Markus Henschel wrote:
-----Original Message-----
You didn't comment any part of my mail that proposed a fix for compatibility of archives prior to 1.42. Does that mean that you consider this issue solved or will you look into that later?
Actually I'm afraid to touch it. But I'm willing to take a look at it. Maybe you should add your patch and email (and maybe my response) as a comment to one of the trac items which touch upon this subject. Another benefit of this - and another reason that I left the items open is that other people how had variations on this problem will get notified when you append to the trac item and that might also be helpful.
Robert Ramey
I posted my analysis of the problem to this track item: https://svn.boost.org/trac/boost/ticket/4660#comment:51
I will provide a patch against the current svn trunk next week. May be it will be possible to reduce the problem to archives created by 1.44.0.
I took a short look at your analysis and it's clear you've put in a lot of work. When you attach the patch, I'll take a very good look at it and test it on my own machine. I very much appreciate your efforts here and will endeavor to make them count. As an aside, I should mention that these sorts of efforts have been essential to whatever success the serialization library has enjoyed. Without this sort of help, I could never, ever address all the issue of compiler dependencies, C++ ambiguities, incredibly wide breadth of applicability of the library. So I would like to thank all those who have made trac items with patches and analysis. I'm aware I haven't been able to incorporate all of them for one reason or another. But the majority have eventually found their way into the library to everyone's benefit. It's also sort of interesting to make note of the "maintainence" of boot libraries. The effort to maintain the serialization library is really not all that much - at least these days. I think a lot of that is that I try to address all the items posted even if that only means enhancing the documentation so that next time I can just point to the manual. Some items I've responded to by enhancing the code to detect/trap/warn of user errors. Basically if someone complains it's an indication that something somewhere needs to change. Upshot is that the more maintainence you're prepared to do, the less you actually have to do. I've noticed some authors have the tendency to respond to complaints defensively. It's understandable - but counter productive. Of course, after all this something just can't be fixed. Either they are side effects of fundemental design decisions. In interesting one in the serialization library is that a pointer to a pointer can't be serialized. Fixing this would require siginificant code re-organization that would likely produce more problems that would be worthwhile to address this corner case.
Have a nice weekend
aaaa - thanks, I'll try Robert Ramey