I am creating a Differential revision by manually uploading a diff (which I have created by running hg diff --git …). Everything is encoded in UTF-8, and the diff file starts with the UTF-8 Byte Order Mark (BOM), i.e., the bytes 0x EF BB BF. This results in the first hunk of the diff just being silently ignored by Differential.

I assume that with a BOM, Differential fails to parse the first line of the file, which is the header for the first hunk, and thus just ignores that first hunk.

Example

Take this very simple diff that just adds two files:

diff --git a/path/to/file1.cs b/path/to/file1.cs
new file mode 100644
--- /dev/null
+++ b/path/to/file1.cs
@@ -0,0 +1,1 @@
+foo1
diff --git a/path/to/file2.cs b/path/to/file2.cs
new file mode 100644
--- /dev/null
+++ b/path/to/file2.cs
@@ -0,0 +1,1 @@
+foo2

Note that even though I copy-pasted this from a file with a BOM, the paste here of course does not carry a BOM. Here is a file that contains exactly this diff, with an UTF-8 BOM:

In a hex editor, this file should start like this (just to make sure that Phorge does not re-encode the uploaded file or strip the BOM):

00000000: efbb bf64 6966 6620 2d2d 6769 7420 612f  ...diff --git a/

If I upload this file to Differential, the resulting Differential diff shows file2.cs being added, but there is no trace of file1.cs.