Do not assume g_content_type_guess() always returns valid results

On Chrome OS the database of magic is either incomplete or missing and so
provide important fallbacks manually.

src/meson.build

@@ -45,6 +45,7 @@ libxmlb = library(
     'xb-builder-node.c',
     'xb-builder-source.c',
     'xb-builder-source-ctx.c',
+    'xb-common.c',
     'xb-machine.c',
     'xb-opcode.c',
     'xb-node.c',
@@ -115,6 +116,7 @@ if get_option('introspection')
       'xb-builder-source.h',
       'xb-builder-source-ctx.c',
       'xb-builder-source-ctx.h',
+      'xb-common.c',
       'xb-machine.c',
       'xb-machine.h',
       'xb-node.c',
@@ -198,6 +200,7 @@ if get_option('tests')
       'xb-builder-node.c',
       'xb-builder-source.c',
       'xb-builder-source-ctx.c',
+      'xb-common.c',
       'xb-machine.c',
       'xb-node.c',
       'xb-node-query.c',

src/xb-builder-source-ctx.c

@@ -11,6 +11,7 @@
 #include <gio/gio.h>
 
 #include "xb-builder-source-ctx-private.h"
+#include "xb-common-private.h"
 
 typedef struct {
 	GInputStream		*istream;
@@ -100,39 +101,22 @@ xb_builder_source_ctx_get_content_type (XbBuilderSourceCtx *self,
 					GError **error)
 {
 	XbBuilderSourceCtxPrivate *priv = GET_PRIVATE (self);
-	g_autofree gchar *content_type = NULL;
+	gsize bufsz = 0;
+	guchar buf[4096] = { 0x00 };
 
 	g_return_val_if_fail (XB_IS_BUILDER_SOURCE_CTX (self), NULL);
 
 	if (G_IS_SEEKABLE (priv->istream)) {
-		gsize bufsz = 0;
-		guchar buf[4096] = { 0x00 };
 		if (!g_input_stream_read_all (priv->istream, buf, sizeof(buf),
 					      &bufsz, cancellable, error))
 			return NULL;
 		if (!g_seekable_seek (G_SEEKABLE (priv->istream), 0, G_SEEK_SET,
 				      cancellable, error))
 			return NULL;
-		if (bufsz > 0)
-			content_type = g_content_type_guess (priv->filename, buf, bufsz, NULL);
 	}
-
-	/* either unseekable, or empty */
-	if (content_type == NULL)
-		content_type = g_content_type_guess (priv->filename, NULL, 0, NULL);
-
-#ifdef _WIN32
-	/* map Windows "mime-type" to a content type */
-	if (g_strcmp0 (content_type, ".gz") == 0)
-		return g_strdup ("application/gzip");
-	if (g_strcmp0 (content_type, ".txt") == 0 ||
-	    g_strcmp0 (content_type, ".xml") == 0)
-		return g_strdup ("application/xml");
-	if (g_strcmp0 (content_type, ".desktop") == 0)
-		return g_strdup ("application/x-desktop");
-#endif
-
-	return g_steal_pointer (&content_type);
+	if (bufsz > 0)
+		return xb_content_type_guess (priv->filename, buf, bufsz);
+	return xb_content_type_guess (priv->filename, NULL, 0);
 }
 
 /* private */

src/xb-common-private.h

@@ -0,0 +1,13 @@
+/*
+ * Copyright (C) 2020 Richard Hughes <richard@hughsie.com>
+ *
+ * SPDX-License-Identifier: LGPL-2.1+
+ */
+
+#pragma once
+
+#include <glib.h>
+
+gchar		*xb_content_type_guess			(const gchar	*filename,
+							 const guchar	*buf,
+							 gsize		 bufsz);

src/xb-common.c

@@ -0,0 +1,98 @@
+/*
+ * Copyright (C) 2020 Richard Hughes <richard@hughsie.com>
+ *
+ * SPDX-License-Identifier: LGPL-2.1+
+ */
+
+#define G_LOG_DOMAIN				"XbCommon"
+
+#include "config.h"
+
+#include <string.h>
+#include <gio/gio.h>
+
+#include "xb-common-private.h"
+
+static const gchar *
+xb_content_type_guess_from_fn (const gchar *filename)
+{
+	gchar *ext; /* no ownership */
+
+	g_return_val_if_fail (filename != NULL, NULL);
+
+	/* get file extension with dot */
+	ext = g_strrstr (filename, ".");
+	if (ext == NULL)
+		return NULL;
+
+	/* map Windows "mime-type" to a content type */
+	if (g_strcmp0 (ext, ".gz") == 0)
+		return "application/gzip";
+	if (g_strcmp0 (ext, ".txt") == 0 ||
+	    g_strcmp0 (ext, ".xml") == 0)
+		return "application/xml";
+	if (g_strcmp0 (ext, ".desktop") == 0)
+		return "application/x-desktop";
+	return NULL;
+}
+
+static gboolean
+xb_content_type_match (const guchar *buf, gsize bufsz, gsize offset,
+		       const gchar *magic, gsize magic_size)
+{
+	/* document too small */
+	if (offset + magic_size > bufsz)
+		return FALSE;
+	return memcmp (buf + offset, magic, magic_size) == 0;
+}
+
+/**
+ * xb_content_type_guess: (skip)
+ * @filename: (nullable): filename
+ * @buf: (nullable): file data buffer
+ * @bufsz: size of file data buffer
+ *
+ * Guesses the content type based on example data. Either @filename or @buf may
+ * be %NULL, in which case the guess will be based solely on the other argument.
+ *
+ * Returns: a string indicating a guessed content type
+ **/
+gchar *
+xb_content_type_guess (const gchar *filename, const guchar *buf, gsize bufsz)
+{
+	g_autofree gchar *content_type = NULL;
+
+	/* check for bad results, e.g. from Chrome OS */
+	content_type = g_content_type_guess (filename, buf, bufsz, NULL);
+	if (g_strcmp0 (content_type, "application/octet-stream") == 0 ||
+	    g_strcmp0 (content_type, "text/plain") == 0) {
+
+		/* magic */
+		if (bufsz > 0) {
+			if (xb_content_type_match (buf, bufsz, 0x0, "\x1f\x8b", 2))
+				return g_strdup ("application/gzip");
+			if (xb_content_type_match (buf, bufsz, 0x0, "<?xml", 5))
+				return g_strdup ("application/xml");
+			if (xb_content_type_match (buf, bufsz, 0x0, "[Desktop Entry]", 15))
+				return g_strdup ("application/x-desktop");
+		}
+
+		/* file extensions */
+		if (filename != NULL) {
+			const gchar *tmp = xb_content_type_guess_from_fn (filename);
+			if (tmp != NULL)
+				return g_strdup (tmp);
+		}
+	}
+
+#ifdef _WIN32
+	/* fall back harder as there is no mime data at all */
+	if (filename != NULL) {
+		const gchar *tmp = xb_content_type_guess_from_fn (filename);
+		if (tmp != NULL)
+			return g_strdup (tmp);
+	}
+#endif
+
+	return g_steal_pointer (&content_type);
+}