Introspection does not detect most error domains

[v1993]

As of right now, only SourceError is properly declared as an error in generated files (see e.g. valadoc). This is a big issue for language bindings, since it removes the ability to (nicely) selectively catch exceptions from GUsb.

[v1993]

I'm not quite sure if it's the cause of this issue, but it's worth noting that corresponding quark functions are appropriated by similarly named classes, e.g. Device.error_quark.

[hughsie]

Is it because the GUsbDevice error uses G_DEFINE_QUARK? the g_usb_source_error_quark() seems to be written out manually.

[v1993]

That's plausible but would be pretty odd; G_DEFINE_QUARK is what official GLib docs recommend. It might have to do with header layout, although I'm not sure.

As an aside, I'm currently using a manually modified VAPI file on my end to work around this issue.

[hughsie]

As an aside, I'm currently using a manually modified VAPI file

Can you share what you're adding? I see

[CCode (cheader_filename = "gusb.h", cprefix = "G_USB_DEVICE_ERROR_", has_type_id = false)]
public enum DeviceError {
    INTERNAL,
    IO,
    TIMED_OUT,
    NOT_SUPPORTED,
    NO_DEVICE,
    NOT_OPEN,
    ALREADY_OPEN,
    CANCELLED,
    FAILED,
    PERMISSION_DENIED,
    BUSY,
    LAST
}

[v1993]

All I do is replace enum with errordomain for affected domains. This is enough to get things working, although static quark() methods also, ideally, should be relocated there from classes they got stuck to.

[hughsie]

@v1993 If you can work out the fix I'm happy to review and then tag a new release.

[v1993]

I'm currently bundling a modified vapi file with my application, which is good enough to get it working; hopefully will look into a proper fix later.