defint.lisp and limit.lisp ver. 1-12-2012: gcl bug?

I consider this apparent bug as fixed, based on Dan Gildea's
latest files on git and his test of gcl using the current
git. Those (like me, at present) who do not maintain a
clone of the current git and hence do not check the compiled
behavior, should see correct behavior with ver. 5.26.1.

Another issue, however, is the requirement of setting
domain:complex to get correct results from integrate
in some cases (such as here). How is the novice etc user
to know when and when not to set domain=complex?

We always want a user to get a correct answer to a symbolic
integral (for its float value) if possible. Is there any reason
why domain=complex should not be the default?

Or, perhaps integrate (or defint) should set domain:false
locally?

Ted Woollett