couchdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Noah Slater <>
Subject Re: Problems blocking 1.2.0
Date Fri, 23 Mar 2012 13:27:18 GMT
Review after commit, by default, is fine. It should be review before commit on a hot release
branch, and the active release manager should be consulted in all instances. These changes
have wasted a significant amount of my time, and pushed this release back by half a week.

In fairness, this is not written down anywhere, and Randal was acting in good faith, so I
do not blame him for these issues. I will be documenting some of this stuff, for clarification
and future reference, after the release.

Let's leave it at that, and revisit best practices later.

On 23 Mar 2012, at 13:17, Robert Newson <> wrote:

> Finally, the review-after-commit approach bothers me, especially at
> crunch times like this. Can we get a chance to preview and discuss
> changes while we're trying to push out a release please?
> B.
> On 23 March 2012 13:15, Robert Newson <> wrote:
>> I'm also -1 on your revised solution. We go to the trouble of
>> carefully logging and formatting these errors and then log them at a
>> level that approximately no one ever runs at (debug is far too noisy
>> to use in production, for instance).
>> B.
>> On 23 March 2012 13:14, Robert Newson <> wrote:
>>> " Is there a good reason why we don't honor
>>> the create option in the way I expected?"
>>> Is there a good reason you committed a fix to a release branch without
>>> testing it?
>>> :)
>>> B.
>>> On 23 March 2012 13:12, Randall Leeds <> wrote:
>>>> On Fri, Mar 23, 2012 at 03:21, Robert Newson <> wrote:
>>>>> Release is still blocked as a clean startup now logs spurious errors
>>>>> (1.1.x and 1.2.x);
>>>> Filipe had asked me to include a log message there after my logging
>>>> related commits. I've just committed a change that reduces this to
>>>> debug level.
>>>> Unfortunately, there doesn't appear to be a way to request that
>>>> couch_file create the file if it doesn't exist, but just open it if it
>>>> does. Instead, we require the overwrite option in order not to bail
>>>> out. Otherwise, couch_replication_manager and couch_auth_cache could
>>>> just add the create option when opening these system databases rather
>>>> than detecting the error themselves and issuing couch_db:create after
>>>> the failed couch_db:open. Is there a good reason why we don't honor
>>>> the create option in the way I expected? The only reason I see is so
>>>> that we can return an error to clients that POST/PUT to create a new
>>>> database, but specifying the exclusive option to file:open actually
>>>> addresses this.
>>>> Anyway, the simple fix (to lower the log level) is committed to both branches.
>>>>> Apache CouchDB 1.1.2a3e2280b-git (LogLevel=info) is starting.
>>>>> [error] [<0.87.0>] Error opening file
>>>>> /Users/robertnewson/Source/couchdb/tmp/lib/_users.couch: no such file
>>>>> or directory
>>>>> [error] [<0.87.0>] Error opening file
>>>>> /Users/robertnewson/Source/couchdb/tmp/lib/_users.couch.compact: no
>>>>> such file or directory
>>>>> [error] [<0.100.0>] Error opening file
>>>>> /Users/robertnewson/Source/couchdb/tmp/lib/_replicator.couch: no such
>>>>> file or directory
>>>>> [error] [<0.100.0>] Error opening file
>>>>> /Users/robertnewson/Source/couchdb/tmp/lib/_replicator.couch.compact:
>>>>> no such file or directory
>>>>> Apache CouchDB has started. Time to relax.
>>>>> [info] [<0.31.0>] Apache CouchDB has started on
>>>>> B.
>>>>> On 21 March 2012 21:44, Randall Leeds <>
>>>>>> On Wed, Mar 21, 2012 at 13:47, Noah Slater <>
>>>>>>> This is great! Browser tests too, please! Thanks!
>>>>>>> On Wed, Mar 21, 2012 at 8:42 PM, Randall Leeds <>wrote:
>>>>>>>> On Wed, Mar 21, 2012 at 13:21, Noah Slater <>
>>>>>>>>> Have you added the appropriate entries in NEWS and CHANGES?
>>>>>>>> I have just now. Running make check so I can give you the
green light.
>>>>>>>> Would appreciate if anyone else would do the same.
>>>>>> Make check and futon tests pass here. Bob Dionne confirms the same.
>>>>>> found one tiny issue that remained with my changes and squashed it.
>>>>>> Green light from me.
>>>>>> I'm online for several hours if you need me.

View raw message