t.CPtr 指针，
|
||
函数将存储器数组指针写入这些位置。
|
||
返回条目数，-1 失败
|
||
"""
|
||
if out_sha1_arr is None or out_mod_arr is None or out_rel_arr is None:
|
||
return -1
|
||
if _sha1_store_arr is None or _sha1_store_mod is None or _sha1_store_rel is None:
|
||
return -1
|
||
sp: t.CPtr = (t.CPtr | t.CPtr)(t.CVoid(out_sha1_arr, t.CPtr))
|
||
if sp is not None:
|
||
sp[0] = _sha1_store_arr
|
||
mp: t.CPtr = (t.CPtr | t.CPtr)(t.CVoid(out_mod_arr, t.CPtr))
|
||
if mp is not None:
|
||
mp[0] = _sha1_store_mod
|
||
rp: t.CPtr = (t.CPtr | t.CPtr)(t.CVoid(out_rel_arr, t.CPtr))
|
||
if rp is not None:
|
||
rp[0] = _sha1_store_rel
|
||
return _sha1_store_count
|
||
|
||
|
||
# ============================================================
|
||
# _IsFuncDeclaredOrDefined - 检查 out_buf 中是否已有函数的 declare/define
|
||
#
|
||
# 逐字节strncmp搜索 func_sig（如 @"sha1.func"(/ @printf( ），每次
|
||
# 找到匹配后检查行首是否以 "declare " 或 "define " 开头，跳过 call 语句。
|
||
# 避免 strstr 子串匹配误匹配 call ... @"func"(...) 中的 @"func"( 子串。
|
||
#
|
||
# Args:
|
||
# out_buf: 已有的 IR 文本缓冲区
|
||
# func_sig: 函数签名片段，如 @"sha1.func"(/ @printf(
|
||
#
|
||
# Returns:
|
||
# 1=已存在 declare/define, 0=不存在
|
||
# ============================================================
|
||
def _IsFuncDeclaredOrDefined(out_buf: bytes, func_sig: str) -> int:
|
||
"""检查 out_buf 中是否已有函数的 declare 或 define 行"""
|
||
if out_buf is None or func_sig is None:
|
||
return 0
|
||
sig_len: t.CSizeT = string.strlen(func_sig)
|
||
out_len: t.CSizeT = string.strlen(out_buf)
|
||
if sig_len == 0 or out_len == 0 or sig_len > out_len:
|
||
return 0
|
||
pos: t.CSizeT = 0
|
||
while pos + sig_len <= out_len:
|
||
if string.strncmp(out_buf + pos, func_sig, sig_len) == 0:
|
||
ls: t.CSizeT = pos
|
||
while ls > 0 and out_buf[ls - 1] != '\n':
|
||
ls -= 1
|
||
if string.strncmp(out_buf + ls, "declare ", 8) == 0:
|
||
return 1
|
||
if string.strncmp(out_buf + ls, "define ", 7) == 0:
|
||
return 1
|
||
pos += 1
|
||
return 0
|
||
|
||
|
||
# ============================================================
|
||
# _CommentOutDeclareInBuf - 将 out_buf 中匹配 func_sig 的 declare 行注释掉
|
||
#
|
||
# 用于解决 declare/define 冲突：当本地 text.ll 包含 define 而依赖
|
||
# text.ll 已生成了同函数的 declare 时，将 declare 行首字符改为 ';'
|
||
# 使其成为注释，然后追加 define 行。
|
||
# ============================================================
|
||
def _CommentOutDeclareInBuf(out_buf: bytes, func_sig: str) -> int:
|
||
"""在 out_buf 中搜索 declare 行（包含 func_sig），将其注释掉"""
|
||
if out_buf is None or func_sig is None:
|
||
return 0
|
||
sig_len: t.CSizeT = string.strlen(func_sig)
|
||
out_len: t.CSizeT = string.strlen(out_buf)
|
||
if sig_len == 0 or out_len == 0 or sig_len > out_len:
|
||
return 0
|
||
pos: t.CSizeT = 0
|
||
while pos + sig_len <= out_len:
|
||
if string.strncmp(out_buf + pos, func_sig, sig_len) == 0:
|
||
ls: t.CSizeT = pos
|
||
while ls > 0 and out_buf[ls - 1] != '\n':
|
||
ls -= 1
|
||
if string.strncmp(out_buf + ls, "declare ", 8) == 0:
|
||
out_buf[ls] = ';'
|
||
return 1
|
||
pos += 1
|
||
return 0
|
||
|
||
|
||
# ============================================================
|
||
# _CommentOutGlobalInBuf - 将 out_buf 中匹配全局变量名的行注释掉
|
||
#
|
||
# 解决 stub/text 全局变量重复定义问题：stub 中可能是 external 声明
|
||
# 或完整定义，以 text 中的定义为准，将 stub 中的同名行注释掉。
|
||
# ============================================================
|
||
def _CommentOutGlobalInBuf(out_buf: bytes, global_name: str) -> int:
|
||
"""在 out_buf 中搜索全局变量定义行（以 global_name 开头），将行首改为 ';'"""
|
||
if out_buf is None or global_name is None:
|
||
return 0
|
||
name_len: t.CSizeT = string.strlen(global_name)
|
||
out_len: t.CSizeT = string.strlen(out_buf)
|
||
if name_len == 0 or out_len == 0 or name_len > out_len:
|
||
return 0
|
||
pos: t.CSizeT = 0
|
||
while pos + name_len <= out_len:
|
||
if string.strncmp(out_buf + pos, global_name, name_len) == 0:
|
||
# 确认是行首（前一个是换行或 pos==0）
|
||
if pos == 0 or out_buf[pos - 1] == '\n':
|
||
out_buf[pos] = ';'
|
||
return 1
|
||
pos += 1
|
||
return 0
|
||
|
||
|
||
# ============================================================
|
||
# _CommentOutOpaqueTypeInBuf - 将 out_buf 中匹配类型名的 opaque 声明行注释掉
|
||
#
|
||
# 解决跨模块类型定义冲突：本地 stub 可能包含自动生成的 opaque 声明
|
||
#（如 %"sha1.Widget" = type opaque），而依赖 stub 包含完整定义
|
||
#（如 %"sha1.Widget" = type { i8*, i32 }）。去重时需要将 opaque 声明
|
||
# 注释掉，让完整定义被追加，否则 LLC 报错 "Cannot allocate unsized type"。
|
||
# ============================================================
|
||
def _CommentOutOpaqueTypeInBuf(out_buf: bytes, type_name_prefix: str) -> int:
|
||
"""在 out_buf 中搜索 type_name_prefix 对应的 opaque 声明行，将行首改为 ';'
|
||
|
||
type_name_prefix 格式: '%"sha1.ClassName"'（不含 '=' 和后续内容）
|
||
返回 1=已注释, 0=未找到
|
||
"""
|
||
if out_buf is None or type_name_prefix is None:
|
||
return 0
|
||
prefix_len: t.CSizeT = string.strlen(type_name_prefix)
|
||
out_len: t.CSizeT = string.strlen(out_buf)
|
||
if prefix_len == 0 or out_len == 0 or prefix_len > out_len:
|
||
return 0
|
||
pos: t.CSizeT = 0
|
||
while pos + prefix_len <= out_len:
|
||
if string.strncmp(out_buf + pos, type_name_prefix, prefix_len) == 0:
|
||
# 确认是行首（前一个是换行或 pos==0）
|
||
if pos == 0 or out_buf[pos - 1] == '\n':
|
||
# 检查后面是否是 " = type opaque"
|
||
p: t.CSizeT = pos + prefix_len
|
||
# 跳过空格
|
||
while p < out_len and out_buf[p] == ' ':
|
||
p += 1
|
||
if p < out_len and out_buf[p] == '=':
|
||
p += 1
|
||
while p < out_len and out_buf[p] == ' ':
|
||
p += 1
|
||
if p + 11 <= out_len:
|
||
if string.strncmp(out_buf + p, "type opaque", 11) == 0:
|
||
# 找到 opaque 声明，将行首改为 ';'
|
||
out_buf[pos] = ';'
|
||
return 1
|
||
pos += 1
|
||
return 0
|
||
|
||
|
||
# ============================================================
|
||
# _is_submodule_of_imported - 检查模块名是否是已导入包的子模块
|
||
#
|
||
# deps.txt 只记录直接导入的模块名（如 "ast"），但包的 __init__.py
|
||
# 内部 from .lexer import Lexer 会让 IR 引用 ast.lexer 子模块类型。
|
||
# 当模块名包含 '.' 时，检查其顶层包名是否在 deps_buf 中。
|
||
#
|
||
# 例如:
|
||
# deps_buf = "ast stdio string"
|
