[antlir] move debuginfo splitter to its own file

Summary: Moving the split file into its own file and not as part of the bzl file... make edit and using linter much easier, also the testing and debugging story is much better this way

Test Plan:
```
❯ buck2 test fbcode//antlir/antlir2/features/install/... fbcode//antlir/antlir2/features/extract/...
Buck UI: https://www.internalfb.com/buck2/d6837da1-66d8-4051-b5fb-7d08cb435390
Tests finished: Pass 35. Fail 0. Fatal 0. Skip 0. Infra Failure 0. Build failure 0
```
https://www.internalfb.com/intern/testinfra/testrun/14636698917489311

Reviewed By: naveedgol

Differential Revision: D85864708

fbshipit-source-id: d8091c2406edbab1085f4f95088eeadf114d20cb

Author: vmagro

antlir/antlir2/bzl/debuginfo.bzl

@@ -33,127 +33,9 @@ def _split_binary_impl(ctx: AnalysisContext) -> list[Provider]:
     # to write
     objcopy_tmp = ctx.actions.declare_output("objcopy_tmp")
 
-    split = ctx.actions.write("split.py", """#!/usr/bin/env python3
-import argparse
-import json
-import os
-import shutil
-import subprocess
-import sys
-from pathlib import Path
-
-parser = argparse.ArgumentParser()
-parser.add_argument("--objcopy", required=True)
-parser.add_argument("--binary", required=True, type=Path)
-parser.add_argument("--binary-dwp", type=Path)
-parser.add_argument("--stripped", required=True, type=Path)
-parser.add_argument("--debuginfo", required=True, type=Path)
-parser.add_argument("--dwp", required=True, type=Path)
-parser.add_argument("--metadata", required=True, type=Path)
-parser.add_argument("--objcopy-tmp", required=True, type=Path)
-
-args = parser.parse_args()
-
-# ensure this exists or buck2 will get mad
-args.objcopy_tmp.touch()
-
-if args.binary.is_dir():
-    is_elf = False
-else:
-    with open(args.binary, mode="rb") as src_f:
-        first_4 = src_f.read(4)
-        is_elf = first_4 == b"\x7fELF"
-
-if args.binary_dwp:
-    shutil.copy2(args.binary_dwp, args.dwp)
-else:
-    with open(args.dwp, "w") as _f:
-        pass
-
-# If this is not an ELF binary, it can't be stripped so just copy the original
-if not is_elf:
-    if args.binary.is_dir():
-        shutil.copytree(args.binary, args.stripped, symlinks=True)
-    else:
-        shutil.copy2(args.binary, args.stripped)
-    with open(args.debuginfo, "w") as _f:
-        pass
-    with open(args.metadata, "w") as f:
-        json.dump({}, f)
-    sys.exit(0)
-
-# Save debug symbols to a separate debuginfo file
-proc = subprocess.run(
-    [
-        args.objcopy,
-        "--only-keep-debug",
-        args.binary,
-        args.debuginfo,
-    ],
-    capture_output=True,
-)
-if proc.returncode != 0:
-    raise RuntimeError("Failed to extract debug symbols for {}:\\n{}\\n{}".format(
-        args.binary,
-        proc.stdout.decode("utf-8", errors = "surrogateescape"),
-        proc.stderr.decode("utf-8", errors = "surrogateescape"),
-    ))
-
-# Remove the debug symbols from the stripped binary
-proc = subprocess.run(
-    [
-        args.objcopy,
-        "--strip-debug",
-        "--remove-section=.pseudo_probe",
-        "--remove-section=.pseudo_probe_desc",
-        args.binary,
-        args.stripped,
-    ],
-    capture_output=True,
-)
-if proc.returncode != 0:
-    raise RuntimeError("Failed to extract debug symbols for {}:\\n{}\\n{}".format(
-        args.binary,
-        proc.stdout.decode("utf-8", errors = "surrogateescape"),
-        proc.stderr.decode("utf-8", errors = "surrogateescape"),
-    ))
-
-# Find the BuildID of the binary. This determines where it should go for gdb to
-# look it up under /usr/lib/debug
-# https://sourceware.org/gdb/onlinedocs/gdb/Separate-Debug-Files.html
-buildid_proc = subprocess.run(
-    [
-        args.objcopy,
-        "--dump-section",
-        ".note.gnu.build-id=/dev/stdout",
-        args.binary,
-        args.objcopy_tmp,
-    ],
-    capture_output=True,
-)
-if buildid_proc.returncode != 0:
-    raise RuntimeError("Failed to get build-id for {}:\\n{}\\n{}".format(
-        args.binary,
-        buildid_proc.stdout.decode("utf-8", errors = "surrogateescape"),
-        buildid_proc.stderr.decode("utf-8", errors = "surrogateescape"),
-    ))
-buildid = buildid_proc.stdout
-
-# Prefer to install the debug info by BuildID since it does not require another
-# objcopy invocation and is more standard
-with open(args.metadata, "w") as f:
-    if buildid := buildid[len(buildid) - 20 :].hex():
-        json.dump({"buildid": buildid}, f)
-    else:
-        # Can't setup debuglink here as we don't reliably know the location the binary
-        # will end up being placed under, which debuglink relies on, so opt to no-op
-        # here and linking will ultimately be handled in the install feature.
-        json.dump({}, f)
-    """, is_executable = True)
-
     ctx.actions.run(
         cmd_args(
-            split,
+            ctx.attrs.debuginfo_splitter[RunInfo],
             cmd_args(objcopy, format = "--objcopy={}"),
             cmd_args(src, format = "--binary={}"),
             (cmd_args(src_dwp, format = "--binary-dwp={}") if src_dwp else []),
@@ -185,6 +67,7 @@ split_binary = anon_rule(
     impl = _split_binary_impl,
     attrs = {
         "cxx_toolchain": attrs.option(attrs.toolchain_dep(default = "toolchains//:cxx", providers = [CxxToolchainInfo]), default = None),
+        "debuginfo_splitter": attrs.exec_dep(default = "antlir//antlir/antlir2/tools:debuginfo-splitter"),
         "objcopy": attrs.option(attrs.exec_dep(), default = None),
         "src": attrs.dep(),
     },
@@ -200,8 +83,10 @@ def split_binary_anon(
         *,
         ctx: AnalysisContext,
         src: Dependency,
-        objcopy: Dependency) -> AnonTarget:
+        objcopy: Dependency,
+        debuginfo_splitter: Dependency) -> AnonTarget:
     return ctx.actions.anon_target(split_binary, {
+        "debuginfo_splitter": debuginfo_splitter,
         "name": "debuginfo//" + src.label.package + ":" + src.label.name,
         "objcopy": objcopy,
         "src": src,

