Add support for malformed LF HTTP responses

http-response-bodies.c

@@ -158,12 +158,21 @@ checkstate:
 			}
 			if (memcmp(bufp, "HTTP/1.1 ", 9) == 0 || memcmp(bufp, "HTTP/1.0 ", 9) == 0) {
 				// Got some headers; find transfer encoding, content length, and end of headers
-				eoh = memmem(bufp, n, "\r\n\r\n", 4);
+				m0 = memmem(bufp, n, "\r\n\r\n", 4);
+				m1 = memmem(bufp, n, "\n\n", 2);
+				if (m0 && m1) {
+					eoh = (m0 < m1 ? m0 + 4 : m1 + 2);
+				} else if (m0) {
+					eoh = m0 + 4;
+				} else if (m1) {
+					eoh = m1 + 2;
+				} else {
+					eoh = NULL;
+				}
 				if (!eoh) {
 					fprintf(stderr, "Error: end of headers not found\n");
 					return 1;
 				}
-				eoh += 4;
 				DEBUG_PRINTF("Response body begins at %p (offset %zu)\n", (void*)eoh, eoh - bufp);
 
 				if (memcmp(bufp, "HTTP/1.0 ", 9) == 0) {
@@ -174,14 +183,15 @@ checkstate:
 					}
 					length = response_length - (eoh - bufp);
 					state = STATE_BODY;
-				} else if ((m0 = memcasemem(bufp, eoh - bufp, "\r\ncontent-length:", 17)) && m0 < eoh) {
+				} else if ((m0 = memcasemem(bufp, eoh - bufp, "\ncontent-length:", 16)) && m0 < eoh) {
 					DEBUG_PRINTF("Found Content-Length header at %p (offset %zu)\n", (void*)(m0 + 2), m0 + 2 - bufp);
-					m1 = memmem(m0 + 1, eoh - (m0 + 1), "\r\n", 2);
+					m1 = memmem(m0 + 1, eoh - (m0 + 1), "\n", 1);
 					if (!m1) {
 						fprintf(stderr, "Error: CRLF after Content-Length missing\n");
 						return 1;
 					}
-					m0 += 17;
+					if (*(m1 - 1) == '\r') --m1;
+					m0 += 16;
 					while (m0 < bufp + n && (*m0 == ' ' || *m0 == '\t')) ++m0;
 					if (sscanf(m0, "%zu%ln", &length, &nscan) <= 0) {
 						fprintf(stderr, "Error: invalid Content-Length\n");
@@ -205,18 +215,19 @@ checkstate:
 
 					state = STATE_BODY;
 				} else {
-					m0 = memcasemem(bufp, eoh - bufp, "\r\ntransfer-encoding:", 20);
+					m0 = memcasemem(bufp, eoh - bufp, "\ntransfer-encoding:", 19);
 					if (!m0 || m0 >= eoh) {
 						fprintf(stderr, "Error: Content-Length and Transfer-Encoding missing\n");
 						return 1;
 					}
-					DEBUG_PRINTF("Found Transfer-Encoding header at %p (offset %zu)\n", (void*)(m0 + 2), m0 + 2 - bufp);
-					m1 = memmem(m0 + 1, eoh - (m0 + 1), "\r\n", 2);
-					if (!m1 || m1 >= eoh - 2) {
+					DEBUG_PRINTF("Found Transfer-Encoding header at %p (offset %zu)\n", (void*)(m0 + 1), m0 + 1 - bufp);
+					m1 = memmem(m0 + 1, eoh - (m0 + 1), "\n", 1);
+					if (!m1 || m1 >= eoh - 1) {
 						fprintf(stderr, "Error: CRLF after Transfer-Encoding missing\n");
 						return 1;
 					}
-					m0 += 20;
+					m0 += 19;
+					if (*(m1 - 1) == '\r') --m1;
 					while (m0 < bufp + n && (*m0 == ' ' || *m0 == '\t')) ++m0;
 					if (memcmp(m0, "chunked", 7) != 0) {
 						fprintf(stderr, "Error: unsupported Transfer-Encoding\n");
@@ -234,15 +245,15 @@ checkstate:
 				}
 
 				if (html_fake_base) {
-					m0 = memcasemem(bufp, eoh - bufp, "\r\ncontent-type:", 15);
+					m0 = memcasemem(bufp, eoh - bufp, "\ncontent-type:", 14);
 					if (m0 && m0 < eoh) {
-						DEBUG_PRINTF("Found Content-Type header at %p (offset %zu)\n", (void*)(m0 + 2), m0 + 2 - bufp);
-						m1 = memmem(m0 + 1, eoh - (m0 + 1), "\r\n", 2);
+						DEBUG_PRINTF("Found Content-Type header at %p (offset %zu)\n", (void*)(m0 + 1), m0 + 1 - bufp);
+						m1 = memmem(m0 + 1, eoh - (m0 + 1), "\n", 1);
 						if (!m1) {
 							fprintf(stderr, "Error: CRLF after Content-Type missing\n");
 							return 1;
 						}
-						m0 += 15;
+						m0 += 14;
 						while (m0 < bufp + n && (*m0 == ' ' || *m0 == '\t')) ++m0;
 						if (memcmp(m0, "text/html", 9) == 0) {
 							DEBUG_PRINTF("Is HTML response, inserting fake base tag\n");