||
# mod_name = "ast.lexer" → 顶层包 "ast" 在 deps_buf 中 → 返回 1
|
||
# mod_name = "ast.token" → 顶层包 "ast" 在 deps_buf 中 → 返回 1
|
||
# mod_name = "foo.bar" → 顶层包 "foo" 不在 deps_buf 中 → 返回 0
|
||
# ============================================================
|
||
def _is_submodule_of_imported(deps_buf: str, mod_name: str) -> int:
|
||
"""检查 mod_name 是否是 deps_buf 中已导入包的子模块"""
|
||
if deps_buf is None or mod_name is None:
|
||
return 0
|
||
# 查找第一个 '.'，提取顶层包名
|
||
dot_pos: str = string.strchr(mod_name, 46) # '.' = 46
|
||
if dot_pos is None:
|
||
return 0
|
||
pkg_len: t.CSizeT = t.CSizeT(t.CUInt64T(dot_pos) - t.CUInt64T(mod_name))
|
||
if pkg_len == 0 or pkg_len >= 256:
|
||
return 0
|
||
# 构造顶层包名（截断到 '.' 之前）
|
||
# 直接在 mod_name 上临时截断检查，避免分配
|
||
saved_ch: t.CChar = mod_name[pkg_len]
|
||
mod_name[pkg_len] = '\0'
|
||
result: int = HandlesImports.is_module_imported(deps_buf, mod_name)
|
||
mod_name[pkg_len] = saved_ch
|
||
return result
|
||
|
||
|
||
# ============================================================
|
||
# _HasFullTypeDefinition - 检查 out_buf 中是否已有完整类型定义（非 opaque）
|
||
#
|
||
# type_name_prefix 格式: '%"sha1.ClassName"'（不含 '='）
|
||
# 完整定义: %"name" = type { ... } 或 %"name" = type < { ... } >
|
||
# 返回 1=已有完整定义, 0=无完整定义
|
||
# ============================================================
|
||
def _HasFullTypeDefinition(out_buf: bytes, type_name_prefix: str) -> int:
|
||
"""检查 out_buf 中是否已有 type_name_prefix 的完整定义"""
|
||
if out_buf is None or type_name_prefix is None:
|
||
return 0
|
||
prefix_len: t.CSizeT = string.strlen(type_name_prefix)
|
||
out_len: t.CSizeT = string.strlen(out_buf)
|
||
if prefix_len == 0 or out_len == 0 or prefix_len > out_len:
|
||
return 0
|
||
pos: t.CSizeT = 0
|
||
while pos + prefix_len <= out_len:
|
||
if string.strncmp(out_buf + pos, type_name_prefix, prefix_len) == 0:
|
||
# 确认是行首（前一个是换行或 pos==0），且未被注释（行首不是 ';'）
|
||
is_line_start: int = 0
|
||
if pos == 0:
|
||
is_line_start = 1
|
||
elif out_buf[pos - 1] == '\n':
|
||
is_line_start = 1
|
||
if is_line_start == 1:
|
||
# 检查后面是否是 " = type {" 或 " = type <"
|
||
p: t.CSizeT = pos + prefix_len
|
||
while p < out_len and out_buf[p] == ' ':
|
||
p += 1
|
||
if p < out_len and out_buf[p] == '=':
|
||
p += 1
|
||
while p < out_len and out_buf[p] == ' ':
|
||
p += 1
|
||
if p + 5 < out_len:
|
||
if string.strncmp(out_buf + p, "type ", 5) == 0:
|
||
p += 5
|
||
if p < out_len:
|
||
if out_buf[p] == '{' or out_buf[p] == '<':
|
||
return 1
|
||
pos += 1
|
||
return 0
|
||
|
||
|
||
# ============================================================
|
||
# _LoadAndAppendStub - 加载依赖 stub 并追加到 out_buf
|
||
#
|
||
# 读取 {temp_dir}/{sha1}.stub.ll，跳过 header（;, target, source_filename），
|
||
# 去重 declare/global，追加到 out_buf。返回新 out_pos。
|
||
# ============================================================
|
||
def _LoadAndAppendStub(temp_dir: str, td_len: t.CSizeT, dep_sha1: str,
|
||
dep_buf: bytes, out_buf: bytes, out_size: t.CSizeT,
|
||
out_pos: t.CSizeT) -> t.CSizeT:
|
||
"""加载依赖 stub 并追加（跳过 header，去重）"""
|
||
if temp_dir is None or dep_sha1 is None or dep_buf is None or out_buf is None:
|
||
return out_pos
|
||
|
||
dep_path: str = _sliced_path(temp_dir, td_len, dep_sha1, "stub.ll")
|
||
if dep_path is None:
|
||
return out_pos
|
||
df_ls: fileio.File | t.CPtr = fileio.File(dep_path, fileio.MODE.R)
|
||
if df_ls.closed:
|
||
stdlib.free(dep_path)
|
||
return out_pos
|
||
dep_br_ls: t.CInt64T = df_ls.read_all(dep_buf, STUB_READ_BUF_SIZE)
|
||
df_ls.close()
|
||
stdlib.free(dep_path)
|
||
if dep_br_ls <= 0:
|
||
return out_pos
|
||
if dep_br_ls < STUB_READ_BUF_SIZE:
|
||
dep_buf[dep_br_ls] = '\0'
|
||
else:
|
||
dep_buf[STUB_READ_BUF_SIZE - 1] = '\0'
|
||
|
||
# 注释掉依赖 stub 中的字符串常量行（模块内部的，不需要）
|
||
# 简单方案：搜索 "= external unnamed_addr constant" 并将行首的 @ 改为 ;@
|
||
cfs_pos: t.CSizeT = 0
|
||
while cfs_pos + 30 < t.CSizeT(dep_br_ls):
|
||
ext_p: t.CPtr = string.strstr(dep_buf + cfs_pos, "= external unnamed_addr constant")
|
||
if ext_p is None:
|
||
break
|
||
ext_off2: t.CSizeT = t.CSizeT(t.CUInt64T(ext_p) - t.CUInt64T(dep_buf))
|
||
# 向前查找行首的 @
|
||
ls: t.CSizeT = ext_off2
|
||
while ls > 0:
|
||
if dep_buf[ls] == '\n':
|
||
ls += 1
|
||
break
|
||
ls -= 1
|
||
if ls < ext_off2 and dep_buf[ls] == '@':
|
||
dep_buf[ls] = ';'
|
||
cfs_pos = ext_off2 + 30
|
||
|
||
# 跳过 header，追加内容（去重 declare/global）
|
||
dep_pos: t.CSizeT = 0
|
||
skipping: int = 1
|
||
while dep_pos < dep_br_ls:
|
||
line_start: t.CSizeT = dep_pos
|
||
while dep_pos < dep_br_ls:
|
||
if dep_buf[dep_pos] == '\n':
|
||
break
|
||
dep_pos += 1
|
||
line_len: t.CSizeT = dep_pos - line_start
|
||
if dep_pos < dep_br_ls:
|
||
dep_pos += 1
|
||
if skipping != 0:
|
||
if line_len == 0:
|
||
continue
|
||
ch0: int = dep_buf[line_start]
|
||
if ch0 == ';':
|
||
continue
|
||
if ch0 == 't' and line_len >= 6:
|
||
if string.strncmp(dep_buf + line_start, "target", 6) == 0:
|
||
continue
|
||
if ch0 == 's' and line_len >= 15:
|
||
if string.strncmp(dep_buf + line_start, "source_filename", 15) == 0:
|
||
continue
|
||
skipping = 0
|
||
# 去重：declare（用 _IsFuncDeclaredOrDefined 避免误匹配 call 语句）
|
||
if line_len >= 8 and string.strncmp(dep_buf + line_start, "declare ", 8) == 0:
|
||
at: t.CSizeT = line_start + 8
|
||
while at < line_start + line_len:
|
||
if dep_buf[at] == '@':
|
||
break
|
||
at += 1
|
||
if at < line_start + line_len:
|
||
lp: t.CSizeT = at + 1
|
||
while lp < line_start + line_len:
|
||
if dep_buf[lp] == '(':
|
||
break
|
||
lp += 1
|
||
if lp < line_start + line_len:
|
||
saved: t.CChar = dep_buf[lp + 1]
|
||
dep_buf[lp + 1] = '\0'
|
||
if _IsFuncDeclaredOrDefined(out_buf, dep_buf + at) != 0:
|
||
dep_buf[lp + 1] = saved
|
||
continue
|
||
dep_buf[lp + 1] = saved
|
||
# 去重：external global
|
||
if line_len > 0 and dep_buf[line_start] == '@':
|
||
eq: t.CSizeT = line_start + 1
|
||
while eq < line_start + line_len:
