HEADS UP: Berkeley DB 4...4.7 port removals/upgrades may require manual preparation

[Scot Hetzel]

On Mon, Aug 25, 2014 at 5:15 AM, Matthias Andree <matthias.andree at gmx.de> wrote:
> On 25. August 2014 07:57:36 MESZ, Scot Hetzel <swhetzel at gmail.com> wrote:
>>
>> On Sun, Aug 24, 2014 at 12:16 PM, Michael Gmelin <grembo at freebsd.org>
>> wrote:
>>>
>>>
>>>
>>>  On Sun, 24 Aug 2014 11:59:37 -0500
>>>  Scot Hetzel <swhetzel at gmail.com> wrote:
>>>
>>>>  On Sun, Aug 24, 2014 at 7:20 AM, Michael Gmelin <grembo at freebsd.org>
>>>>  wrote:
>>>>>
>>>>>  @Olli: Checking Mk/bsd.databasemk it seems like you're right. It's
>>>>>  quite amusing how this went unnoticed for so long, it has been in
>>>>>  there for eight years :)
>>>>
>>>>
>>>>
>>>>  As the original author of the USE_BDB code, the USE_BDB=5  had worked
>>>>  in the past.  As it is supposed to mean that this port only uses that
>>>>  version of BDB.  The code was changed on Aug 21 (r365599) and the
>>>>  author of the new code might have forgot to test this case.
>>>
>>>
>>>
>>>  Not sure if you read the rest of the thread, but this is still working
>>>  as intended (I tested using both the old and new version).
>>>
>>>  The issue is, that USE_BDB=version means USE_BDB=yes,
>>>  WITH_BDB_VER=version, but Kurt has WITH_BDB_VER=6 in make.conf,
>>>  which overrides this version number and version =6 is invalid for
>>>  devel/ice. So basically it works as designed, Kurt wanted a specific
>>>  version of bdb, which doesn't work for devel/ice.
>>>
>>>  The fix was to add WANT_BDB_VER= 5 in devel/ice, which I guess is ok,
>>>  since this is the only version it really works with (and I guess I
>>>  could remove the INVALID directive now, since WANT means really
>>>  *want*).
>>
>>
>>
>> Hmm, I thought I had it implemented this way at one time.  Any way.
>> What we should be doing with the WITH_BDB_VER is not overriding the
>> USE_BDB value.  Instead we should see if it is a VALID version to use
>> for this port, if it is not, then use the USE_BDB value.
>>
>> So basically the code should be doing the following:
>>
>> USE_BDB=yes
>>  - use the default version (48+) or the installed version if higher
>>
>> USE_BDB=yes, WITH_BDB_VER=6
>>  - use version 6
>>
>> USE_BDB=5
>>  - should mean the same as USE_BDB=yes, WANT_BDB_VER=5
>>  - shouldn't be able to override by setting WITH_BDB_VER=6
>>  - this would also allow the removal of the WANT_BDB_VER variable.
>>  - no error should be generated when WITH_BDB_VER is set to an invalid
>> version
>>
>> USE_BDB=48 6
>>  - use (either 48 or 6) or the installed version that matches one of
>> these versions
>>  - no error should be generated when WITH_BDB_VER is set to an invalid
>> version
>>
>> USE_BDB=5+, WITH_BDB_VER=48
>>  - use version 5+ or the installed version if higher
>>  - shouldn't allow downgrade to a lower unsupported version by setting
>> the WITH_BDB_VER
>>  -- should we still display an error in this case or just install the
>> port with bdb 5+?
>>
>> If we make the above changes to the code, then INVALID_BDB_VER and
>> WANT_BDB_VER could disappear.
>>
>>>  My question is, what the point of INVALID_BDB_VER really is in this
>>>  case, it seems a bit pointless to me given the trouble it caused Kurt
>>>  and how we resolved this. Having a fully specified list of supported
>>>  versions in WANT_BDB_VER seems better in this case (assuming WANT
>>>  supports listing multiple versions).
>>
>>
>>
>> At the time I had implemented the code, INVALID_BDB_VER was meant to
>> poke holes in the VER+:
>>
>> USE_BDB=2+
>> INVALID_BDB_VER=3
>>
>> So the port would be able to use version 2 and 40+, and not 3.
>> Probably should have just implanted it as:
>>
>> USE_BDB=2 40+
>>
>>  and skipped the INVALID_BDB_VER entirely.
>>
>> I was also thinking that we should make WITH_BDB_HIGHEST the default.
>> It is used by VER+ when multiple versions of bdb are installed.   That
>> way we can remove this variable also.
>
>
> Scot,
>
> I rewrote all of the Berkeley DB detection code to match documentation, and
> I will check how we can optimise the new code for corner cases.
>
> Sorry about the inconvenience.
>

Thanks for the code re-write.

Even I missed the corner case of checking that the WITH_BDB_VER was a
valid version for the port. If you can fix that, then the last 3 cases
may just work.

WITH_BDB_VER (global) - shouldn't generate an error if it doesn't
match a valid version for the port - use the port specified version
${UNIQUENAME:tu:S,-,_,}_WITH_BDB_VER (port specific) - should generate
an error if it doesn't match a valid version for the port

-- 
DISCLAIMER:

No electrons were maimed while sending this message. Only slightly bruised.