Smalltalk › Squeak › Squeak - Dev

SqueakMap in 4.1

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

11 messages Options

Göran Krampe

SqueakMap in 4.1

Hi folks!

Ok, so I am meddling with SM in trunk etc. SMBase is updated in trunk,
and so is SMLoader. SMBase had several issues in trunk, one is more
interesting than others:

I have tried to maintain SM outside of the image earlier - but since the
code sits in the image I understand that people start hacking on it. :)
The effect is that there have been some changes to SMBase in trunk that
only work in *later images*. For example, changing ifNotNilDo: to
ifNotNil:. That does NOT work in Squeak 3.8 for example.

So now I have backed out a few of these fixes so that the code still
works in 3.8. For the moment - since I don't want to "disturb" the trunk
way of things too much - just keep in mind that if you fix stuff in
SMBase, the code is CURRENTLY meant to work in 3.8 and higher!

I am just about now upgrading the server image - it still runs 3.8 (I
just don't have the time to do a full upgrade there), but uses the same
SMBase code as trunk has.

Finally, for SMLoader there are TONS of fixes now available on trunk. I
don't know exactly why this stuff didn't get into the trunk image
earlier - it is work done years back! Anyway, my advice for 4.1:

- Upgrade SMBase to the latest I committed to trunk. Otherwise it will
fail when trying to talk to the server!

- Hold the SMLoader upgrade. It is larger and there are THREE slightly
different SMLoader classes in that snapshot... More on that in another
thread.

regards, Göran

Andreas.Raab

Re: SqueakMap in 4.1

On 4/11/2010 3:27 PM, Göran Krampe wrote:

> Finally, for SMLoader there are TONS of fixes now available on trunk. I
> don't know exactly why this stuff didn't get into the trunk image
> earlier - it is work done years back! Anyway, my advice for 4.1:
>
> - Upgrade SMBase to the latest I committed to trunk. Otherwise it will
> fail when trying to talk to the server!
>
> - Hold the SMLoader upgrade. It is larger and there are THREE slightly
> different SMLoader classes in that snapshot... More on that in another
> thread.

Sounds good to me. And of course a big thanks for these updates.

Cheers,
- Andreas

Bert Freudenberg

ifNotNilDo: (was Re: [squeak-dev] SqueakMap in 4.1)

In reply to this post by Göran Krampe

On 12.04.2010, at 00:27, Göran Krampe wrote:
>
> I have tried to maintain SM outside of the image earlier - but since the code sits in the image I understand that people start hacking on it. :) The effect is that there have been some changes to SMBase in trunk that only work in *later images*. For example, changing ifNotNilDo: to ifNotNil:. That does NOT work in Squeak 3.8 for example.
>
> So now I have backed out a few of these fixes so that the code still works in 3.8. For the moment - since I don't want to "disturb" the trunk way of things too much - just keep in mind that if you fix stuff in SMBase, the code is CURRENTLY meant to work in 3.8 and higher!

I ran into exactly the same problem this weekend. Wanted to work on packetizing the Etoys image, using Karl's script as a start. But I got stuck when filing in recent Monticello versions gave a *syntax* error because ifNotNilDo: was replaced by ifNotNil:.

I agree with you that some packages should be kept compatible with older image versions. In particular those packages that are necessary to upgrade older code bases.

Does anyone have a snippet to find in a package all ifNotNil: sends that have a 1-argument block?

- Bert -

Göran Krampe

Re: ifNotNilDo: (was Re: [squeak-dev] SqueakMap in 4.1)

Bert Freudenberg wrote:
> I ran into exactly the same problem this weekend.

Another one was #fileNamed:do: (or perhaps #newFileNamed:do:) IIRC.

regards, Göran

Bert Freudenberg

Re: ifNotNilDo: (was Re: [squeak-dev] SqueakMap in 4.1)

On 12.04.2010, at 13:53, Göran Krampe wrote:
>
> Bert Freudenberg wrote:
>> I ran into exactly the same problem this weekend.
>
> Another one was #fileNamed:do: (or perhaps #newFileNamed:do:) IIRC.

But that's easier dealt with since it is not a syntax error.

Monticello had another syntax error: method annotations (a preference declaration).

Hmm, seems futile at this point. Guess I'll have to apply Eliot's changes first.