|
||
if dep_buf[eq] == '=':
|
||
break
|
||
eq += 1
|
||
if eq < line_start + line_len:
|
||
saved: t.CChar = dep_buf[eq + 1]
|
||
dep_buf[eq + 1] = '\0'
|
||
if string.strstr(out_buf, dep_buf + line_start) is not None:
|
||
dep_buf[eq + 1] = saved
|
||
continue
|
||
dep_buf[eq + 1] = saved
|
||
# 去重：type 定义（如 %"sha1.Point" = type {i32, i32}）
|
||
# 区分完整定义（= type { ... } / = type < ... >）和 opaque 声明（= type opaque）
|
||
# 当本地 stub 有 opaque 声明而依赖 stub 有完整定义时，注释掉 opaque 行，
|
||
# 让完整定义被追加，避免 LLC "Cannot allocate unsized type" 错误。
|
||
if line_len > 0 and dep_buf[line_start] == '%':
|
||
eq2: t.CSizeT = line_start + 1
|
||
while eq2 < line_start + line_len:
|
||
if dep_buf[eq2] == '=':
|
||
break
|
||
eq2 += 1
|
||
if eq2 < line_start + line_len:
|
||
saved2: t.CChar = dep_buf[eq2 + 1]
|
||
dep_buf[eq2 + 1] = '\0'
|
||
# 检查依赖行是否为完整定义（= type { 或 = type <）
|
||
dep_is_full_def: int = 0
|
||
if eq2 + 8 <= line_start + line_len:
|
||
if dep_buf[eq2 + 2] == 't' and dep_buf[eq2 + 3] == 'y' and dep_buf[eq2 + 4] == 'p' and dep_buf[eq2 + 5] == 'e' and dep_buf[eq2 + 6] == ' ':
|
||
if dep_buf[eq2 + 7] == '{' or dep_buf[eq2 + 7] == '<':
|
||
dep_is_full_def = 1
|
||
if dep_is_full_def == 1:
|
||
# 依赖行是完整定义
|
||
# 临时截断到 '=' 之前（不含 '='），辅助函数期望纯类型名前缀
|
||
saved_eq: t.CChar = dep_buf[eq2]
|
||
dep_buf[eq2] = '\0'
|
||
# 1. 注释掉 out_buf 中的 opaque 声明（如果有）
|
||
commented: int = _CommentOutOpaqueTypeInBuf(out_buf, dep_buf + line_start)
|
||
# 2. 检查 out_buf 中是否已有完整定义（避免重复）
|
||
has_full: int = _HasFullTypeDefinition(out_buf, dep_buf + line_start)
|
||
dep_buf[eq2] = saved_eq
|
||
if has_full == 1:
|
||
dep_buf[eq2 + 1] = saved2
|
||
continue
|
||
# 3. 追加完整定义
|
||
else:
|
||
# 依赖行是 opaque 声明，用原有子串匹配去重
|
||
found_opaque: t.CPtr = string.strstr(out_buf, dep_buf + line_start)
|
||
if found_opaque is not None:
|
||
dep_buf[eq2 + 1] = saved2
|
||
continue
|
||
dep_buf[eq2 + 1] = saved2
|
||
# 追加
|
||
if out_pos + line_len + 2 < out_size:
|
||
string.strncpy(out_buf + out_pos, dep_buf + line_start, line_len)
|
||
out_pos += line_len
|
||
out_buf[out_pos] = '\n'
|
||
out_pos += 1
|
||
out_buf[out_pos] = '\0'
|
||
return out_pos
|
||
|
||
|
||
# ============================================================
|
||
# _LoadAndAppendTextDeclares - 从依赖 text.ll 提取 declare 追加到 out_buf
|
||
#
|
||
# 读取 {temp_dir}/{sha1}.text.ll，将 define 行转为 declare（提取函数签名），
|
||
# 跳过函数体，去重后追加到 out_buf。返回新 out_pos。
|
||
# ============================================================
|
||
def _LoadAndAppendTextDeclares(temp_dir: str, td_len: t.CSizeT, dep_sha1: str,
|
||
dep_buf: bytes, out_buf: bytes, out_size: t.CSizeT,
|
||
out_pos: t.CSizeT) -> t.CSizeT:
|
||
"""从依赖 text.ll 提取 declare（define 转 declare），追加到 out_buf"""
|
||
if temp_dir is None or dep_sha1 is None or dep_buf is None or out_buf is None:
|
||
return out_pos
|
||
|
||
text_path: str = _sliced_path(temp_dir, td_len, dep_sha1, "text.ll")
|
||
if text_path is None:
|
||
return out_pos
|
||
tf_lt: fileio.File | t.CPtr = fileio.File(text_path, fileio.MODE.R)
|
||
if tf_lt.closed:
|
||
stdlib.free(text_path)
|
||
return out_pos
|
||
text_br: t.CInt64T = tf_lt.read_all(dep_buf, STUB_READ_BUF_SIZE)
|
||
tf_lt.close()
|
||
stdlib.free(text_path)
|
||
if text_br <= 0:
|
||
return out_pos
|
||
if text_br < STUB_READ_BUF_SIZE:
|
||
dep_buf[text_br] = '\0'
|
||
else:
|
||
dep_buf[STUB_READ_BUF_SIZE - 1] = '\0'
|
||
|
||
dep_pos_lt: t.CSizeT = 0
|
||
in_body: int = 0
|
||
while dep_pos_lt < text_br:
|
||
ls: t.CSizeT = dep_pos_lt
|
||
while dep_pos_lt < text_br:
|
||
if dep_buf[dep_pos_lt] == '\n':
|
||
break
|
||
dep_pos_lt += 1
|
||
ll: t.CSizeT = dep_pos_lt - ls
|
||
if dep_pos_lt < text_br:
|
||
dep_pos_lt += 1
|
||
|
||
# 函数体跳过：直到单独的 } 行
|
||
if in_body != 0:
|
||
if ll == 1 and dep_buf[ls] == '}':
|
||
in_body = 0
|
||
continue
|
||
|
||
# 跳过空行和 header
|
||
if ll == 0:
|
||
continue
|
||
c0: int = dep_buf[ls]
|
||
if c0 == ';':
|
||
continue
|
||
if c0 == 't' and ll >= 6 and string.strncmp(dep_buf + ls, "target", 6) == 0:
|
||
continue
|
||
if c0 == 's' and ll >= 15 and string.strncmp(dep_buf + ls, "source_filename", 15) == 0:
|
||
continue
|
||
|
||
# 处理 define 行：转为 declare
|
||
if ll >= 7 and string.strncmp(dep_buf + ls, "define ", 7) == 0:
|
||
# 提取 @funcname( 用于去重
|
||
at_lt: t.CSizeT = ls + 7
|
||
while at_lt < ls + ll:
|
||
if dep_buf[at_lt] == '@':
|
||
break
|
||
at_lt += 1
|
||
if at_lt >= ls + ll:
|
||
in_body = 1
|
||
continue
|
||
lp_lt: t.CSizeT = at_lt + 1
|
||
while lp_lt < ls + ll:
|
||
if dep_buf[lp_lt] == '(':
|
||
break
|
||
lp_lt += 1
|
||
if lp_lt >= ls + ll:
|
||
in_body = 1
|
||
continue
|
||
# 去重检查：用 _IsFuncDeclaredOrDefined 避免误匹配 call 语句
|
||
saved_lt: t.CChar = dep_buf[lp_lt + 1]
|
||
dep_buf[lp_lt + 1] = '\0'
|
||
if _IsFuncDeclaredOrDefined(out_buf, dep_buf + at_lt) != 0:
|
||
dep_buf[lp_lt + 1] = saved_lt
|
||
in_body = 1
|
||
continue
|
||
dep_buf[lp_lt + 1] = saved_lt
|
||
|
||
# 找行尾的 ' {' 并截断
|
||
decl_end: t.CSizeT = ls + ll
|
||
if decl_end > ls + 1:
|
||
if dep_buf[decl_end - 1] == '{' and dep_buf[decl_end - 2] == ' ':
|
||
decl_end -= 2
|
||
|
||
# 输出: declare + content (skip "define " 7 chars, up to decl_end)
|
||
copy_len_lt: t.CSizeT = decl_end - (ls + 7)
|
||
if out_pos + 8 + copy_len_lt + 2 < out_size:
|
||
string.strncpy(out_buf + out_pos, "declare ", 8)
|
||
out_pos += 8
|
||
string.strncpy(out_buf + out_pos, dep_buf + ls + 7, copy_len_lt)
|
||
out_pos += copy_len_lt
|
||
out_buf[out_pos] = '\n'
|
||
out_pos += 1
|
||
out_buf[out_pos] = '\0'
|
||
in_body = 1
|
||
continue
|
||
|
||
# 提取类型定义（%... = type { ... } 或 %... = type < ... >）
|
||
# stub.ll 中只有 opaque 声明，需要从 text.ll 中提取完整定义
|
||
# 用于 alloca 和 getelementptr 等需要完整类型的指令
|
||
if ll > 0 and dep_buf[ls] == '%':
|
||
eq_td: t.CSizeT = ls + 1
|
||
while eq_td < ls + ll:
|
||
if dep_buf[eq_td] == '=':
|
||
break
|
||
eq_td += 1
|
||
if eq_td < ls + ll:
|
||
