Changelog of @hackage/haskell-gi-base 0.21.1

0.21.1

• Remove the ::= and ::~ constructors in AttrOp, since they cannot really be used for anything, since they are pure functions.

0.21.0

• New release to keep major version parity with the haskell-gi package, no changes otherwise.

0.20.8

• Fix a bug which could lead to crashes when releasing boxed objects, see issue #130.

0.20.7

• Fix a memory leak in doConstructGObject.

0.20.6

• Use g_object_new_with_properties instead of g_object_newv in GLib versions 2.54 or later, to avoid a deprecation warning.

0.20.5

• Run object finalizers in the main loop. The reason is that for some types the destructor is not thread safe, and assumes that it is being run from the same thread as the thread that created the object, which can lead to crashes when using the threaded runtime.

0.20.4

• Better error diagnostics for wrapObject and newObject.

0.20.3

• Fixes for GHC 8.2.1 (and the corresponding base-4.10.0).

0.20.2

• Fix fromGVariant for empty arrays, see #91 for details.

0.20.1

• Add Data.GI.Base.CallStack, abstracting (and backporting to the extent possible) the HasCallStack constraint present in newer GHCs. Using this, we now include callstacks pervasively in the generated code.

• Improve the WrappedPtr implementation.

• Deprecate nulltoNothing, it is better to simply fix the overrides when necessary.

• Make the semantics of GObject ownership transfer closer to those used by the Python bindings.