comp.lang.ada
 help / color / mirror / Atom feed
From: madmats%elcit.epfl.ch@CMUCCVMA.cc.cmu.edu (Mats Weber)
Subject: Portability of Unchecked Deallocation in Ada
Date: 16 Nov 89 12:41:44 GMT	[thread overview]
Message-ID: <891116134144.2080179c@SIC.Epfl.CH> (raw)


Unchecked_Deallocation is not required to do anymore than setting its parameter

to null. Effective deallocation of the memory is not required. However, most
systems implement it correctly.

I have submitted an Ada-9X revision request that requires Unchecked_Deallocation

to work (i.e. to deallocate storage) if automatic garbage collection is not
provided. This, however, is quite hard to specify in the way the semantics of
the language are given in the LRM.

There are other related deficiencies in the LRM. For example, it is not stated
that constrained objects must occupy as little space as possible. Even worse,
Storage_Error can be raised even at places where it is not 'sensible' to do so.

In these areas, it seems that all kinds of abstract specifications of the
semantics will fail.

Mats Weber
Swiss Federal Institute of Technology
EPFL DI LITh
1015 Lausanne
Switzerland

e-mail : madmats@elcit.epfl.ch
phone  : +41 21 693 42 43

             reply	other threads:[~1989-11-16 12:41 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
1989-11-16 12:41 Mats Weber [this message]
  -- strict thread matches above, loose matches on Subject: below --
1989-11-15 19:42 Portability of Unchecked Deallocation in Ada planck!hercules!westley
replies disabled

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox