>> IOW you are replacing crashes with leaks.
>>
>> IMO it would be simpler to just never delete any objects...
>>
>> Now all these intentional leaks ("persistent objects")
>> will have to be refixed one by one once ever HBQT gets
>> fixed in the proper place. This means another (even
>> lengthier) chain of fixes in the future. Unless all of
>> these current hacks are clearly marked as such in source.
>>
>
> This is exactly what I am doing, locals to class variable.
> OR am I missing something from your message?
You're fixing low-level crashes by introducing
unnecessary memory consumption, and/or by solving
"reference counting" manually on the high level.
Both are wrong solutions. They are masking/hiding
the real problem by introducing unnecessary and/or
wrong logic in high-level code. Plus they create
impossible to document and follow high-level
programming requirements for any HBQT programs.
> If I leave them "as is", next time a new object will be
> created and will consume another chunk of memory.
> If an object is created once and reused, no harms,
> even if we fix the root issue. This also prevents memory
> fregmentation.
In this sense the best way to avoid memory fragmentation
is to never release any memory. In this sense your
reasoning is right.
All I'm seeing is huge and endless efforts to hide
very serious low-level errors on the high-level.
I've written about this dozens of times, Przemek
wrote about it, but the phenomena just continues.
Probably it would be more fruitful method to start
experimenting with and understanding the basic
memory/object handling concept of QT using some
_small_ and _representative_ examples, and create
a clean wrapper design for these. Then apply this
logic for all objects covered by HBQT.
[ I understand HBQT is so heavy by now, that such
experimenting is not very practical on it, hence
this suggestion. ]
Brgds,
Viktor
_______________________________________________
Harbour mailing list (attachment size limit: 40KB)
Harbour@harbour-project.org
http://lists.harbour-project.org/mailman/listinfo/harbour
