[squeak-dev] How to find...?

  Hello,

  I've been looking at the relicensing agreement issue, and trying to
make better mapping from the initial to the person.  It is almost
straightforward if it is not in the preparation of 3.9 time; I have
all changes in the update stream so I can grep and look at the
preamble and find it out.  (I built a little browser so that you can
visit all old versions, more or less similar to Maurice has, I
believe).

  However, in the 3.9 era the changeset typically looks like:

|repository|
repository := MCHttpRepository
                location: 'http://source.squeakfoundation.org/39a'
                user: ''
                password: ''.
(repository loadVersionFromFileNamed: 'ScriptLoader-sd.183.mcz') load.

and cannot get the great detail.  For example, if I would like to find
out who 'tp' is in the following method:

!DependentsArray methodsFor: 'copying' stamp: 'tp 7/6/2006 11:11'!
size
        ^self inject: 0 into: [ :count :dep | dep ifNil: [ count ] ifNotNil: [ count + 1 ]]! !

What is the best way?

-- Yoshiki

On Wed, 2008-08-20 at 15:42 -0700, Yoshiki Ohshima wrote: I'm not sure what sort of detail you are looking for.  If you mean the
sort of textual information you might find in a changeset preamble you
might be able to find that information in the MC repository, hopefully
including a reference to a Mantis report with further information.  Go
to http://source.squeakfoundation.org/39a.html and click on Versions and
search for the specific version you want to examine, then look at the
details, you can even browse the code as of that version if you want.

Alternatively if you want a 3.9 or later image with full history it
seems like it should work to start with a 3.8 image and force it to load
updates.  This will be a long and arduous task but in time should result
in an image with method versions reflecting the changes during 3.9 (and
3.10 if you continue that far).

Ken

  Thank you, Ken,

> I'm not sure what sort of detail you are looking for.  If you mean the
> sort of textual information you might find in a changeset preamble you
> might be able to find that information in the MC repository, hopefully
> including a reference to a Mantis report with further information.  Go
> to http://source.squeakfoundation.org/39a.html and click on Versions and
> search for the specific version you want to examine, then look at the
> details, you can even browse the code as of that version if you
> want.

  The detail I'm looking for is to make the mapping from the author
initials to the authors full name.  I don't mind to look at the
textual information in the preamble, but it is hard to find which one
contains the method and the version in question.


  I managed to fill most of the '?'  marks on:

http://www.netjam.org/squeak/contributors/missingSignatories

but not the 3.9 era contributors.  How would you find who 'tp' was,
for example?

  (How about:

'ASF', 'edt', 'EP', 'EW', 'm', 'NDCC', 'sk', 'YE'?

-----
!MIMEDocument class methodsFor: 'initialize-release' stamp: 'ASF 4/30/2005 17:14'!
extendedMIMEdatabase
-----

-----
!ObjectExplorerWrapper methodsFor: 'converting' stamp: 'edt 5/26/2003 12:36'!
asString
-----

-----
!BagTest methodsFor: 'tests' stamp: 'EP 2/28/2006 09:48'!
testSortedElements
-----

-----
!HttpUrl methodsFor: 'downloading' stamp: 'EW 4/30/2006 12:00'!
retrieveContentsArgs: args
-----

-----
!ProcessTerminateBug methodsFor: 'tests' stamp: 'm 7/28/2003 19:10'!
testSchedulerTermination
-----

-----
!DictionaryTest methodsFor: 'association tests' stamp: 'NDCC 3/8/2006 08:14'!
testAddAssociation
-----

-----
!WorldState methodsFor: 'stepping' stamp: 'sk 1/26/2003 18:31'!
runStepMethodsIn: aWorld
-----

-----
!SkipListTest methodsFor: 'public' stamp: 'YE 3/8/2006 10:00'!
testAtPutAlreadyPresent
-----


> Alternatively if you want a 3.9 or later image with full history it
> seems like it should work to start with a 3.8 image and force it to load
> updates.  This will be a long and arduous task but in time should result
> in an image with method versions reflecting the changes during 3.9 (and
> 3.10 if you continue that far).

