From 99e86bbe29c03e3d52ac49e714c950b24e1f8d6f Mon Sep 17 00:00:00 2001
From: Kovid Goyal <kovid@kovidgoyal.net>
Date: Fri, 9 Oct 2020 09:29:07 +0530
Subject: [PATCH] Windows: Fix some MOBI files with non-BMP characters not
 being processed correctly. Fixes #1898894 [Conversion from epub to mobi en
 v5.2](https://bugs.launchpad.net/calibre/+bug/1898894)

lxml passes unicode to libxml2 as UCS-4 when non-BMP chars are present,
and libxml needs iconv to process UCS4 (god knows why) which was not
being built on windows.
---
 bypy/sources.json         | 6 +++++-
 src/calibre/test_build.py | 2 ++
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/bypy/sources.json b/bypy/sources.json
index 5fd5e7ae3e..f9cca7cf3f 100644
--- a/bypy/sources.json
+++ b/bypy/sources.json
@@ -246,11 +246,15 @@
 
     {
         "name": "iconv",
-        "os": "macos, linux",
         "unix": {
             "filename": "libiconv-1.16.tar.gz",
             "hash": "sha256:e6a1b1b589654277ee790cce3734f07876ac4ccfaecbee8afa0b649cf529cc04",
             "urls": ["https://ftp.gnu.org/pub/gnu/libiconv/{filename}"]
+        },
+        "windows": {
+            "filename": "libiconv-for-Windows-1.16.zip",
+            "hash": "sha256:d542e635cad954a62a2b451b5644e855b848398917e93adf46d8da4c9cc88b6d",
+            "urls": ["https://github.com/pffang/libiconv-for-Windows/archive/v1.16.zip"]
         }
     },
 
diff --git a/src/calibre/test_build.py b/src/calibre/test_build.py
index 8f1dcde6ff..05da265348 100644
--- a/src/calibre/test_build.py
+++ b/src/calibre/test_build.py
@@ -124,6 +124,8 @@ class BuildTest(unittest.TestCase):
         raw = b'<a/>'
         root = etree.fromstring(raw, parser=etree.XMLParser(recover=True, no_network=True, resolve_entities=False))
         self.assertEqual(etree.tostring(root), raw)
+        from lxml import html
+        html.fromstring("<p>\U0001f63a")
 
     def test_certgen(self):
         from calibre.utils.certgen import create_key_pair