# 检查是否为完整定义（= type { 或 = type <）
|
||
dep_is_full_td: int = 0
|
||
if eq_td + 8 <= ls + ll:
|
||
if dep_buf[eq_td + 2] == 't' and dep_buf[eq_td + 3] == 'y' \
|
||
and dep_buf[eq_td + 4] == 'p' and dep_buf[eq_td + 5] == 'e' \
|
||
and dep_buf[eq_td + 6] == ' ':
|
||
if dep_buf[eq_td + 7] == '{' or dep_buf[eq_td + 7] == '<':
|
||
dep_is_full_td = 1
|
||
if dep_is_full_td == 1:
|
||
# 临时截断到 '=' 之前（不含 '='），辅助函数期望纯类型名前缀
|
||
saved_eq_td: t.CChar = dep_buf[eq_td]
|
||
dep_buf[eq_td] = '\0'
|
||
# 1. 注释掉 out_buf 中的 opaque 声明（如果有）
|
||
_CommentOutOpaqueTypeInBuf(out_buf, dep_buf + ls)
|
||
# 2. 检查 out_buf 中是否已有完整定义（避免重复）
|
||
has_full_td: int = _HasFullTypeDefinition(out_buf, dep_buf + ls)
|
||
dep_buf[eq_td] = saved_eq_td
|
||
if has_full_td == 0:
|
||
# 追加完整类型定义
|
||
if out_pos + ll + 2 < out_size:
|
||
string.strncpy(out_buf + out_pos, dep_buf + ls, ll)
|
||
out_pos += ll
|
||
out_buf[out_pos] = '\n'
|
||
out_pos += 1
|
||
out_buf[out_pos] = '\0'
|
||
continue
|
||
|
||
# 提取全局变量定义（@... = ...）
|
||
# 依赖 text.ll 中的全局变量带初值定义（如 @_mbuddy = global ...）
|
||
# stub.ll 中只有 external 声明，需要从 text.ll 提取完整定义
|
||
# 注释掉 out_buf 中的同名行（external 声明或旧定义），追加 text.ll 中的完整定义
|
||
if ll > 0 and dep_buf[ls] == '@':
|
||
eq_gv: t.CSizeT = ls + 1
|
||
while eq_gv < ls + ll:
|
||
if dep_buf[eq_gv] == '=':
|
||
break
|
||
eq_gv += 1
|
||
if eq_gv < ls + ll:
|
||
# 截断到 '=' 之前（不含 '='），得到全局变量名 @name
|
||
saved_gv: t.CChar = dep_buf[eq_gv]
|
||
dep_buf[eq_gv] = '\0'
|
||
# 注释掉 out_buf 中的同名行（external 声明或旧定义）
|
||
_CommentOutGlobalInBuf(out_buf, dep_buf + ls)
|
||
dep_buf[eq_gv] = saved_gv
|
||
# 追加完整定义
|
||
if out_pos + ll + 2 < out_size:
|
||
string.strncpy(out_buf + out_pos, dep_buf + ls, ll)
|
||
out_pos += ll
|
||
out_buf[out_pos] = '\n'
|
||
out_pos += 1
|
||
out_buf[out_pos] = '\0'
|
||
continue
|
||
|
||
# 跳过其他行
|
||
return out_pos
|
||
|
||
|
||
# ============================================================
|
||
# _PathToModuleName - 将文件路径转换为模块名
|
||
#
|
||
# 输入: "includes/stdio.py" 或 "includes/lib/core/Handles/HandlesType.py"
|
||
# 输出: "stdio" 或 "lib.core.Handles.HandlesType"
|
||
# 去除 includes/ 前缀和 .py 后缀，/ 替换为 .
|
||
# 返回 stdlib.malloc 分配的字符串，调用者负责释放
|
||
# ============================================================
|
||
def _PathToModuleName(path: str) -> str:
|
||
"""将文件路径转换为模块名
|
||
|
||
特殊处理 __init__.py：includes/ast/__init__.py → ast（而非 ast.__init__）
|
||
"""
|
||
if path is None:
|
||
return None
|
||
plen: t.CSizeT = string.strlen(path)
|
||
# 跳过 includes/ 或 includes\ 前缀（Windows 反斜杠兼容）
|
||
prefix_len: t.CSizeT = 9
|
||
path_start: t.CSizeT = 0
|
||
if plen > prefix_len and string.strncmp(path, "includes", 8) == 0:
|
||
psep_ch: t.CChar = path[8]
|
||
if psep_ch == '/' or psep_ch == '\\':
|
||
path_start = prefix_len
|
||
plen = plen - prefix_len
|
||
# 去掉 .py 后缀
|
||
if plen > 3 and path[path_start + plen - 3] == '.' and path[path_start + plen - 2] == 'p' and path[path_start + plen - 1] == 'y':
|
||
plen -= 3
|
||
if plen == 0:
|
||
return None
|
||
# 检查是否以 /__init__ 或 \__init__ 结尾（包的 __init__.py）
|
||
# /__init__ 是 9 个字符：/ _ _ i n i t _ _
|
||
# 如果是，去掉分隔符+__init__ 后缀，模块名就是包名（如 ast/__init__ → ast）
|
||
# 注意：顶级 __init__.py（plen=8）不会进入此分支，因为 plen > 9 才检查
|
||
# 同时支持 / 和 \ 两种分隔符（TPC 写入的路径可能混合使用）
|
||
if plen > 9:
|
||
tail_off: t.CSizeT = path_start + plen - 9
|
||
sep_char: t.CChar = path[tail_off]
|
||
if (sep_char == '/' or sep_char == '\\') \
|
||
and path[tail_off + 1] == '_' and path[tail_off + 2] == '_' \
|
||
and path[tail_off + 3] == 'i' and path[tail_off + 4] == 'n' and path[tail_off + 5] == 'i' \
|
||
and path[tail_off + 6] == 't' and path[tail_off + 7] == '_' and path[tail_off + 8] == '_':
|
||
# 去掉 分隔符+__init__ 后缀，plen 就是包名部分长度
|
||
plen = tail_off - path_start
|
||
if plen == 0:
|
||
return None
|
||
# 复制并替换 / 为 .
|
||
result: str = stdlib.malloc(plen + 1)
|
||
if result is None:
|
||
return None
|
||
for ri in range(plen):
|
||
ch: t.CChar = path[path_start + ri]
|
||
if ch == '/' or ch == '\\':
|
||
result[ri] = '.'
|
||
else:
|
||
result[ri] = ch
|
||
result[plen] = '\0'
|
||
return result
|
||
|
||
|
||
# ============================================================
|
||
# _BuildIncludesSha1Map - 从全局内存存储器构建 sha1 → module_name 映射
|
||
#
|
||
# 数据来源是 PopulateSha1MapStore 填充的全局存储器（不读取 _sha1_map.txt）。
|
||
# 返回两个数组（sha1_list 和 mod_list），长度存储在 count 中。
|
||
# 所有内存由 stdlib.malloc 分配，调用者负责释放。
|
||
# ============================================================
|
||
def _BuildIncludesSha1Map(temp_dir: str, td_len: t.CSizeT,
|
||
sha1_list: t.CChar | t.CPtr,
|
||
mod_list: t.CChar | t.CPtr) -> int:
|
||
"""从全局内存存储器复制 sha1→module_name 映射到调用者数组
|
||
|
||
temp_dir/td_len 参数保留以兼容现有调用者，但不再使用。
|
||
数据来源是 PopulateSha1MapStore 填充的全局存储器。
|
||
"""
|
||
if sha1_list is None or mod_list is None:
|
||
return 0
|
||
if _sha1_store_arr is None or _sha1_store_mod is None:
|
||
return 0
|
||
count: int = 0
|
||
for i in range(_sha1_store_count):
|
||
if count >= MAX_INCLUDES:
|
||
break
|
||
idx: t.CSizeT = t.CSizeT(i) * 17
|
||
idx2: t.CSizeT = t.CSizeT(i) * 64
|
||
string.strcpy(sha1_list + t.CSizeT(count) * 17, _sha1_store_arr + idx)
|
||
string.strcpy(mod_list + t.CSizeT(count) * 64, _sha1_store_mod + idx2)
|
||
count += 1
|
||
return count
|
||
|
||
|
||
# ============================================================
|
||
# _FindSha1ByModName - 从模块名查找 SHA1（精确匹配）
|
||
#
|
||
# 在 sha1_arr/mod_arr 映射中查找模块名对应的 SHA1。
|
||
# 用于依赖图分析：模块名 → SHA1。
|
||
#
|
||
# Args:
|
||
# sha1_arr: SHA1 数组（每个 17 字节）
|
||
# mod_arr: 模块名数组（每个 64 字节）
|
||
# count: 映射条目数
|
||
# mod_name: 要查找的模块名
|
||
#
|
||
# Returns:
|
||
# SHA1 字符串指针（指向 sha1_arr 内部，无需释放），未找到返回 None
|
||
# ============================================================
|
||
def _FindSha1ByModName(sha1_arr: t.CChar | t.CPtr, mod_arr: t.CChar | t.CPtr,
|
||
count: int, mod_name: str) -> str:
|
||
"""从模块名查找 SHA1（精确匹配 + 包名回退）
|
||
|
||
精确匹配失败时，尝试 "{mod_name}.__init__" 回退，
|
||
支持 import ast（包）→ ast.__init__ 的 SHA1 查找。
|
||
"""
|
||
if sha1_arr is None or mod_arr is None or mod_name is None:
|
||
return None
|
||
if count <= 0:
|
||
return None
|
||
# 第一遍：精确匹配
|
||
for i in range(count):
|
||
idx: t.CSizeT = t.CSizeT(i) * 64
|
||
if string.strcmp(mod_arr + idx, mod_name) == 0:
|
||
sidx: t.CSizeT = t.CSizeT(i) * 17
|
||
return sha1_arr + sidx
|
||
# 第二遍：包名回退（"ast" → "ast.__init__"）
|
||
nm_len: t.CSizeT = string.strlen(mod_name)
|
||
init_buf: bytes = stdlib.malloc(nm_len + 12)
|
||
if init_buf is not None:
|
||
viperlib.snprintf(init_buf, nm_len + 12, "%s.__init__", mod_name)
|
||
for i in range(count):
|
||
idx2: t.CSizeT = t.CSizeT(i) * 64
|
||
if string.strcmp(mod_arr + idx2, init_buf) == 0:
|
||
sidx2: t.CSizeT = t.CSizeT(i) * 17
|
||
stdlib.free(init_buf)
|
||
return sha1_arr + sidx2
|
||
stdlib.free(init_buf)
|
||
return None
|
||
|
||
|
||
# ============================================================
|
||
# _AddSha1ToSet - 将 SHA1 加入集合（去重）
|
||
#
|
||
# 检查 SHA1 是否已在集合中，若不在则加入。返回新的 count。
|
||
#
|
||
# Args:
|
||
# set_buf: SHA1 集合缓冲区（每个 17 字节）
|
||
# count: 当前集合中的 SHA1 数量
|
||
# sha1: 要加入的 SHA1
|
||
#
|
||
# Returns:
|
||
# 新的 count（如果已存在则不变）
|
||
# ============================================================
|
||
def _AddSha1ToSet(set_buf: t.CChar | t.CPtr, count: int, sha1: str) -> int:
|
||
"""将 SHA1 加入集合，返回新 count"""
|
||
if set_buf is None or sha1 is None:
|
||
return count
|
||
for i in range(count):
|
||
idx: t.CSizeT = t.CSizeT(i) * 17
|
||
if string.strcmp(set_buf + idx, sha1) == 0:
|
||
return count
|
||
if count < MAX_INCLUDES_SHA1:
|
||
idx: t.CSizeT = t.CSizeT(count) * 17
|
||
string.strcpy(set_buf + idx, sha1)
|
||
return count + 1
|
||
return count
|
||
|
||
|
||
# ============================================================
|
||
# _BuildReachableSha1Set - 构建可达 SHA1 集合（依赖图按需翻译）
|
||
#
|
||
# 从 source_dir 的源文件开始，解析 import 语句，递归收集可达的
|
||
# includes 文件 SHA1。Phase 1b 只翻译可达集合中的文件，避免
|
||
# 翻译不需要的 includes（如 Test 不依赖 llvmlite，则不翻译）。
|
||
#
|
||
# 算法（工作列表）：
|
||
# 1. 扫描 source_dir 下的 .py 文件，解析 AST 获取 _imported_modules
|
||
# 2. 构建 SHA1→module_name 映射（从全局内存存储器）
|
||
# 3. 工作列表递归：模块名 → 查 SHA1 → 加入 reachable_set → 读 .deps.txt → 追加新模块名
|
||
#
|
||
# Args:
|
||
# mb: 内存池
|
||
# source_dir: 源文件目录（Config.SourceDir）
|
||
# temp_dir: 临时目录（.deps.txt 所在位置；_sha1_map.txt 仅人类可读输出，不读取）
|
||
# reachable_set: 输出参数，可达 SHA1 集合缓冲区（MAX_INCLUDES_SHA1 * 17 字节）
|
||
#
|
||
# Returns:
|
||
# 可达 SHA1 数量（>0 成功，<=0 失败）
|
||
# ============================================================
|
||
def _BuildReachableSha1Set(mb: memhub.MemBuddy | t.CPtr, source_dir: str,
|
||
temp_dir: str,
|
||
reachable_set: t.CChar | t.CPtr) -> int:
|
||
"""构建可达 SHA1 集合（依赖图按需翻译）"""
|
||
if source_dir is None or temp_dir is None or reachable_set is None:
|
||
return -1
|
||
|
||
SRC_BUF_SIZE_R: t.CSizeT = 1048576
|
||
|
||
# 1. 构建 SHA1→module_name 映射
|
||
td_len_r: t.CSizeT = string.strlen(temp_dir)
|
||
sha1_arr: bytes = stdlib.malloc(MAX_INCLUDES * 17)
|
||
mod_arr: bytes = stdlib.malloc(MAX_INCLUDES * 64)
|
||
if sha1_arr is None or mod_arr is None:
|
||
if sha1_arr is not None:
|
||
stdlib.free(sha1_arr)
|
||
if mod_arr is not None:
|
||
stdlib.free(mod_arr)
|
||
return -1
|
||
map_count: int = _BuildIncludesSha1Map(temp_dir, td_len_r, sha1_arr, mod_arr)
|
||
if map_count <= 0:
|
||
VLogger.error("无法构建 SHA1 映射", "Reachable")
|
||
stdlib.free(sha1_arr)
|
||
stdlib.free(mod_arr)
|
||
return -1
|
||
fb: t.CChar | t.CPtr = VLogger.fmt_buf()
|
||
if fb is not None:
|
||
viperlib.snprintf(fb, 1024, "SHA1 映射: %d 个", map_count)
|
||
VLogger.debug(fb, "Reachable")
|
||
|
||
# 2. 扫描 source_dir 下的 .py 文件，收集直接依赖
|
||
dir_len: t.CSizeT = string.strlen(source_dir)
|
||
pattern: bytes = stdlib.malloc(dir_len + 8)
|
||
if pattern is None:
|
||
stdlib.free(sha1_arr)
|
||
stdlib.free(mod_arr)
|
||
return -1
|
||
viperlib.snprintf(pattern, dir_len + 8, "%s/*.py", source_dir)
|
||
|
||
find_data_size: t.CSizeT = win32file.WIN32_FIND_DATAA.__sizeof__()
|
||
find_data: win32file.WIN32_FIND_DATAA | t.CPtr = stdlib.malloc(find_data_size + 16)
|
||
if find_data is None:
|
||
stdlib.free(pattern)
|
||
stdlib.free(sha1_arr)
|
||
stdlib.free(mod_arr)
|
||
return -1
|
||
string.memset(find_data, 0, find_data_size + 16)
|
||
|
||
# 工作列表（空格分隔的模块名）
|
||
worklist: bytes = stdlib.malloc(8192)
|
||
if worklist is None:
|
||
stdlib.free(pattern)
|
||
stdlib.free(find_data)
|
||
stdlib.free(sha1_arr)
|
||
stdlib.free(mod_arr)
|
||
return -1
|
||
worklist[0] = '\0'
|
||
wl_len: t.CSizeT = 0
|
||
|
||
reachable_count: int = 0
|
||
|
||
handle: win32base.HANDLE = win32file.FindFirstFileA(pattern, find_data)
|
||
if handle == win32base.INVALID_HANDLE_VALUE:
|
||
fb: t.CChar | t.CPtr = VLogger.fmt_buf()
|
||
if fb is not None:
|
||
viperlib.snprintf(fb, 1024, "未找到 .py 文件: %s", pattern)
|
||
VLogger.warning(fb, "Reachable")
|
||
stdlib.free(pattern)
|
||
stdlib.free(find_data)
|
||
stdlib.free(sha1_arr)
|
||
stdlib.free(mod_arr)
|
||
stdlib.free(worklist)
|
||
return -1
|
||
|
||
while True:
|
||
fname: str = find_data.cFileName
|
||
if fname is not None:
|
||
fname_len: t.CSizeT = string.strlen(fname)
|
||
if fname_len > 3:
|
||
is_py: int = 0
|
||
if fname[fname_len - 3] == '.' and fname[fname_len - 2] == 'p' and fname[fname_len - 1] == 'y':
|
||
is_py = 1
|
||
if is_py != 0:
|
||
full_path: bytes = stdlib.malloc(dir_len + fname_len + 2)
|
||
if full_path is not None:
|
||
viperlib.snprintf(full_path, dir_len + fname_len + 2, "%s/%s", source_dir, fname)
|
||
sf: fileio.File | t.CPtr = fileio.File(full_path, fileio.MODE.R)
|
||
if not sf.closed:
|
||
sbuf: bytes = stdlib.malloc(SRC_BUF_SIZE_R)
|
||
if sbuf is not None:
|
||
br: LONG = sf.read_all(sbuf, SRC_BUF_SIZE_R)
|
||
sf.close()
|
||
if br > 0:
|
||
if br < SRC_BUF_SIZE_R:
|
||
sbuf[br] = 0
|
||
