[squeak-dev] Update number -> commit mapping?

Bert Freudenberg bert at freudenbergs.de
Mon Jan 7 14:50:15 UTC 2013




On 07.01.2013, at 15:40, Frank Shearar <frank.shearar at gmail.com> wrote:

> On 7 January 2013 13:46, Bert Freudenberg <bert at freudenbergs.de> wrote:
>> On 03.01.2013, at 19:49, Chris Muller <asqueaker at gmail.com> wrote:
>> 
>>> No, I don't think we currently capture enough state to determine that.
>>> 
>>> On Thu, Jan 3, 2013 at 10:08 AM, Frank Shearar <frank.shearar at gmail.com> wrote:
>>>> Is there a convenient way of mapping an update number to the commit
>>>> that produced that update number?
>>>> 
>>>> I know that SqueakTrunk build #105 updated to update number 12360, but
>>>> I don't know of a way of knowing "ah, that's Environments-fbs.11" (if
>>>> indeed it was).
>>>> 
>>>> frank
>> 
>> It's not even a unique mapping anyway. What you could do is compare the package versions between one build and the next to see which packages changed.
> 
> I'd argue that it _must_ be a unique mapping, from update number to a
> set of package versions. An update might point to multiple packages -
> 2121 -> {Foo-fbs.1 Bar-fbs.2}.
> 
> Further, if any update N refers to a package version Foo-blah.M, then
> any update N+1, ... that refers to Foo should refer to a Foo-blah.K
> where K > M.
> 
> But I take your point that - at least for now - a pragmatic approach
> to _determining_ the mapping (of the current update at least) is as
> you describe: take the set of mappings now, update one step, take the
> set of mappings again, and diff the two.
> 
> frank

For a single image, the mapping is indeed unique. But it is not necessarily the same for two given images. Close, usually, but not guaranteed, because it depends on the time you update. That's because packages are not loaded in the order they are committed, but in the order defined in the update config map. 

- Bert -


More information about the Squeak-dev mailing list