[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [aspectj-dev] AJDE 1.1 status

I'd also like to release tomorrow, but I think delaying would be better.

It's also a question of when we do a 1.1.1 release. For users, if releasing now means we just have to cut a new one in two weeks, we're going to annoy the many people who've waited almost a year to start working with 1.1, who'll find they immediately have to consider upgrading to a bug-fix release. Conversely, if we wait a month or more for 1.1.1, then non-Eclipse developers would have a strong signal that we're not supporting them. That message would be stronger if we save people the initial agony of upgrading by telling those interested in IDE support to wait for 1.1.1. We haven't been on the radar in a while, but I suspect we will be with the 1.1.0 release, so noise might be amplified.

For us, except for Mik, releasing in two weeks should mean mostly that we sit tight until then. Conversely, if we release now, users have a reasonable expectation that some of the bugs deferred for 1.1 (and found since 1.1) will require fixing, so everyone will have to consider more work on AspectJ in the next 2-4 weeks for the 1.1.1 release. And if there are bugs in the IDE's, it makes for a lot less work and noise on the mail lists to fix them rather than to have to write them up and support users who find them. So delaying means less work overall.

I'm not sure there's any cost to us or users for delay. There's not much different between 1.1rc2 and the currently-proposed official release, which means that people wanting to be productive now aren't hindered. Nor would we be breaking any promises, AFAIK.

Mik should have had tests and fixes for these issues long ago, but I don't want to use releasing now as a way to enforce that, because it seems bad for users and for us.

In the interest of triage, this is the order I'd like to see Mik's findings addressed (ordered by seriousness, high-traffic...), at least by isolating test cases.

1) the VerifyError

2) Parsing of "../" paths in ".lst" files is broken.
   (works for me?)

3) There is a new suspicious bug causing *all* Eclipse compiler
   warnings to show when no "-Xlint" option is passed.
   (I haven't seen that)

4) Type information doesn't show up

5) Pointcut references don't show up

6) advice on constructions is missing

7) inter-type declarations info missing

My standards:

1) is necessary to identify if not resolve before any release.

2),3) I'd really like to see investigated before any release.
   I'm hoping this is not true for the compiler, ant, or ajbrowser.
   My concern is that options processing is broken, so
   people won't get what they expect.  If we know these are
   corner cases, we need not fix them.

4),5) I think would be complete enough as a 1.1.0 structure model;
    I'd really like 4) to be fixed.

6),7) I can see omitting for 1.1.0

So I'd vote +1, and 30 lashes for Mik for his delay*. I'm willing to provide a day or so of help to Mik, and to release tomorrow.


* plus thanks for finding these and bringing them up.

Jim.Hugunin@xxxxxxxx wrote:

Mik Kersten wrote:

I want us to figure out if it is worth delaying the 1.1 release by 2 weeks
in order to give me time to fix the remaining high priority issues listed
below. Help from others could shorten that time. If we don't delay I,
*sigh*, think that we should postpone the Eclipse, JBuilder, and NetBeans
plugin releases until 1.1.1 comes out.

It's too late in the game to delay the release two weeks for a large set of missing features/bugs that could have been detected months ago.  We should release AspectJ-1.1.0 either today or tomorrow (preferably today).  I'm happy to release any build that both George and Wes can agree has been properly built and tested.  Personally, I think the build that George produced and tested manually would be quite reasonable.

The current plugin support will be very useful to people even with the issues you list and it would be nice to see a release to go along with 1.1.0.  However, that decision rests in the hands of the different plugin development groups.


PS - If any committers disagree with this decision on the release, you are entitled to call for a vote on postponing the release in order to include Mik's changes.
aspectj-dev mailing list