I am basically fine with it, but
suggest with check explicitly with Scott Lewis (ECF) and the
cross-project list proactively rather than just spring it on them.
Even if their "source code still compiles" there is a chance
behavior might change, as always, so it implies some amount of
additional testing as well as re-builds and re-packaging.
Are you concerned about conflicts because it uses singletons?
In addition to conflicts within Sim Release there would be
conflicts from unknown adopters.
Also, it would be best to be sure the bundles "import what they
export" (since the package may come from different class loader
"sources"). And, I suggest, we just tell everyone they *must* use
"import package" for this case. (I am not sure we say that with
ICU, but is a similar case where "import package" is the only
option that makes sense.) Have we found a case where "require
bundle" is used?
Those are just my quick thoughts -- I have only been skim reading
the longer on-going discussion, so perhaps all this has been
discussed already.
Thanks for reading,
On 02/09/2017 03:04 AM, Steffen Pingel wrote:
Hi,
I would like to go ahead with a change that updates HttpClient
Core to version 4.4.6 [1].
As part of this change I would like to remove 4.4.4 from Orbit
to avoid any potential conflicts. Please let me know if anyone
needs version 4.4.4 specifically and we can consider leaving
that in the repository as well.
The main driver are the bug fixes in the latest version of
HttpClient Core: http://www.apache.org/dist/httpcomponents/httpcore/RELEASE_NOTES-4.4.x.txt
Steffen
_______________________________________________
orbit-dev mailing list
orbit-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/orbit-dev
|