else:
|
||
sbuf[SRC_BUF_SIZE_R - 1] = 0
|
||
# 解析 AST 获取 _imported_modules
|
||
lx: ast.Lexer | t.CPtr = ast.new_lexer(mb)
|
||
if lx is not None:
|
||
ast._lexer_init(lx, sbuf, mb)
|
||
tokens: ast.Token | t.CPtr = ast.tokenize(lx)
|
||
tree: ast.AST | t.CPtr = ast.parse_tokens(mb, tokens)
|
||
if tree is not None:
|
||
tr: HandlesTranslator.Translator | t.CPtr = HandlesTranslator.Translator()
|
||
if tr is not None:
|
||
tr._declare_only = 2
|
||
# 源文件包名：fname 是顶级文件名（无目录分隔符），包为 None
|
||
tr.CurrentPackage = HandlesImports.compute_package_from_relpath(mb, fname)
|
||
HandlesStruct.reset_visible_structs(mb, 0)
|
||
tr.translate(tree)
|
||
# 获取 _imported_modules
|
||
if tr._imported_modules is not None:
|
||
im_len: t.CSizeT = string.strlen(tr._imported_modules)
|
||
if im_len > 0 and wl_len + im_len + 1 < 8192:
|
||
string.strcpy(worklist + wl_len, tr._imported_modules)
|
||
wl_len += im_len
|
||
worklist[wl_len] = ' '
|
||
wl_len += 1
|
||
worklist[wl_len] = '\0'
|
||
# 释放 Translator 资源
|
||
if tr._global_names is not None:
|
||
stdlib.free(tr._global_names)
|
||
if tr._nonlocal_names is not None:
|
||
stdlib.free(tr._nonlocal_names)
|
||
stdlib.free(sbuf)
|
||
else:
|
||
stdlib.free(sbuf)
|
||
else:
|
||
sf.close()
|
||
stdlib.free(full_path)
|
||
|
||
if win32file.FindNextFileA(handle, find_data) == 0:
|
||
break
|
||
win32base.FindClose(handle)
|
||
stdlib.free(pattern)
|
||
stdlib.free(find_data)
|
||
|
||
# 去掉末尾多余空格
|
||
if wl_len > 0 and worklist[wl_len - 1] == ' ':
|
||
worklist[wl_len - 1] = '\0'
|
||
wl_len -= 1
|
||
|
||
fb: t.CChar | t.CPtr = VLogger.fmt_buf()
|
||
if fb is not None:
|
||
viperlib.snprintf(fb, 1024, "源文件直接依赖: '%s'", worklist)
|
||
VLogger.debug(fb, "Reachable")
|
||
|
||
# 2.5. 全局预加入基础容器模块（_list/_dict/json）
|
||
#
|
||
# 这些模块定义了 list/dict 等内建容器的泛型模板。
|
||
# 使用 list[...]/dict[...] 语法时需要这些模块被翻译并注册模板。
|
||
# 即使源文件和 includes 文件没有显式 import _list，也预加入以确保可达。
|
||
# 这对应 TPC 的 _inject_auto_imports 机制（TPC 在翻译前自动注入 import _list）。
|
||
# _list
|
||
if HandlesImports.is_module_imported(worklist, "_list") == 0:
|
||
bsha1_l: str = _FindSha1ByModName(sha1_arr, mod_arr, map_count, "_list")
|
||
if bsha1_l is not None:
|
||
if wl_len + 7 < 8192:
|
||
if wl_len > 0:
|
||
worklist[wl_len] = ' '
|
||
wl_len += 1
|
||
string.strcpy(worklist + wl_len, "_list")
|
||
wl_len += 5
|
||
worklist[wl_len] = ' '
|
||
wl_len += 1
|
||
worklist[wl_len] = '\0'
|
||
# _dict
|
||
if HandlesImports.is_module_imported(worklist, "_dict") == 0:
|
||
bsha1_d: str = _FindSha1ByModName(sha1_arr, mod_arr, map_count, "_dict")
|
||
if bsha1_d is not None:
|
||
if wl_len + 7 < 8192:
|
||
if wl_len > 0:
|
||
worklist[wl_len] = ' '
|
||
wl_len += 1
|
||
string.strcpy(worklist + wl_len, "_dict")
|
||
wl_len += 5
|
||
worklist[wl_len] = ' '
|
||
wl_len += 1
|
||
worklist[wl_len] = '\0'
|
||
# json
|
||
if HandlesImports.is_module_imported(worklist, "json") == 0:
|
||
bsha1_j: str = _FindSha1ByModName(sha1_arr, mod_arr, map_count, "json")
|
||
if bsha1_j is not None:
|
||
if wl_len + 6 < 8192:
|
||
if wl_len > 0:
|
||
worklist[wl_len] = ' '
|
||
wl_len += 1
|
||
string.strcpy(worklist + wl_len, "json")
|
||
wl_len += 4
|
||
worklist[wl_len] = ' '
|
||
wl_len += 1
|
||
worklist[wl_len] = '\0'
|
||
fb: t.CChar | t.CPtr = VLogger.fmt_buf()
|
||
if fb is not None:
|
||
viperlib.snprintf(fb, 1024, "预加入容器模块后: '%s'", worklist)
|
||
VLogger.debug(fb, "Reachable")
|
||
|
||
# 3. 工作列表算法：递归收集可达 SHA1
|
||
processed: bytes = stdlib.malloc(8192)
|
||
if processed is None:
|
||
stdlib.free(sha1_arr)
|
||
stdlib.free(mod_arr)
|
||
stdlib.free(worklist)
|
||
return -1
|
||
processed[0] = '\0'
|
||
proc_len: t.CSizeT = 0
|
||
|
||
wl_pos: t.CSizeT = 0
|
||
wl_iter: int = 0
|
||
while wl_pos < wl_len:
|
||
wl_iter += 1
|
||
# 提取一个模块名（到空格或末尾）
|
||
name_start: t.CSizeT = wl_pos
|
||
while wl_pos < wl_len and worklist[wl_pos] != ' ':
|
||
wl_pos += 1
|
||
name_len: t.CSizeT = wl_pos - name_start
|
||
if wl_pos < wl_len:
|
||
wl_pos += 1
|
||
if name_len == 0:
|
||
continue
|
||
|
||
# 复制模块名
|
||
mod_buf: bytes = stdlib.malloc(name_len + 1)
|
||
if mod_buf is None:
|
||
continue
|
||
string.strncpy(mod_buf, worklist + name_start, name_len)
|
||
mod_buf[name_len] = '\0'
|
||
|
||
# 检查是否已处理
|
||
already: int = 0
|
||
if HandlesImports.is_module_imported(processed, mod_buf) != 0:
|
||
already = 1
|
||
if already == 0:
|
||
# 加入 processed（嵌套 if 拆分 + 用 = 代替 += 绕过 TPC AUGASGN bug）
|
||
if processed is not None:
|
||
if proc_len + name_len + 1 < 8192:
|
||
string.strcpy(processed + proc_len, mod_buf)
|
||
proc_len = proc_len + name_len
|
||
processed[proc_len] = ' '
|
||
proc_len = proc_len + 1
|
||
processed[proc_len] = '\0'
|
||
|
||
# 查找 SHA1
|
||
found_sha1: str = _FindSha1ByModName(sha1_arr, mod_arr, map_count, mod_buf)
|
||
if found_sha1 is not None:
|
||
# 加入 reachable_set
|
||
reachable_count = _AddSha1ToSet(reachable_set, reachable_count, found_sha1)
|
||
|
||
# 读取 .deps.txt 追加到 worklist
|
||
deps_path: str = _sliced_path(temp_dir, td_len_r, found_sha1, "deps.txt")
|
||
if deps_path is not None:
|
||
df: fileio.File | t.CPtr = fileio.File(deps_path, fileio.MODE.R)
|
||
if not df.closed:
|
||
deps_buf: bytes = stdlib.malloc(2048)
|
||
if deps_buf is not None:
|
||
dbr: t.CInt64T = df.read_all(deps_buf, 2048)
|
||
df.close()
|
||
if dbr > 0:
|
||
if dbr < 2048:
|
||
deps_buf[dbr] = '\0'
|
||
else:
|
||
deps_buf[2047] = '\0'
|
||
# 追加到 worklist（嵌套 if 拆分 + 用 = 代替 += 绕过 TPC AUGASGN bug）
|
||
dl: t.CSizeT = string.strlen(deps_buf)
|
||
if worklist is not None:
|
||
if dl > 0 and wl_len + dl + 2 < 8192:
|
||
if wl_len > 0 and worklist[wl_len - 1] != ' ':
|
||
worklist[wl_len] = ' '
|
||
wl_len = wl_len + 1
|
||
string.strcpy(worklist + wl_len, deps_buf)
|
||
wl_len = wl_len + dl
|
||
worklist[wl_len] = ' '
|
||
wl_len = wl_len + 1
|
||
worklist[wl_len] = '\0'
|
||
stdlib.free(deps_buf)
|
||
else:
