Files
mongo/buildscripts/convert_bazel_headers.py
Daniel Moody 00da7f4428 SERVER-94055 fix bazel header list gen to not list gen headers redundantly (#26535)
GitOrigin-RevId: ef54bd205937b736caf6053a899e505a1d77638c
2024-08-27 17:16:17 +00:00

372 lines
14 KiB
Python

import subprocess
import sys
import json
import platform
import os
import shlex
import concurrent.futures
import glob
import traceback
import shutil
from typing import Annotated, List
import typer
def work(target_library: str, silent: bool, cpu_count: int, cc: List[str]):
headers = set()
original_headers = set()
def get_headers(line):
nonlocal headers
try:
with open(target_library + ".bazel_headers") as f:
bazel_headers = [line.strip() for line in f.readlines()]
bazel_headers += [
"src/mongo/platform/basic.h",
"src/mongo/platform/windows_basic.h",
]
with open(target_library + ".env_vars") as f:
tmp_env_vars = json.load(f)
env_vars = {}
# subprocess requies only strings
for k, v in tmp_env_vars.items():
env_vars[str(k)] = str(v)
for command in cc:
cmd_output = command["output"].replace("\\", "/").strip("'").strip('"')
line_output = line.replace("\\", "/")
if cmd_output == line_output:
os.makedirs(os.path.dirname(line), exist_ok=True)
if os.name == "nt":
header_arg = " /showIncludes"
else:
header_arg = " -H"
if not silent:
print(f"compiling {line}")
p = subprocess.run(
shlex.split((command["command"].replace("\\", "/") + header_arg)),
env=env_vars,
capture_output=True,
text=True,
)
if p.returncode != 0:
print(f"Error compiling, exitcode: {p.returncode}", file=sys.stderr)
print(f"STDOUT: {p.stdout}", file=sys.stderr)
print(f"STDERR: {p.stderr}", file=sys.stderr)
sys.exit(1)
if os.name == "nt":
for line in p.stdout.split("\n"):
line = (
line.replace("Note: including file:", "")
.strip(" ")
.replace("\\", "/")
)
if not line.startswith(os.getcwd().replace("\\", "/")):
continue
line = os.path.relpath(
line, start=os.getcwd().replace("\\", "/")
).replace("\\", "/")
if line not in bazel_headers:
if line.startswith("src/") or line.startswith("bazel-out/"):
original_headers.add(line)
line = "//" + line
line = ":".join(line.rsplit("/", 1))
headers.add(line)
else:
for line in p.stderr.split("\n"):
if ". src/" in line or ". bazel-out/" in line:
while line.startswith("."):
line = line[1:]
line = line.replace("\\", "/")
if line[1:] not in bazel_headers:
original_headers.add(line[1:])
line = "//" + line[1:]
line = ":".join(line.rsplit("/", 1))
headers.add(line)
except Exception as exc:
print(traceback.format_exc(), file=sys.stderr)
raise exc
sources = []
with open(target_library + ".obj_files") as f:
lines = f.readlines()
for line in lines:
line = line.strip()
line = line.replace("build/opt", "//src")
line = line[: line.find(".")] + ".cpp"
src_header = os.path.splitext(line[2:])[0] + ".h"
if os.path.exists(src_header):
src_header = "//" + ":".join(src_header.rsplit("/", 1))
headers.add(src_header)
line = ":".join(line.rsplit("/", 1))
if line.endswith("_gen.cpp"):
line = line[:-4]
sources.append(line)
with concurrent.futures.ThreadPoolExecutor(max_workers=cpu_count) as executor:
jobs = {executor.submit(get_headers, line.strip()): line.strip() for line in lines}
for completed_job in concurrent.futures.as_completed(jobs):
if not silent:
print(f"finished {jobs[completed_job]}")
with open(".bazel_include_info.json") as f:
bazel_include_info = json.load(f)
header_map = bazel_include_info["header_map"]
gen_header_map = bazel_include_info["gen_header_map"]
