diff options
| author | Elliott Hughes <enh@google.com> | 2024-04-08 19:57:59 +0000 |
|---|---|---|
| committer | Elliott Hughes <enh@google.com> | 2024-04-08 19:57:59 +0000 |
| commit | c5acbd0bdb3e407a1ce1a4857ac5e88032c9deb4 (patch) | |
| tree | c3625e43ca667c1b761ebce1ee7e67971d900708 | |
| parent | 956d6939bf93f762c20c3e07e75398b9487403a9 (diff) | |
| download | xz-embedded-c5acbd0bdb3e407a1ce1a4857ac5e88032c9deb4.tar.gz | |
Disable BCJ filters.
It's just a coincidence that we happened to wonder about these just
before the recent https://en.wikipedia.org/wiki/XZ_Utils_backdoor
incident, and the results came in just after!
Extra code we don't need, and extra runtime that doesn't win us any
size benefits --- it would be embarrassing to feel compelled to do
the riscv64 work for "feature parity" with a useless feature...
Bug: http://b/329112384
Change-Id: If258d57ab8a860ede52eb124e04434ba647734aa
| -rw-r--r-- | Android.bp | 10 |
1 files changed, 5 insertions, 5 deletions
@@ -29,12 +29,12 @@ cc_library_static { ], local_include_dirs: ["userspace"], - // Enable branch/call/jump filters. See http://b/27817327. + // We don't enable branch/call/jump filters because they have no impact on + // Android OTA sizes (measured in http://b/329112384). The frequent + // function calls this is supposed to optimize will all go via the PLT, so + // the actual branch/call/jump instructions in the ELF file will be zeroed + // out anyway. cflags: [ - "-DXZ_DEC_X86", - "-DXZ_DEC_ARM", - "-DXZ_DEC_ARM64", - "-DXZ_DEC_ARMTHUMB", "-Wall", "-Werror", ], |