antlir/antlir2/features/extract/extract.bzl

@@ -101,6 +101,7 @@ def extract_buck_binary(
         },
         exec_deps = {
             "_analyze": "antlir//antlir/antlir2/features/extract:extract-buck-binary-analyze",
+            "_debuginfo_splitter": "fbcode//antlir/antlir2/tools:debuginfo-splitter",
             "_objcopy": internal_external(
                 fb = "fbsource//third-party/binutils:objcopy",
                 oss = "toolchains//:objcopy",
@@ -141,6 +142,7 @@ def _extract_buck_binary_impl(ctx: AnalysisContext) -> list[Provider]:
             ctx = ctx,
             src = ctx.attrs.src,
             objcopy = ctx.attrs._objcopy,
+            debuginfo_splitter = ctx.attrs._debuginfo_splitter,
         )
         src = split_anon_target.artifact("src")
     else:
@@ -189,6 +191,7 @@ extract_buck_binary_rule = rule(
         "strip": attrs.bool(default = True),
         "target_arch": attrs.string(),
         "_analyze": attrs.exec_dep(),
+        "_debuginfo_splitter": attrs.option(attrs.exec_dep(), default = None),
         "_objcopy": attrs.option(attrs.exec_dep(), default = None),
     },
 )

antlir/antlir2/features/install/install.bzl

@@ -102,6 +102,7 @@ def install(
         fail("always_use_gnu_debuglink requires split_debuginfo=True")
 
     exec_deps = {
+        "_debuginfo_splitter": "fbcode//antlir/antlir2/tools:debuginfo-splitter",
         "_mac_signer": "fbcode//python/runtime/tools:recursive_mac_signer",
         "_objcopy": internal_external(
             fb = "fbsource//third-party/binutils:objcopy",
@@ -384,6 +385,7 @@ def _impl(ctx: AnalysisContext) -> list[Provider] | Promise:
                 ctx = ctx,
                 src = src,
                 objcopy = ctx.attrs._objcopy,
+                debuginfo_splitter = ctx.attrs._debuginfo_splitter,
             )
             binary_info = binary_record(
                 installed = installed_binary(
@@ -549,6 +551,7 @@ install_rule = rule(
         ),
         "xattrs": attrs.dict(attrs.string(), attrs.string(), default = {}),
         "_binaries_require_repo": binaries_require_repo.optional_attr,
+        "_debuginfo_splitter": attrs.option(attrs.exec_dep(), default = None),
         "_ensure_dir_exists_plugin": attrs.option(attrs.label(), default = None),
         "_mac_signer": attrs.option(attrs.exec_dep(), default = None),
         "_objcopy": attrs.option(attrs.exec_dep(), default = None),