|
||
df.close()
|
||
stdlib.free(deps_path)
|
||
stdlib.free(mod_buf)
|
||
stdlib.free(sha1_arr)
|
||
stdlib.free(mod_arr)
|
||
stdlib.free(worklist)
|
||
stdlib.free(processed)
|
||
|
||
fb: t.CChar | t.CPtr = VLogger.fmt_buf()
|
||
if fb is not None:
|
||
viperlib.snprintf(fb, 1024, "可达 SHA1: %d 个", reachable_count)
|
||
VLogger.debug(fb, "Reachable")
|
||
return reachable_count
|
||
|
||
|
||
# ============================================================
|
||
# BuildCombinedIR - 组合本地 stub + 依赖 stubs + 本地 text → 完整 IR
|
||
#
|
||
# 按需加载：读取 deps.txt 获取导入模块名，通过内存存储器查找 SHA1，
|
||
# 仅加载实际使用的依赖 stub（而非全部扫描）。
|
||
# _sha1_map.txt 仅作为人类可读输出，程序内部不读取。
|
||
# ============================================================
|
||
def BuildCombinedIR(temp_dir: str, local_sha1: str,
|
||
out_buf: bytes, out_size: t.CSizeT) -> t.CSizeT:
|
||
"""组合本地 stub + 依赖 stubs + 本地 text → out_buf（按需加载）"""
|
||
if temp_dir is None or local_sha1 is None or out_buf is None or out_size == 0:
|
||
return 0
|
||
|
||
td_len: t.CSizeT = string.strlen(temp_dir)
|
||
out_buf[0] = '\0'
|
||
out_pos: t.CSizeT = 0
|
||
|
||
# 1. 读取并追加本地 stub.ll（含 header）
|
||
stub_path: str = _sliced_path(temp_dir, td_len, local_sha1, "stub.ll")
|
||
if stub_path is None:
|
||
return 0
|
||
sf: fileio.File | t.CPtr = fileio.File(stub_path, fileio.MODE.R)
|
||
if sf.closed:
|
||
stdlib.free(stub_path)
|
||
return 0
|
||
stub_content: bytes = stdlib.malloc(STUB_READ_BUF_SIZE)
|
||
if stub_content is None:
|
||
sf.close()
|
||
stdlib.free(stub_path)
|
||
return 0
|
||
stub_br: t.CInt64T = sf.read_all(stub_content, STUB_READ_BUF_SIZE)
|
||
sf.close()
|
||
stdlib.free(stub_path)
|
||
if stub_br <= 0:
|
||
stdlib.free(stub_content)
|
||
return 0
|
||
if stub_br < STUB_READ_BUF_SIZE:
|
||
stub_content[stub_br] = '\0'
|
||
else:
|
||
stub_content[STUB_READ_BUF_SIZE - 1] = '\0'
|
||
slen: t.CSizeT = stub_br
|
||
# 修复本地 stub.ll 中字符串常量的非法 IR：
|
||
# llvmlite OUTPUT_STUB 模式输出 "= external unnamed_addr constant" 的非法形式。
|
||
# 简单方案：搜索 "= external unnamed_addr constant" 并将 external 替换为 8 个空格。
|
||
fix_pos: t.CSizeT = 0
|
||
while fix_pos + 30 < slen:
|
||
ext_ptr: t.CPtr = string.strstr(stub_content + fix_pos, "= external unnamed_addr constant")
|
||
if ext_ptr is None:
|
||
break
|
||
# ext_ptr 指向 "= external ..." 中的 '='
|
||
# 'external' 开始于 ext_ptr + 2
|
||
ext_off: t.CSizeT = t.CSizeT(t.CUInt64T(ext_ptr) - t.CUInt64T(stub_content)) + 2
|
||
if ext_off + 8 <= slen:
|
||
stub_content[ext_off] = ' '
|
||
stub_content[ext_off + 1] = ' '
|
||
stub_content[ext_off + 2] = ' '
|
||
stub_content[ext_off + 3] = ' '
|
||
stub_content[ext_off + 4] = ' '
|
||
stub_content[ext_off + 5] = ' '
|
||
stub_content[ext_off + 6] = ' '
|
||
stub_content[ext_off + 7] = ' '
|
||
fix_pos = ext_off + 8
|
||
# 直接复制修复后的 stub 内容
|
||
if out_pos + slen + 2 < out_size:
|
||
string.strcpy(out_buf + out_pos, stub_content)
|
||
out_pos += slen
|
||
if out_buf[out_pos - 1] != '\n':
|
||
out_buf[out_pos] = '\n'
|
||
out_pos += 1
|
||
out_buf[out_pos] = '\0'
|
||
stdlib.free(stub_content)
|
||
|
||
# 2. 按需加载依赖 stubs（根据 deps.txt 过滤，而非扫描全部）
|
||
# 先加载依赖 stub（含 type 定义），确保 type 定义在本地 text.ll 的 define 块之前
|
||
# 2a. 构建 includes SHA1 → module_name 映射（从全局内存存储器）
|
||
sha1_arr: bytes = stdlib.malloc(MAX_INCLUDES * 17)
|
||
mod_arr: bytes = stdlib.malloc(MAX_INCLUDES * 64)
|
||
inc_count: int = 0
|
||
if sha1_arr is not None and mod_arr is not None:
|
||
inc_count = _BuildIncludesSha1Map(temp_dir, td_len, sha1_arr, mod_arr)
|
||
|
||
# 2b. 读取 deps.txt 获取导入模块名集合
|
||
deps_buf: str = None
|
||
deps_loaded: int = 0
|
||
deps_path: str = _sliced_path(temp_dir, td_len, local_sha1, "deps.txt")
|
||
if deps_path is not None:
|
||
df_deps: fileio.File | t.CPtr = fileio.File(deps_path, fileio.MODE.R)
|
||
if not df_deps.closed:
|
||
deps_buf = stdlib.malloc(4096)
|
||
if deps_buf is not None:
|
||
deps_br: t.CInt64T = df_deps.read_all(deps_buf, 4096)
|
||
if deps_br > 0:
|
||
if deps_br < 4096:
|
||
deps_buf[deps_br] = '\0'
|
||
else:
|
||
deps_buf[4095] = '\0'
|
||
deps_loaded = 1
|
||
df_deps.close()
|
||
stdlib.free(deps_path)
|
||
|
||
# 2c. 递归扫描 temp_dir 中所有 .stub.ll，按需加载
|
||
all_sha1_arr: str = stdlib.malloc(MAX_INCLUDES_SHA1 * 17)
|
||
all_count: int = 0
|
||
if all_sha1_arr is not None:
|
||
string.memset(all_sha1_arr, 0, MAX_INCLUDES_SHA1 * 17)
|
||
all_count = _collect_stub_sha1s(temp_dir, td_len, Config.Sha1SliceLevel, all_sha1_arr, 0, MAX_INCLUDES_SHA1)
|
||
if all_count > 0:
|
||
dep_buf: bytes = stdlib.malloc(STUB_READ_BUF_SIZE)
|
||
if dep_buf is not None:
|
||
si: int = 0
|
||
while si < all_count:
|
||
dep_sha1: str = all_sha1_arr + t.CSizeT(si) * 17
|
||
if string.strcmp(dep_sha1, local_sha1) != 0:
|
||
should_load: int = 0
|
||
# 检查是否在 includes 映射中
|
||
is_include: int = 0
|
||
if inc_count > 0:
|
||
for ii in range(inc_count):
|
||
idx_ii: t.CSizeT = t.CSizeT(ii) * 17
|
||
if string.strcmp(sha1_arr + idx_ii, dep_sha1) == 0:
|
||
is_include = 1
|
||
if deps_loaded != 0:
|
||
idx_mi: t.CSizeT = t.CSizeT(ii) * 64
|
||
dep_mod_name: str = mod_arr + idx_mi
|
||
if HandlesImports.is_module_imported(deps_buf, dep_mod_name) != 0:
|
||
should_load = 1
|
||
else:
|
||
# 传递性依赖：deps.txt 只记录直接导入（如 ast），
|
||
# 但包的 __init__.py 内部 from .lexer import Lexer
|
||
# 会让 IR 引用 ast.lexer 子模块类型。
|
||
# 检查 dep_mod_name 是否是已导入包的子模块
|
||
# （如 ast.lexer 是 ast 的子模块）
|
||
if _is_submodule_of_imported(deps_buf, dep_mod_name) != 0:
|
||
should_load = 1
|
||
break
|
||
if is_include == 0:
|
||
# 非 includes stub（用户文件），总是加载
|
||
should_load = 1
|
||
if should_load != 0:
|
||
out_pos = _LoadAndAppendStub(temp_dir, td_len, dep_sha1, dep_buf, out_buf, out_size, out_pos)
|
||
out_pos = _LoadAndAppendTextDeclares(temp_dir, td_len, dep_sha1, dep_buf, out_buf, out_size, out_pos)
|
||
si += 1
|
||
stdlib.free(dep_buf)
|
||
stdlib.free(all_sha1_arr)
|
||
|
||
# 2d. 验证 deps.txt 中所有依赖模块的 stub 文件都存在（fail-fast）
|
