Subject: Re: [Boost-bugs] [Boost C++ Libraries] #4214: Use collection_size_type/version_type consistently (fixes Boost.MPI)
From: Boost C++ Libraries (noreply_at_[hidden])
Date: 2010-07-13 18:21:57
#4214: Use collection_size_type/version_type consistently (fixes Boost.MPI)
--------------------------------------+-------------------------------------
Reporter: dgregor | Owner: ramey
Type: Patches | Status: reopened
Milestone: Boost 1.44.0 | Component: serialization
Version: Boost Development Trunk | Severity: Problem
Resolution: | Keywords:
--------------------------------------+-------------------------------------
Comment (by ramey):
Doug,
I'm still struggling with this. Not so much with the MPI aspect but with
maintaining backward compatibility.
I had looked at your patch. I didn't (and still don't) understand the
"skeleton" concept but
I had concluded (maybe incorrectly) that the problem was that somewhere in
MPI there was a dependency upon the size of version_type. And of course
the existence of 3 different version types didn't help any. So it seemed
the correct fix was just to make the version_type consistent at one byte
wide.
I made changes to implement this, and still some changes had to be made in
MPI and all tests pass. BUT it's still not right for loading old
archives.
So my question is: suppose version_type and item_version_type are made
larger than 1 char. Archives already now need specialized code to handle
these particular types. But even though
I might save/load only one byte for current archives, I need to retrieve a
larger type older archives. So I would like to make the types larger than
one byte, even though the current version only saves/loads 1 byte.
Would this break MPI again?
Any insight you can give me would be appreciated.
Robert Ramey
-- Ticket URL: <https://svn.boost.org/trac/boost/ticket/4214#comment:7> Boost C++ Libraries <http://www.boost.org/> Boost provides free peer-reviewed portable C++ source libraries.
This archive was generated by hypermail 2.1.7 : 2017-02-16 18:50:03 UTC