Created attachment 166847 [details] [review]
Add new API g_typelib_require_private()

This is equivalent to g_typelib_require() but intended for use with
private typelibs, which get loaded from the provided directory.

Created attachment 166848 [details] [review]
Use the new g_irepository_require_private API

N.B.: the second patch is for gedit.

Review of attachment 166847 [details] [review]:

A few minor comments, otherwise looks good to commit.

::: girepository/girepository.c
@@ +1166,3 @@
 }
+check_typelib_internal_info (GTypelib     *typelib,
+static gboolean

"check_typelib_header" might be a better name.

@@ +1183,3 @@
+			     const gchar  *path,
+check_typelib_internal_info (GTypelib     *typelib,
+static gboolean

Should quote the filename formatter like: '%s', as we're doing for namespace.

@@ +1338,3 @@
+  g_free (tmp_version);
+ out:
+  ret = typelib;

Missing (allow-none)

@@ +1345,3 @@
+  g_free (tmp_version);
+ out:
+  ret = typelib;

"requires a valid version number"

@@ +1387,3 @@
+		   namespace, version, version_conflict);
+		   "Requiring namespace '%s' version '%s', but '%s' is already loaded",
+		   G_IREPOSITORY_ERROR_NAMESPACE_VERSION_CONFLICT,

Hm, there will be at least 3 areas of code which format %s-%s.typelib with namespace and version.  May be worth refactoring into format_typelib_filename (namespace, version)

Created attachment 167040 [details] [review]
Add new API g_typelib_require_private()

This is equivalent to g_typelib_require() but intended for use with
private typelibs, which get loaded from the provided directory.

I didn't create a format_typelib_filename() function because all
typelib filename generation are actually different:
 - %s.typelib
 - %s-%s.typelib
 - %s/%s-%s.typelib

Review of attachment 167040 [details] [review]:

::: girepository/girepository.c
@@ +1177,3 @@
+		      const gchar  *path,
+check_typelib_header (GTypelib     *typelib,
+static gboolean

Only fetch the version when you use it

@@ +1192,3 @@
+      g_set_error (error, G_IREPOSITORY_ERROR,
+		   G_IREPOSITORY_ERROR_NAMESPACE_MISMATCH,
+		   "Typelib file %s for namespace '%s' contains "

Missing quotes for file

@@ +1319,2 @@
+    goto out;
+  if (!check_typelib_header (typelib, path, namespace, version, error))

Can't see all the code above, but typelib might be leaked here

@@ +1334,3 @@
+  g_free (tmp_version);
+ out:
+  ret = typelib;

Add this to gi-sections.txt

@@ +1344,3 @@
+  g_free (tmp_version);
+ out:
+  ret = typelib;

It should probably make the version optional, eg accept NULL for consistency with the other APIs

@@ +1414,3 @@
+    typelib = g_typelib_new_from_mapped_file (mfile, &temp_error);
+    GError *temp_error = NULL;
+  {

typelib is leaked here.

Review tool failed, let's clarify:  New API should be added to gi-sections.txt, typelib is leaed after the check_typelib_header calls.

Created attachment 167071 [details] [review]
Add new API g_typelib_require_private()

This is equivalent to g_typelib_require() but intended for use with
private typelibs, which get loaded from the provided directory.

This approach reuses most of the code of g_irepository_require(), and
allows version=NULL (the previous patch didn't).

Created attachment 167072 [details] [review]
Do not leak typelibs with wrong header info.

Previously the typelibs that were loaded but whose header information
weren't right were just leaked.

Review of attachment 167072 [details] [review]:

Looks fine.

Review of attachment 167071 [details] [review]:

Looks fine, thanks!

Attachment 167071 [details] pushed as e59232e - Add new API g_typelib_require_private()
Attachment 167072 [details] pushed as 3f4a605 - Do not leak typelibs with wrong header info.

[Mass-moving gobject-introspection tickets to its own Bugzilla product - see bug 708029. Mass-filter your bugmail for this message: introspection20150207 ]