This is the mail archive of the mailing list for the Archer project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [python] don't try to free "independent" types

Thiago> Is it possible that one struct type can be present in two hierarchies?

It turns out that we have a problem if we ever let the user
dynamically construct a type from its constituents.

If we did this then it would be trivial to construct a type that
straddles hierarchies -- for instance you could make a struct with
fields whose types are from different ones.

I think we will eventually want to allow this.  For example, libgcj
will create classes and such dynamically -- it would be reasonable to
have Python code read the metadata out of the runtime and create a
gdb-side type representation.

However, I don't think we need to worry about this quite yet.  I don't
think gdb has this capability right now; AFAICT it would at least mean
adding ABI knowledge to gdb.  So, it is a ways off.

I think a specialized GC is not extremely difficult to write later on.
The set of roots is relatively well-defined.  We could GC after every
N calls to value_free, or something along those lines.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]