From 4afd46286ba0c282c67593d060dc9d35377a9f36 Mon Sep 17 00:00:00 2001
From: Alfred Klomp <git@alfredklomp.com>
Date: Tue, 9 Jan 2024 21:57:38 +0100
Subject: [PATCH] bin/base64: decode: ignore newlines

By popular demand, ignore newlines in the encoded input. This achieves
bug compatibility with GNU base64.

Using `memchr(3)' is probably going to slow down things, but we can
vectorize the algorithm later.
---
 bin/base64.c | 62 ++++++++++++++++++++++++++++++++++++++++++----------
 1 file changed, 51 insertions(+), 11 deletions(-)

diff --git a/bin/base64.c b/bin/base64.c
index 6a94074..e7c0125 100644
--- a/bin/base64.c
+++ b/bin/base64.c
@@ -340,7 +340,7 @@ encode (const struct config *config, struct buffer *buf)
 	return true;
 }
 
-static int
+static bool
 decode (const struct config *config, struct buffer *buf)
 {
 	size_t nread, nout;
@@ -352,18 +352,58 @@ decode (const struct config *config, struct buffer *buf)
 	// Read encoded data into the buffer. Use the smallest buffer size to
 	// be on the safe side: the decoded output will fit the raw buffer.
 	while ((nread = fread(buf->enc, 1, BUFFER_RAW_SIZE, config->fp)) > 0) {
+		char *s      = buf->enc;
+		size_t avail = nread;
+
+		// By popular demand, this utility tries to be bug-compatible
+		// with GNU `base64'. That includes silently ignoring newlines
+		// in the input. Use `memchr(3)' to tokenize the input on
+		// newline characters. This is probably quite slow, and at some
+		// point we will want to vectorize this.
+		// Ignore \r for now, because there is no equivalent to
+		// `strpbrk(3)' in the mem* series of library functions. Also,
+		// this program is intended to be used in Unix-like
+		// environments where \r is not treated specially.
+		while (avail > 0) {
+
+			// Find the next newline character.
+			char *p = memchr(s, '\n', avail);
+
+			// Find the length of the next chunk. If no newline is
+			// found, use the rest of the buffer.
+			size_t len = p == NULL ? avail : (size_t) (p - s);
+
+			// Ignore empty chunks.
+			if (len == 0) {
+				s++;
+				avail--;
+				continue;
+			}
 
-		// Decode the input into the raw buffer.
-		if (base64_stream_decode(&state, buf->enc, nread,
-		                         buf->raw, &nout) == 0) {
-			fprintf(stderr, "%s: %s: decoding error\n",
-			        config->name, config->file);
-			return false;
-		}
+			// Decode the input into the raw buffer.
+			if (base64_stream_decode(&state, s, len,
+			                         buf->raw, &nout) == 0) {
+				fprintf(stderr, "%s: %s: decoding error\n",
+				        config->name, config->file);
+				return false;
+			}
 
-		// Append the raw data to the output stream.
-		if (write_stdout(config, buf->raw, nout) == false) {
-			return false;
+			// Append the raw data to the output stream.
+			if (write_stdout(config, buf->raw, nout) == false) {
+				return false;
+			}
+
+			// Bail out if the whole string has been consumed.
+			if (p == NULL) {
+				break;
+			}
+
+			// Add the newline to the chunk length.
+			len++;
+
+			// Move the start pointer and length past the chunk.
+			s     += len;
+			avail -= len;
 		}
 	}