On Thu, 18 Oct 2001, Torge Husfeldt wrote:

> Hi Bert, Hi G.J.,
>
> Thanks Bert, that you jumped in...
> The "Fix" I posted was really only ment as a workaround and I should
> have mentioned that it was likely to youse problems with existing code
> in the image.
> We need to remember that standard swikis are deployed without .changes
> file and with the preference #warnIfNoChangesFile turned _off_ but the
> CMs all still carry their source pointers with them. So if you try to
> file out one of them it is quite natural that you stumble into problems.

Really? I had hoped Jochen would at least use the "abandonSources"
procedure when deploying the swiki. That should get rid of source pointers
and leave all in a usable way. (I usually skip the organisation zapping
part, though: makes browsing much more fun).

> As Bert mentioned this will only work for the code _newly_ added/changed
> (by hand or by fileIn) _after_ the changes file was installed. If you
> want to achieve overall better results with browsing/fileOut you should
> really get the BigTime.changes that belongs to the BigTime.image
> originally. There seems to be another way if you execute in your image
> _before_ you create the .changes file:
> Smalltalk lastQuitLogPosition
> Then on creating the .changes file you pad it with the number obtained
> of a appropriate character. What means appropriate in this circumstance
> I don't know and I would have to poke into it a bit and would only do so
> if somebody badly needed this functionality ($  $! or Charcacter value:0
> are my best guesses so far).

Sounds pretty hackish to me ;-)

> Regards,
> Torge
>
> P.S.: Wouldn't it be nice if there was a way to tell squeak that it
> keeps all new changes even if there is no changes file? Maybe even
> source code pointer that could point to (more or less) arbitrary files -
> for fileIns that is.

I'd rather keep them in the image - otherwise you'll get a lot of dangling
pointers to a lot of changesets. But it sounds like a good idea.

-- Bert

> Bert Freudenberg wrote:
> >
> > On Tue, 16 Oct 2001 [EMAIL PROTECTED] wrote:
> >
> > > OK I did make an empty file bigTime.changes
> >
> > If your image is named bigTime.image that is correct.
> >
> > > Then I did try to file out current change set
> >
> > Why? You should try to file in the fix again. That copies the source code
> > >from the fix file into your changes file so when you browse that
> > method again it will still have the formatting and comments. OTOH, nothing
> > is wrong with looking at decompiled code as long as you just want to use
> > that code.
> >
> > OTTH, if you really want to dig into the ComSwiki sources, you should
> > order the *real* image/changes file from Jochen. Or is that already
> > available for d/l somewhere?
> >
> > > I got an error window messages not undestood: runs
> > > (it ends up after 20 walkbalk lines somewhere in the menuItemMorph with
> > > BlockContext>ensure:
> >
> > That seems to be a bug with filing out decompiled code (which you should
> > not do anyway, but which is still a bug).
> >
> > -- Bert
>

Reply via email to