Fix test_tar_utf8 to pass on macOS Catalina.
For whatever reason, the umlauts when encoded in NFC form were not recognized as valid UTF-8 by macOS Catalina when extracted by Bazel's tar archive extractor. When encoding them in a slightly different form (NFD), it all works.
I'm keeping the NFC form commented out so that we can revisit this in the future, but for now this test is blocking us from upgrading the CI machines to macOS Catalina.
RELNOTES: None.
PiperOrigin-RevId: 291168862
diff --git a/src/test/shell/bazel/archive_contents_test.sh b/src/test/shell/bazel/archive_contents_test.sh
index 1c5a757..bfb5df7 100755
--- a/src/test/shell/bazel/archive_contents_test.sh
+++ b/src/test/shell/bazel/archive_contents_test.sh
@@ -65,7 +65,10 @@
echo 'Hello World' > ext/data.txt
touch ext/$'unrelated\xF0\x9F\x8D\x83.txt'
touch ext/$'cyrillic\xD0\x90\xD0\x91\xD0\x92\xD0\x93\xD0\x94...'
- touch ext/$'umlauts\xC3\x84\xC3\x96\xC3\x9C\xC3\xA4\xC3\xB6\xC3\xBC'
+ touch ext/$'umlauts\x41\xCC\x88\x4F\xCC\x88\x55\xCC\x88\x61\xCC\x88\x6F\xCC\x88\x75\xCC\x88'
+ # TODO(philwo) - figure out why we get an "invalid characters" error on
+ # macOS Catalina when using this form (NFC normalized).
+ #touch ext/$'umlauts\xC3\x84\xC3\x96\xC3\x9C\xC3\xA4\xC3\xB6\xC3\xBC'
tar cf ext.tar ext
rm -rf ext
