This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB 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: [patch] [python] Fix Python 3 build and testsuite issues


>>>>> "Phil" == Phil Muldoon <pmuldoon@redhat.com> writes:

Phil> For the hasattr point, I believe the equivalent check is the same in
Phil> this version, as the incompatible Python 3 frame filters code.  But I
Phil> use several methods of gdb.Symbol so which one would we use?  Remember
Phil> this is not the Symbol/Value like interface that frame_args, and
Phil> frame_locals returns, but the actual return of the symbol() function
Phil> from that interface.  At this point in the code we have already called
Phil> that symbol() API and have the object it returned.

We could pick any unique-to-Symbol method as a sentinel.
Say, the first one called in this method.

Phil> See SymValueWrapper class in FrameDecorator.py.  This object interface
Phil> specifies that it must return either a string, or a gdb.Symbol.  I
Phil> suppose you could return a Symbol like object, but you would have to
Phil> implement pretty much all of the gdb.Symbol methods.  This would not
Phil> make a great deal of sense.  Instead we should make gdb.Symbol
Phil> inheritable, so the user can sub-class.  This new child class of
Phil> gdb.Symbol would still work with the above code as the code  makes no
Phil> assumptions of the class other than it has the available methods.

I think subclassing Symbol is problematic.
For example I think it means it is hard to make a purely synthetic one.
What would the native state look like?  Where would it come from?

It seems better to go with Python's duck-typing approach instead.

Tom


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