From b5e9ae3595ee27bc194138782fbc529d15681dea Mon Sep 17 00:00:00 2001
From: John MacFarlane <jgm@berkeley.edu>
Date: Tue, 18 Mar 2025 09:09:32 -0700
Subject: Fix parsing of base64 data URIs...

...to allow (a) URI escapes, (b) whitespace (which will be ignored).

Partially addresses #10704.
---
 src/Text/Pandoc/Class/PandocMonad.hs | 3 ++-
 src/Text/Pandoc/URI.hs               | 2 +-
 2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/src/Text/Pandoc/Class/PandocMonad.hs b/src/Text/Pandoc/Class/PandocMonad.hs
index b7f2d5a9b..e58429ee7 100644
--- a/src/Text/Pandoc/Class/PandocMonad.hs
+++ b/src/Text/Pandoc/Class/PandocMonad.hs
@@ -336,7 +336,8 @@ downloadOrRead :: PandocMonad m
                -> m (B.ByteString, Maybe MimeType)
 downloadOrRead s
  | "data:" `T.isPrefixOf` s,
-   Right (bs, mt) <- A.parseOnly pBase64DataURI s
+   Right (bs, mt) <- A.parseOnly pBase64DataURI
+                        (T.pack . unEscapeString . T.unpack $ s)
    = pure (bs, Just mt)
  | otherwise = do
   sourceURL <- getsCommonState stSourceURL
diff --git a/src/Text/Pandoc/URI.hs b/src/Text/Pandoc/URI.hs
index 09a9660d8..556654a28 100644
--- a/src/Text/Pandoc/URI.hs
+++ b/src/Text/Pandoc/URI.hs
@@ -133,7 +133,7 @@ pBase64DataURI = base64uri
       mps <- many mediaParam
       pure $ n1 <> "/" <> n2 <> mconcat mps
     A.string ";base64,"
-    b64 <- A.takeWhile (A.inClass "A-Za-z0-9+/")
+    b64 <- A.takeWhile (A.inClass "A-Za-z0-9+ \t\r\n/")
     A.skipWhile (== '=')
     -- this decode should be lazy:
     pure (decodeLenient (encodeUtf8 b64), mime)
-- 
cgit v1.2.3