bazel_exec = bazel_include_info["bazel_exec"]
bazel_config = bazel_include_info["config"]
global_headers = (
"src/mongo:config.h",
"src/mongo/config.h",
"src/mongo/platform/basic.h",
"src/mongo/platform/windows_basic.h",
)
reverse_header_map = {}
for k, v in header_map.items():
for hdr in v:
if not hdr or hdr.endswith(global_headers):
continue
bazel_header = "//" + hdr.replace("\\", "/")
bazel_header = ":".join(bazel_header.rsplit("/", 1))
if bazel_header.startswith("//src/third_party/SafeInt"):
reverse_header_map[bazel_header] = ["//src/third_party/SafeInt:headers"]
elif bazel_header.startswith("//src/third_party/immer"):
reverse_header_map[bazel_header] = ["//src/third_party/immer:headers"]
elif bazel_header in reverse_header_map:
if bazel_header.startswith("//src/third_party/"):
continue
reverse_header_map[bazel_header].append(k)
else:
reverse_header_map[bazel_header] = [k]
for k, v in gen_header_map.items():
for hdr in v:
if not hdr or hdr.endswith(global_headers):
continue
bazel_header = "//" + hdr.replace("\\", "/")
bazel_header = ":".join(bazel_header.rsplit("/", 1))
if bazel_header not in reverse_header_map:
reverse_header_map[bazel_header] = [k]
else:
reverse_header_map[bazel_header].append(k)
recommended_deps = set()
minimal_headers = set()
basename_sources = [os.path.splitext(src.rsplit(":", 1)[1])[0] for src in sources]
for header in headers:
header_basename = os.path.splitext(header.rsplit(":", 1)[1])[0]
if header_basename in basename_sources:
minimal_headers.add(header)
continue
if header in reverse_header_map:
found = False
for lib in reverse_header_map[header]:
recommended_deps.add(lib)
else:
if not header.endswith(global_headers):
minimal_headers.add(header)
deps_order_by_height = []
deps_queries = {}
with open(target_library + ".bazel_deps") as f:
original_deps = [line.strip() for line in f.readlines()]
for dep in recommended_deps | set(original_deps):
p = subprocess.run(
[bazel_exec, "cquery"]
+ bazel_config
+ [f'kind("extract_debuginfo|idl_generator|render_template", deps("@{dep}"))'],
capture_output=True,
text=True,
)
deps_queries[dep] = [
line.split(" ")[0] for line in p.stdout.splitlines() if line.startswith("//")
]
deps_order_by_height.append((dep, len(deps_queries[dep])))
deps_order_by_height.sort(key=lambda x: x[1])
deps_order_by_height = [dep[0] for dep in deps_order_by_height]
optimal_header_deps = set()
for header in headers:
if header in minimal_headers:
continue
path_header = "/".join(header.rsplit(":", 1))
path_header = path_header[2:]
for dep in deps_order_by_height:
if dep in header_map and path_header in header_map[dep]:
optimal_header_deps.add(dep)
break
found = False
for other_dep in deps_order_by_height:
if other_dep in gen_header_map:
continue
if dep in deps_queries[other_dep]:
optimal_header_deps.add(other_dep)
found = True
break
if found:
continue
if dep in gen_header_map:
minimal_headers.add(dep)
else:
raise Exception(
f"Should not happen, did not find way to add dep {dep} for {target_library}"
)
optimal_header_deps = list(optimal_header_deps)
working_deps = optimal_header_deps.copy()
for dep in optimal_header_deps:
if dep in working_deps:
for test_dep in optimal_header_deps:
if test_dep == dep:
continue
if test_dep in working_deps and test_dep in deps_queries[dep]:
working_deps.remove(test_dep)
link_deps = []
header_deps = []
for dep in sorted(list(set(list(working_deps) + list(set(original_deps))))):
if dep in original_deps:
link_deps.append(dep)
