Better release process [was: CVS Bind on MSVC failure]

From: Beman Dawes (bdawes_at_[hidden])
Date: 2001-11-30 16:41:09

At 03:17 PM 11/29/2001, Douglas Gregor wrote:

 >On Thursday 29 November 2001 12:35 pm, you wrote:
 >> In general, I think that we should freeze and run all regression tests 
at
 >> least three or four days (perhaps even a week) before release so we 
have
 >> time to react. (bind had a failure on SGI MIPSpro as well.)
 >
 >I agree. I would think that 3-4 days before release we should branch
 >CVS and
 >tag it appropriately (e.g., boost_1_26_0) and release only from branches. 

 >That way we can stabilize code on the branch without playing the waiting
 >game on newer (but not stable) changes.

We really need to do something!

Maybe we can try the above procedure for the next release.

--Beman