From: Robert Ramey (ramey_at_[hidden])
Date: 2007-11-23 12:54:16
John Maddock wrote:
> Robert Ramey wrote:
>> I've resolved to move these headers from their current locations
>> into the boost/serialization directory. Right now I'm sort of bogged
>> down in other stuff and I've make fairly extensive
>> changes/enhancements to the library. I'm working on a branch.
>> About a month ago I
>> inquired as to the convenience/inconvenience of merging my changes
>> from the
>> branch into the trunk and the response was that as things were soon
>> to be released, it would not be a good time to do so. This was fine
>> with me. So by now my branch is even more divirgent from the
>> trunk. I'll make these changes on this branch which can be merged
>> into trunk at such time as is deemed convenient.
> However, don't forget that you can always make changes to the trunk
> as well as the branch, and then merge them later :-)
I think that would increase the work I have to do.
> And... without wishing to point fingers... Serialization currently
> accounds for about half of the outstanding regressions at
> http://boost.org/regression/trunk/developer/issues.html, so I guess
> we need to make some sort of decision whether 1.35 will contain:
> 1) The Boost-1.34 version of serialization (in which case we should
> revert the Trunk for the time being).
> 2) Current Trunk version (plus fixes? Looks like many of the
> failures have
> a single cause?)
> 3) The current development branch version that you're actively
> working on.
When I look at the trunk results I see the following:
a) failure due to the fact that wide character serialization won't build
for certain platforms. On the branch, I tweaked to test/Jamfile
to make some tests depended on test results on in test/config.
So that tests which don't make sense aren't run. This works
on my system but I have no idea whether its kosher. In Jamfiles
for V1, this was addressed but that code was broken when
things moved to V2. I don't know if this combination was tested
with 1.34 so I don't know if reverting would help here.
b) tests of portable_binary_demo fail on high endien platforms.
This demo illustrates how to add functionality to an archive
via derivation. However, changes to binary_archive at
a basic level break this idea for binary archives. This this
demo is broken and in no longer serves its original purpose.
I haven't had time to replace it.
I believe that the above accounts for all the known failures.
c)I've noticed that bjam doesn't detect failure of test_exported
which is an imported test. I've got it passing on my branch
but I don't know what the true status is on the trunk.
d)Changes in implemenation of binary_archives prevent the
library from building on older platforms. This doesn't show
up in the boost test matrix.
> Which we go for much depends upon Beman's schedule, so over to the
> release manager I guess ;-)
> Regards, John.
> Unsubscribe & other changes: