GitLab

The directory was removed before and reappeared trough some less
successful merge. The changelog already contains the removal.

Remove some duplicated DESTDIR.
This was reported by Jan Engelhardt via bind10-dev list.

(NOTE: I had the wrong date in the release tarball again;
I changed my checklist again to make sure this is fixed
for next time.)

This is 20110224.

Add release stamp in the ChangeLog

Update versions for bind10, bindctl, and msgq.
Maybe not precise and we should discuss if or how we should
maintain per component version numbers.

In the test for UDPSocket, use io_service::run_one() instead of
poll().

The setproctitle check in configure is not really used by build;
it only provides a suggestion for installation.

Some systems don't need it.

We could consider removing this check (ticket #587).

Discussed via jabber.

It is not used currently.
Discussed in jabber.

my bad, i think the review even pointed it out but it still snuck through

If the e is used explicitly, the type is lost (even when it shouldn't
probably be, as we catch it as a reference). This way the type is
preserved, allowing other catches on the stack handle it properly.

It needed to close sockets to be able to successfully roll back in case
of failure.

so sqlite3 is not the only one.

TODO: still need to document the in-memory datasource.

it used IPPROTO_IDP instead of IPPROTO_UDP, somehow this snuck past my review, and since IPPROTO_IDP exists, it reverted to TCP (which isn't implemented yet). The future-proof fix is to either not allow this to be passed (use a boolen use_udp or something), or raise an error on an unknown protocol.

But for now this quick fix should at least get it working again. Reviewed on jabber by stephen.

should be sufficiently trivial, directly committing to master.

The closures are moved to the top of the function and commented. It'll
hopefully be more readable this way.