  I'd probably build the extention of tool that I have now; instead of
filing in, it just creates the ChangeRecord like object and store in
my dictionary.  I don't necessarily like to see the current code
changing^^; In this way, I can view the long history of methods from
1.13 to the latest in a browser-like tool.

-- Yoshiki

On Wed, 2008-08-20 at 16:26 -0700, Yoshiki Ohshima wrote: I can sort of help regarding 'YE'.  I ran into those initials when
planning to export SkipList and license it as MIT.  I emailed all of the
other authors of SkipList I could find and in the process asked if
anyone recognized those initials.  Stephane Ducasse replied (in the
context of SkipList): "It was one of my student and I forgot to remove
his initial when I  rewrote all his tests."  I'm not sure I really
understand the relevance of this response but I have not as yet
revisited it since I have not heard from a much more relevant
contributor, Scott A. Crosby (SAC).  I don't suppose he has responded
with a signature since I last checked?

Ken

At Wed, 20 Aug 2008 18:55:56 -0500,
Ken Causey wrote:
  The question is still about "how to find" such a change.

  Hmm, I wonder why the version in the OLPC image is older and all
with Leandro's initials...

-- Yoshiki

On Wed, 2008-08-20 at 20:51 -0700, Yoshiki Ohshima wrote: I'm sorry I'm still confused.  What do you mean by 'such a change'?
Uhm, I may be confused again, or maybe I confused you.  I did not mean
to imply that Scott is the only contributor to SkipList, Leandro
Caniglia is certainly another major contributor and Stephane and Tom
Koenig have also contributed.  Both of them have signed the agreement
though and so there are no concerns regarding the licensing, at least.

Ken

> -- Yoshiki
>

At Thu, 21 Aug 2008 08:45:46 -0500,
Ken Causey wrote:
  I meant that "a changeset" that has the preamble where I "might be
able to find that information in the MC repository".  how do I find in
the MC repository which version contains the changeset with a method
with, say, 'ASF'?

> >   Hmm, I wonder why the version in the OLPC image is older and all
> > with Leandro's initials...
>
> Uhm, I may be confused again, or maybe I confused you.  I did not mean
> to imply that Scott is the only contributor to SkipList, Leandro
> Caniglia is certainly another major contributor and Stephane and Tom
> Koenig have also contributed.  Both of them have signed the agreement
> though and so there are no concerns regarding the licensing, at least.

  Sure.  I understand that you didn't imply that Scott is the only
contributor.  I had a wrong sense that we hevested Scott's change by
3.8.

-- Yoshiki

Am 21.08.2008 um 17:45 schrieb Yoshiki Ohshima:

>  I meant that "a changeset" that has the preamble where I "might be
> able to find that information in the MC repository".  how do I find in
> the MC repository which version contains the changeset with a method
> with, say, 'ASF'?


MC is snapshot-based, every MCZ contains all the code. So you would  
have to bisect all MCZs to find the one introducing a change.

An MCZ is a zip file that also contains the plain Smalltalk source  
code, so that could be used.

A more efficient way is to use MCDs which store something like a  
changeset (actually, the old and new snapshots of the changed parts)  
that updates from one MCZ to another.

An up-to-date SqueakSource server can auto-generate those MCDs even  
though they do not appear in the file index list. E.g., this repository

        http://source.squeakfoundation.org/39a/

shows

        http://source.squeakfoundation.org/39a/Collections-md.71.mcz
        http://source.squeakfoundation.org/39a/Collections-md.70.mcz
        http://source.squeakfoundation.org/39a/Collections-md.69.mcz

but you can access

        http://source.squeakfoundation.org/39a/Collections-md.70(md.69).mcd

which is much smaller (though the first request might take long  
because the mcd is built). It has the new methods in new/source.st so  
that's where you would find the changes ...

The full history info including preambles is stored in both MCZs and  
MCDs inside a file named "version" as literal Smalltalk array (very  
similar to s-expr).

- Bert -

On Thu, 2008-08-21 at 08:45 -0700, Yoshiki Ohshima wrote:
>   I meant that "a changeset" that has the preamble where I "might be
> able to find that information in the MC repository".  how do I find in
> the MC repository which version contains the changeset with a method
> with, say, 'ASF'?