else:
header_deps.append(dep)
target_name = os.path.splitext(os.path.basename(target_library))[0]
if target_name.startswith("lib"):
target_name = target_name[3:]
bazel_target = f"{target_library}\n"
bazel_target += "=" * 50 + "\n"
local_bazel_path = os.path.dirname(target_library.replace("build/opt", "//src")) + ":"
bazel_target += "mongo_cc_library(\n"
bazel_target += f' name = "{target_name}",\n'
if sources:
bazel_target += " srcs = [\n"
for src in sorted([src.replace(local_bazel_path, "") for src in sources]):
bazel_target += f' "{src}",\n'
bazel_target += " ],\n"
if minimal_headers:
bazel_target += " hdrs = [\n"
for header in sorted([header.replace(local_bazel_path, "") for header in minimal_headers]):
bazel_target += f' "{header}",\n'
bazel_target += " ],\n"
if header_deps:
bazel_target += " header_deps = [\n"
for dep in sorted([dep.strip().replace(local_bazel_path, "") for dep in header_deps]):
bazel_target += f' "{dep}",\n'
bazel_target += " ],\n"
if link_deps:
bazel_target += " deps = [\n"
for dep in sorted([dep.strip().replace(local_bazel_path, "") for dep in link_deps]):
bazel_target += f' "{dep}",\n'
bazel_target += " ],\n"
bazel_target += ")\n"
return bazel_target
def main(
target_libraries: Annotated[List[str], typer.Argument()],
silent: Annotated[bool, typer.Option()] = False,
skip_scons: Annotated[bool, typer.Option()] = False,
debug_mode: Annotated[bool, typer.Option()] = False,
):
extra_args = []
if os.name == "nt":
extra_args += [
"CPPPATH=C:\sasl\include",
"LIBPATH=C:\sasl\lib",
]
target_library = os.path.join(
os.path.dirname(target_library), os.path.basename(target_library)[3:-2] + "lib"
)
path = shutil.which("icecc")
if path is None:
extra_args += ["ICECC="]
if os.name == "nt":
target_fmt = lambda target_library: os.path.join(
os.path.dirname(target_library), os.path.basename(target_library)[3:-2] + "lib"
)
elif platform.system() == "Darwin":
target_fmt = lambda target_library: target_library[:-2] + "a"
else:
target_fmt = lambda x: None
map(target_fmt, target_libraries)
cmd = [
sys.executable,
"buildscripts/scons.py",
"--build-profile=opt",
" ".join(
[f"--bazel-includes-info={target_library}" for target_library in target_libraries]
),
"--libdeps-linting=off",
"--ninja=disabled",
"compiledb",
] + extra_args
if not skip_scons:
p = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True)
while True:
line = p.stdout.readline()
if not line:
break
print(line.strip(), file=sys.stderr)
_, _ = p.communicate()
if p.returncode != 0:
print(f"SCons build failed, exit code {p.returncode}", file=sys.stderr)
sys.exit(1)
with open("compile_commands.json") as f:
cc = json.load(f)
if platform.system() == "Linux":
cpu_count = len(os.sched_getaffinity(0)) + 4
else:
cpu_count = os.cpu_count() + 4
# Process pool makes it harder to debug what is happening
# so for debug mode, we disabled process pool so things happen in order
# you can just print from the process.
if debug_mode:
bazel_targets = []
for target_library in target_libraries:
bazel_targets += [work(target_library, silent, cpu_count, cc)]
else:
with concurrent.futures.ProcessPoolExecutor(max_workers=cpu_count) as executor:
jobs = {
executor.submit(work, target_library, silent, cpu_count, cc): target_library
for target_library in target_libraries
}
bazel_targets = [job.result() for job in concurrent.futures.as_completed(jobs)]
print("====== Bazel Targets ======\n")
print("\n".join(bazel_targets))
if __name__ == "__main__":
typer.run(main)