||
# 避免到 llc 才报 undefined value 错误。
|
||
# 仅验证在 includes 映射中找到的模块（t/c 等内部库无 SHA1 映射，自动跳过）。
|
||
if deps_loaded != 0 and inc_count > 0:
|
||
dep_total: t.CSizeT = string.strlen(deps_buf)
|
||
dp: t.CSizeT = 0
|
||
while dp < dep_total:
|
||
# 跳过前导空格
|
||
while dp < dep_total:
|
||
if deps_buf[dp] != ' ':
|
||
break
|
||
dp += 1
|
||
if dp >= dep_total:
|
||
break
|
||
# 提取模块名（到下一个空格或末尾）
|
||
ms: t.CSizeT = dp
|
||
while dp < dep_total:
|
||
if deps_buf[dp] == ' ':
|
||
break
|
||
dp += 1
|
||
ml: t.CSizeT = dp - ms
|
||
if ml == 0:
|
||
continue
|
||
# 复制模块名到临时缓冲区并 null 终止
|
||
mod_nm: str = stdlib.malloc(ml + 1)
|
||
if mod_nm is None:
|
||
continue
|
||
string.strncpy(mod_nm, deps_buf + ms, ml)
|
||
mod_nm[ml] = '\0'
|
||
# 在 includes 映射中查找模块对应的 SHA1
|
||
found_s: str = None
|
||
for mi2 in range(inc_count):
|
||
idx_m2: t.CSizeT = t.CSizeT(mi2) * 64
|
||
if string.strcmp(mod_arr + idx_m2, mod_nm) == 0:
|
||
idx_s2: t.CSizeT = t.CSizeT(mi2) * 17
|
||
found_s = sha1_arr + idx_s2
|
||
break
|
||
if found_s is not None:
|
||
# 检查 stub 文件是否存在
|
||
chk_path: str = _sliced_path(temp_dir, td_len, found_s, "stub.ll")
|
||
if chk_path is not None:
|
||
chk_f: fileio.File | t.CPtr = fileio.File(chk_path, fileio.MODE.R)
|
||
if chk_f.closed:
|
||
fb: t.CChar | t.CPtr = VLogger.fmt_buf()
|
||
if fb is not None:
|
||
viperlib.snprintf(fb, 1024, "依赖模块 '%s' (sha1=%s) 的 stub 文件不存在: %s，立即终止编译", mod_nm, found_s, chk_path)
|
||
VLogger.error(fb, "BuildCombinedIR")
|
||
sys.exit(1)
|
||
chk_f.close()
|
||
stdlib.free(chk_path)
|
||
stdlib.free(mod_nm)
|
||
|
||
# 3. 读取并追加本地 text.ll（行级去重，跳过 stub.ll 中已存在的定义）
|
||
text_path: str = _sliced_path(temp_dir, td_len, local_sha1, "text.ll")
|
||
if text_path is None:
|
||
return out_pos
|
||
tf: fileio.File | t.CPtr = fileio.File(text_path, fileio.MODE.R)
|
||
if tf.closed:
|
||
stdlib.free(text_path)
|
||
# text.ll 不存在的情况，继续
|
||
else:
|
||
text_content: bytes = stdlib.malloc(STUB_READ_BUF_SIZE)
|
||
if text_content is None:
|
||
tf.close()
|
||
stdlib.free(text_path)
|
||
else:
|
||
text_br: t.CInt64T = tf.read_all(text_content, STUB_READ_BUF_SIZE)
|
||
tf.close()
|
||
stdlib.free(text_path)
|
||
if text_br > 0:
|
||
if text_br < STUB_READ_BUF_SIZE:
|
||
text_content[text_br] = '\0'
|
||
else:
|
||
text_content[STUB_READ_BUF_SIZE - 1] = '\0'
|
||
# 行级去重追加（与 _LoadAndAppendStub 相同逻辑）
|
||
tx_pos: t.CSizeT = 0
|
||
tx_skip: int = 1
|
||
while tx_pos < text_br:
|
||
tx_ls: t.CSizeT = tx_pos
|
||
while tx_pos < text_br:
|
||
if text_content[tx_pos] == '\n':
|
||
break
|
||
tx_pos += 1
|
||
tx_ll: t.CSizeT = tx_pos - tx_ls
|
||
if tx_pos < text_br:
|
||
tx_pos += 1
|
||
if tx_skip != 0:
|
||
if tx_ll == 0:
|
||
continue
|
||
tx_c0: int = text_content[tx_ls]
|
||
if tx_c0 == ';':
|
||
continue
|
||
if tx_c0 == 't' and tx_ll >= 6:
|
||
if string.strncmp(text_content + tx_ls, "target", 6) == 0:
|
||
continue
|
||
if tx_c0 == 's' and tx_ll >= 15:
|
||
if string.strncmp(text_content + tx_ls, "source_filename", 15) == 0:
|
||
continue
|
||
tx_skip = 0
|
||
# 去重：declare（用 _IsFuncDeclaredOrDefined 避免误匹配 call 语句）
|
||
if tx_ll >= 8 and string.strncmp(text_content + tx_ls, "declare ", 8) == 0:
|
||
tx_at: t.CSizeT = tx_ls + 8
|
||
while tx_at < tx_ls + tx_ll:
|
||
if text_content[tx_at] == '@':
|
||
break
|
||
tx_at += 1
|
||
if tx_at < tx_ls + tx_ll:
|
||
tx_lp: t.CSizeT = tx_at + 1
|
||
while tx_lp < tx_ls + tx_ll:
|
||
if text_content[tx_lp] == '(':
|
||
break
|
||
tx_lp += 1
|
||
if tx_lp < tx_ls + tx_ll:
|
||
tx_sv: t.CChar = text_content[tx_lp + 1]
|
||
text_content[tx_lp + 1] = '\0'
|
||
if _IsFuncDeclaredOrDefined(out_buf, text_content + tx_at) != 0:
|
||
text_content[tx_lp + 1] = tx_sv
|
||
continue
|
||
text_content[tx_lp + 1] = tx_sv
|
||
# 去重：define（注释掉 out_buf 中已有的 declare，避免 declare/define 冲突）
|
||
if tx_ll >= 7 and string.strncmp(text_content + tx_ls, "define ", 7) == 0:
|
||
tx_at_d: t.CSizeT = tx_ls + 7
|
||
while tx_at_d < tx_ls + tx_ll:
|
||
if text_content[tx_at_d] == '@':
|
||
break
|
||
tx_at_d += 1
|
||
if tx_at_d < tx_ls + tx_ll:
|
||
tx_lp_d: t.CSizeT = tx_at_d + 1
|
||
while tx_lp_d < tx_ls + tx_ll:
|
||
if text_content[tx_lp_d] == '(':
|
||
break
|
||
tx_lp_d += 1
|
||
if tx_lp_d < tx_ls + tx_ll:
|
||
tx_sv_d: t.CChar = text_content[tx_lp_d + 1]
|
||
text_content[tx_lp_d + 1] = '\0'
|
||
_CommentOutDeclareInBuf(out_buf, text_content + tx_at_d)
|
||
text_content[tx_lp_d + 1] = tx_sv_d
|
||
# 处理：@global = ...
|
||
# stub 中可能是 external 声明或完整定义，以 text 中的定义为准
|
||
# 找到 stub 中同名全局变量行并注释掉，然后追加 text 中的
|
||
if tx_ll > 0 and text_content[tx_ls] == '@':
|
||
tx_eq: t.CSizeT = tx_ls + 1
|
||
while tx_eq < tx_ls + tx_ll:
|
||
if text_content[tx_eq] == '=':
|
||
break
|
||
tx_eq += 1
|
||
if tx_eq < tx_ls + tx_ll:
|
||
tx_sv2: t.CChar = text_content[tx_eq + 1]
|
||
text_content[tx_eq + 1] = '\0'
|
||
_CommentOutGlobalInBuf(out_buf, text_content + tx_ls)
|
||
text_content[tx_eq + 1] = tx_sv2
|
||
# 去重：%type = ...
|
||
if tx_ll > 0 and text_content[tx_ls] == '%':
|
||
tx_eq2: t.CSizeT = tx_ls + 1
|
||
while tx_eq2 < tx_ls + tx_ll:
|
||
if text_content[tx_eq2] == '=':
|
||
break
|
||
tx_eq2 += 1
|
||
if tx_eq2 < tx_ls + tx_ll:
|
||
tx_sv3: t.CChar = text_content[tx_eq2 + 1]
|
||
text_content[tx_eq2 + 1] = '\0'
|
||
if string.strstr(out_buf, text_content + tx_ls) is not None:
|
||
text_content[tx_eq2 + 1] = tx_sv3
|
||
continue
|
||
text_content[tx_eq2 + 1] = tx_sv3
|
||
# 追加
|
||
if out_pos + tx_ll + 2 < out_size:
|
||
string.strncpy(out_buf + out_pos, text_content + tx_ls, tx_ll)
|
||
out_pos += tx_ll
|
||
out_buf[out_pos] = '\n'
|
||
out_pos += 1
|
||
out_buf[out_pos] = '\0'
|
||
stdlib.free(text_content)
|
||
|
||
if deps_buf is not None:
|
||
stdlib.free(deps_buf)
|
||
if sha1_arr is not None:
|
||
stdlib.free(sha1_arr)
|
||
if mod_arr is not None:
|
||
stdlib.free(mod_arr)
|
||
|
||
return out_pos
|