From 096defc40d15bf3d80d468b83233fb2e5804d8f2 Mon Sep 17 00:00:00 2001
From: Tony Aiuto <aiuto@google.com>
Date: Tue, 26 May 2020 14:48:43 -0400
Subject: [PATCH] Stop doing actions.write UTF-8.  This was causing a double
 encode.

Since Bazel reads BUILD and .bzl files as (essentially) raw bytes, any UTF-8
encoded input is essentially pre-UTF-8 encoded on the way out.

Closes #10174

DO NOT SUBMIT: This is a breaking change, so it needs a migration path.
Current idea:
- introduce a new attribute to write 'encode_as_utf_8'. Default is gated by incompatible_write_action_encodes_as_utf_8.
- for bazel we do the multi-release cycle flip
- for google we agressively try to change the default
  - add encode_as_utf_8 to the few broken rules
  - flip the default
  - fix the user.
---
 .../build/lib/analysis/actions/FileWriteAction.java   | 11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

diff --git a/src/main/java/com/google/devtools/build/lib/analysis/actions/FileWriteAction.java b/src/main/java/com/google/devtools/build/lib/analysis/actions/FileWriteAction.java
index 5cb26b47ee09dd..2b19ac2715382e 100644
--- a/src/main/java/com/google/devtools/build/lib/analysis/actions/FileWriteAction.java
+++ b/src/main/java/com/google/devtools/build/lib/analysis/actions/FileWriteAction.java
@@ -14,8 +14,6 @@
 
 package com.google.devtools.build.lib.analysis.actions;
 
-import static java.nio.charset.StandardCharsets.UTF_8;
-
 import com.google.common.annotations.VisibleForTesting;
 import com.google.devtools.build.lib.actions.ActionExecutionContext;
 import com.google.devtools.build.lib.actions.ActionKeyContext;
@@ -39,7 +37,8 @@
 /**
  * Action to write a file whose contents are known at analysis time.
  *
- * <p>The output is always UTF-8 encoded.
+ * <p>The output is not reencoded in any way. That is, the raw bytes from BUILD
+ * and .bzl files for string attributes will be emitted exactly as read.
  *
  * <p>TODO(bazel-team): Choose a better name to distinguish this class from {@link
  * BinaryFileWriteAction}.
@@ -171,7 +170,7 @@ private static final class CompressedString extends LazyString {
     final int uncompressedSize;
 
     CompressedString(String chars) {
-      byte[] dataToCompress = chars.getBytes(UTF_8);
+      byte[] dataToCompress = chars.getBytes();
       ByteArrayOutputStream byteStream = new ByteArrayOutputStream(dataToCompress.length);
       try (GZIPOutputStream zipStream = new GZIPOutputStream(byteStream)) {
         zipStream.write(dataToCompress);
@@ -202,7 +201,7 @@ public String toString() {
         // This should be impossible since we're reading from a byte array.
         throw new RuntimeException(e);
       }
-      return new String(uncompressedBytes, UTF_8);
+      return new String(uncompressedBytes);
     }
   }
 
@@ -235,7 +234,7 @@ public DeterministicWriter newDeterministicWriter(ActionExecutionContext ctx) {
     return new DeterministicWriter() {
       @Override
       public void writeOutputFile(OutputStream out) throws IOException {
-        out.write(getFileContents().getBytes(UTF_8));
+        out.write(getFileContents().getBytes());
       }
     };
   }