antlir/antlir2/tools/BUCK

@@ -0,0 +1,7 @@
+oncall("antlir")
+
+python_bootstrap_binary(
+    name = "debuginfo-splitter",
+    main = "debuginfo-splitter.py",
+    visibility = ["PUBLIC"],
+)

antlir/antlir2/tools/debuginfo-splitter.py

@@ -0,0 +1,126 @@
+#!/usr/bin/env python3
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the MIT license found in the
+# LICENSE file in the root directory of this source tree.
+
+import argparse
+import json
+import shutil
+import subprocess
+import sys
+from pathlib import Path
+
+parser = argparse.ArgumentParser()
+parser.add_argument("--objcopy", required=True)
+parser.add_argument("--binary", required=True, type=Path)
+parser.add_argument("--binary-dwp", type=Path)
+parser.add_argument("--stripped", required=True, type=Path)
+parser.add_argument("--debuginfo", required=True, type=Path)
+parser.add_argument("--dwp", required=True, type=Path)
+parser.add_argument("--metadata", required=True, type=Path)
+parser.add_argument("--objcopy-tmp", required=True, type=Path)
+
+args = parser.parse_args()
+
+# ensure this exists or buck2 will get mad
+args.objcopy_tmp.touch()
+
+if args.binary.is_dir():
+    is_elf = False
+else:
+    with open(args.binary, mode="rb") as src_f:
+        first_4 = src_f.read(4)
+        is_elf = first_4 == b"\x7fELF"
+
+if args.binary_dwp:
+    shutil.copy2(args.binary_dwp, args.dwp)
+else:
+    with open(args.dwp, "w"):
+        pass
+
+# If this is not an ELF binary, it can't be stripped so just copy the original
+if not is_elf:
+    if args.binary.is_dir():
+        shutil.copytree(args.binary, args.stripped, symlinks=True)
+    else:
+        shutil.copy2(args.binary, args.stripped)
+    with open(args.debuginfo, "w"):
+        pass
+    with open(args.metadata, "w") as f:
+        json.dump({}, f)
+    sys.exit(0)
+
+# Save debug symbols to a separate debuginfo file
+proc = subprocess.run(
+    [
+        args.objcopy,
+        "--only-keep-debug",
+        args.binary,
+        args.debuginfo,
+    ],
+    capture_output=True,
+)
+if proc.returncode != 0:
+    raise RuntimeError(
+        "Failed to extract debug symbols for {}:\\n{}\\n{}".format(
+            args.binary,
+            proc.stdout.decode("utf-8", errors="surrogateescape"),
+            proc.stderr.decode("utf-8", errors="surrogateescape"),
+        )
+    )
+
+# Remove the debug symbols from the stripped binary
+proc = subprocess.run(
+    [
+        args.objcopy,
+        "--strip-debug",
+        "--remove-section=.pseudo_probe",
+        "--remove-section=.pseudo_probe_desc",
+        args.binary,
+        args.stripped,
+    ],
+    capture_output=True,
+)
+if proc.returncode != 0:
+    raise RuntimeError(
+        "Failed to extract debug symbols for {}:\\n{}\\n{}".format(
+            args.binary,
+            proc.stdout.decode("utf-8", errors="surrogateescape"),
+            proc.stderr.decode("utf-8", errors="surrogateescape"),
+        )
+    )
+
+# Find the BuildID of the binary. This determines where it should go for gdb to
+# look it up under /usr/lib/debug
+# https://sourceware.org/gdb/onlinedocs/gdb/Separate-Debug-Files.html
+buildid_proc = subprocess.run(
+    [
+        args.objcopy,
+        "--dump-section",
+        ".note.gnu.build-id=/dev/stdout",
+        args.binary,
+        args.objcopy_tmp,
+    ],
+    capture_output=True,
+)
+if buildid_proc.returncode != 0:
+    raise RuntimeError(
+        "Failed to get build-id for {}:\\n{}\\n{}".format(
+            args.binary,
+            buildid_proc.stdout.decode("utf-8", errors="surrogateescape"),
+            buildid_proc.stderr.decode("utf-8", errors="surrogateescape"),
+        )
+    )
+buildid = buildid_proc.stdout
+
+# Prefer to install the debug info by BuildID since it does not require another
+# objcopy invocation and is more standard
+with open(args.metadata, "w") as f:
+    if buildid := buildid[len(buildid) - 20 :].hex():
+        json.dump({"buildid": buildid}, f)
+    else:
+        # Can't setup debuglink here as we don't reliably know the location the binary
+        # will end up being placed under, which debuglink relies on, so opt to no-op
+        # here and linking will ultimately be handled in the install feature.
+        json.dump({}, f)