Quoting Basile STARYNKEVITCH <bas...@starynkevitch.net>:
My current way of thinking is that MELT is incompatibly with
precompiled headers. And I tend to think that any sufficient flexible
plugin machinery is incompatible with pch. Intuitively I would believe
that pch (in their current form) and plugins are incompatible.
I suppose it should still be OK if no plugin becomes active before the pch
is saved or used.
If your pointers are such that you can ignore them for garbage collection,
but they cannot be saved sanely in a precompiled header, you could add a
GTY marker that says just that.
An attempt to save a pch with such a pointer should then fail.
That should probably be an internal error; if we assume that plugin use
makes pch unusable, we should set a flag when a plugin becomes active, and
refuse to save or load a pch if that flag is set.
If you think it could be OK for some plugins if they are used in exactly the
same way for pch generation and use, you can have each plugin generate data
to be compared for pch, or a failure indication if the plugin is fundamentally
incompatible with pch (or if the plugin hasn't implemented the
necessary logic).
