Been 7 months since the last checkup and a lot has changed so I thought
its about time for a new planning round.
We have passed feature freeze on 3.2 now. Feature-like additions are
needing to show more than just a need to be added now. They will also
need to show some good level of bug-free stability in use on 3.HEAD
before I'll be happy porting back.
* Comm streamlining
IPv4 / IPv6 handling - done
Accept API - done
Open API - one last bug patch (in 3397) to be tested and applied
Close API - done
Timeout API - done
Write API - done. writev() conversion deferred to later release.
Read API - needs SourceLayout shuffling but is operating okay AFAICT.
UDP handler APIs - deferred to later release
* StringNG upgrade merged - Kinkie, Alex ?
As far as I understand the current state:
MemBuf - done
StringArea - done
SBuf - stalled behind more urgent work
Code rollout - Going with incremental merges indefinitely rather than
a flag-day commit for 3.2
* SMP cache/store support (RockStore) - Alex
Down to bug fixes now?
* Squid3 performance regressions - Alex, Everybody (with help Dave Lang,
Jenny Lee, Frederico )
Kinkie polygraph testing to map out the impacts and regression points
that need work.
Any timeline on when we might see the graph Kinkie?
* Bugs - Everybody.
We are down to 61 release blockers. Including all the 2.x series bugs
which need to be confirmed whether or not they exist in 3.2.
5 bugs major or higher from 2.6 stable need to be checked for relevance
12 bugs major or higher from 2.7 stable need to be checked for relevance
5 bugs major or higher outstanding from 3.0 stable
23 bugs major or higher outstanding from 3.1 stable
10 new bugs major or higher open in 3.2 and 3.HEAD
plus 4 more major ones in the long term basket. # 7, 2833, 2956, 3390
Some themes seems to stand out if you have an interest in these areas:
breakage after reading disk files
delay pools behaviour
NTLM behaviour
HTTP/1.1 enhancements and compliance
* Windows support - Amos, Kinkie, Guido ?
MinGW compat fixes are stalled by the test machines being down.
Kinkie, Guido can you look into that please?
CommIO needs shuffling into the DiskIO library. This will resolve one
blocker build issue on Windows.
SMP IPC support on Windows needs to be designed, built and tested.
Amos
