Created attachment 15154 [details] [review]
API docs patch

Is this based on the C documentation?

I don't like the mention of reference counting - C++ people shouldn't
care about manual reference counting.

The C++ object is called Glib::Module, or Module, not GModule.

> If nothing can be found, at the moment a valid GModule is
> +   * returned but gobj() will be NULL. This will be fixed in 2.4.
Nobody would expect an invalid C++ instance. That has no meaning. Just
say "check that gobj() != NULL to see whether this succeeded. glibmm
2.4 will have an operator=() for this."

The documentation for the destructor seems to be superfluous. Of
course destructors release memory.,

std::strings can not be NULL. If the C function can take this then we
need to add a method overload without that parameter.

Created attachment 15181 [details] [review]
Updated patch

So, I think this is based on the C docs?

This should be applied to both branches with the following changes:

s/operation=()/operator bool()/ - sorry, my typo.

Near "It can be an empty string" there should be a "//TODO: " saying
that we should add an override. Actually I suspect that an empty
string != NULL so the documentation might be wrong. If so, we could
fix our implementation to use null if the string is empty. A patch or
TODO for that might be appropriate.

Thanks, Ross.

Created attachment 15197 [details] [review]
Updated patch

Thanks, Ross, I didn't expect you to provide a revised patch again,
but it's very helpful.

Could someone please apply this, on both branches. We should probably do
s/operation bool/operator bool/ 
though before commiting though.

Ross, if you have cvs write access, then that means you.

I'm busy too! :)

Committed (with typo fix) to gtkmm2 and glibmm.