Well, I started to write up my ideas on how to go about this.  But I run
into a roadblock for any post-3.8 and I'm struggling to find a workable
solution.  I'll keep thinking about it and get back to you if I come up
with anything.

One solution (tedious) might be to start with the date in the timestamp
and check first the squeak-dev archive for that date, then
bugs.squeak.org.  This is going to take a lot of manual checking though.

Ken

On Thu, 2008-08-21 at 12:20 -0500, Ken Causey wrote: OK, I hit some paydirt with this method:

http://lists.squeakfoundation.org/pipermail/squeak-dev/2005-April/091252.html

I'll try it with some of the other initials you mentioned.

Ken

Am 21.08.2008 um 19:30 schrieb Ken Causey <[hidden email]>:

Yay!

- Bert -

On Thu, 2008-08-21 at 08:45 -0700, Yoshiki Ohshima wrote:
> > > > >   (How about:
> > > > >
> > > > > 'ASF', 'edt', 'EP', 'EW', 'm', 'NDCC', 'sk', 'YE'?

For 'm' it looks like the credit should go to Stephen Pair, the 'm' may
be a mistake by Marcus or someone else who actually created a changeset.

http://lists.squeakfoundation.org/pipermail/squeak-dev/2003-July/063646.html

Ken

On Thu, 2008-08-21 at 08:45 -0700, Yoshiki Ohshima wrote:
> > > > >   (How about:
> > > > >
> > > > > 'ASF', 'edt', 'EP', 'EW', 'm', 'NDCC', 'sk', 'YE'?

For 'sk':

http://lists.squeakfoundation.org/pipermail/squeak-dev/2003-January/051738.html

Ken

On Thu, 2008-08-21 at 13:37 -0500, Ken Causey wrote:
> On Thu, 2008-08-21 at 08:45 -0700, Yoshiki Ohshima wrote:
> > > > > >   (How about:
> > > > > >
> > > > > > 'ASF', 'edt', 'EP', 'EW', 'm', 'NDCC', 'sk', 'YE'?

'EP', 'NDCC', and 'YE' all are related collection tests around March of
2006.  Could these all be students of Stephane or Marcus?

Ken

On Thu, 2008-08-21 at 13:37 -0500, Ken Causey wrote:
> On Thu, 2008-08-21 at 08:45 -0700, Yoshiki Ohshima wrote:
> > > > > >   (How about:
> > > > > >
> > > > > > 'ASF', 'edt', 'EP', 'EW', 'm', 'NDCC', 'sk', 'YE'?

EW = Etan Wexler <[hidden email]>

http://bugs.squeak.org/view.php?id=2454

Ken

At Thu, 21 Aug 2008 18:09:52 +0200,
Bert Freudenberg wrote:
  Cool.  I'll fetch these mcd, expand and suck the changes in.  Thank
you!

-- Yoshiki

At Thu, 21 Aug 2008 12:30:08 -0500,
Ken Causey wrote:
  Thanks!  Can somebody at UIUC should be able to nudge him, or
contact him?

-- Yoshiki

At Thu, 21 Aug 2008 13:25:01 -0500,
Ken Causey wrote:
  Quite likely.

-- Yoshiki

On Thu, 2008-08-21 at 13:37 -0500, Ken Causey wrote:
> On Thu, 2008-08-21 at 08:45 -0700, Yoshiki Ohshima wrote:
> > > > > >   (How about:
> > > > > >
> > > > > > 'ASF', 'edt', 'EP', 'EW', 'm', 'NDCC', 'sk', 'YE'?

For 'edt':

1 method: ObjectExplorerWrapper>>asString

Matthew Fulmer recommended a method using the Dual Changes Browser (a MC
1.5 tool, not the classic one) and a manual binary search.  Using this I
was able to narrow it down to somewhere between
Morphic-stephaneducasse.51 and Morphic-CdG.55.  Unfortunately none of
the intermediate versions are available for checking.  I strongly
suspect that this is part of the

http://bugs.squeak.org/view.php?id=1915

issue related to the addition of PlusTools to the image.

Ring any bells Andreas?

Ken

At Thu, 21 Aug 2008 14:00:45 -0500,
Ken Causey wrote:
  Ah, good!

-- Yoshiki