- Bert -

Levente Uzonyi-2

Re: ifNotNilDo: (was Re: [squeak-dev] SqueakMap in 4.1)

In reply to this post by Bert Freudenberg

On Mon, 12 Apr 2010, Bert Freudenberg wrote:

> On 12.04.2010, at 00:27, Göran Krampe wrote:
>>
>> I have tried to maintain SM outside of the image earlier - but since the code sits in the image I understand that people start hacking on it. :) The effect is that there have been some changes to SMBase in trunk that only work in *later images*. For example, changing ifNotNilDo: to ifNotNil:. That does NOT work in Squeak 3.8 for example.
>>
>> So now I have backed out a few of these fixes so that the code still works in 3.8. For the moment - since I don't want to "disturb" the trunk way of things too much - just keep in mind that if you fix stuff in SMBase, the code is CURRENTLY meant to work in 3.8 and higher!
>
> I ran into exactly the same problem this weekend. Wanted to work on packetizing the Etoys image, using Karl's script as a start. But I got stuck when filing in recent Monticello versions gave a *syntax* error because ifNotNilDo: was replaced by ifNotNil:.
>
> I agree with you that some packages should be kept compatible with older image versions. In particular those packages that are necessary to upgrade older code bases.
>
> Does anyone have a snippet to find in a package all ifNotNil: sends that have a 1-argument block?

This should do it (without high-level tools):

| methods |
methods := (PackageInfo named: 'EToys') methods
collect: #compiledMethod
as: IdentitySet.
SystemNavigation default browseAllSelect: [ :method |
(methods includes: method) and: [
(method getSource asString reject: #isSeparator) includesSubString: 'ifNotNil:[:' ] ].

Comments may be a problem with this approach, but that's pretty unlikely
to happen.

Levente

>
> - Bert -
>
>
>

Tobias Pape

Re: ifNotNilDo: (was Re: [squeak-dev] SqueakMap in 4.1)

Hi,

Am 2010-04-12 um 18:26 schrieb Levente Uzonyi:

> (method getSource asString reject: #isSeparator) includesSubString: 'ifNotNil:[:' ] ].

Given _I_ would have written Code there,
you wouldn't find any of it with this line,
as I would have used 'ifNotNil: [:'.
some other would have used 'ifNotNil:[ :'
and even other 'ifNotNil: [ :'.
anyhow, what about bytecode inspection? *duck&hide* ;)

So Long,
-Tobias

Levente Uzonyi-2

Re: ifNotNilDo: (was Re: [squeak-dev] SqueakMap in 4.1)

On Mon, 12 Apr 2010, Tobias Pape wrote:

> Hi,
>
> Am 2010-04-12 um 18:26 schrieb Levente Uzonyi:
>
>> (method getSource asString reject: #isSeparator) includesSubString: 'ifNotNil:[:' ] ].
>
> Given _I_ would have written Code there,
> you wouldn't find any of it with this line,
> as I would have used 'ifNotNil: [:'.
> some other would have used 'ifNotNil:[ :'
> and even other 'ifNotNil: [ :'.

The separators are rejected, so all variations you mentioned will be
found.

> anyhow, what about bytecode inspection? *duck&hide* ;)

That could work too, but that's probably a lot more complicated.

Levente

>
> So Long,
> -Tobias
>
>
>

Eliot Miranda-2

Re: ifNotNilDo: (was Re: [squeak-dev] SqueakMap in 4.1)

In reply to this post by Bert Freudenberg

Hi Bert,

On Mon, Apr 12, 2010 at 2:18 AM, Bert Freudenberg <[hidden email]> wrote:

On 12.04.2010, at 00:27, Göran Krampe wrote:
>
> I have tried to maintain SM outside of the image earlier - but since the code sits in the image I understand that people start hacking on it. :) The effect is that there have been some changes to SMBase in trunk that only work in *later images*. For example, changing ifNotNilDo: to ifNotNil:. That does NOT work in Squeak 3.8 for example.
>
> So now I have backed out a few of these fixes so that the code still works in 3.8. For the moment - since I don't want to "disturb" the trunk way of things too much - just keep in mind that if you fix stuff in SMBase, the code is CURRENTLY meant to work in 3.8 and higher!

I ran into exactly the same problem this weekend. Wanted to work on packetizing the Etoys image, using Karl's script as a start. But I got stuck when filing in recent Monticello versions gave a *syntax* error because ifNotNilDo: was replaced by ifNotNil:.

Why not upgrade the compiler with Vassili's ifNotNil: enhancement? I believe versions of his changes go back to 3.8. He has a change set for 3.8 or thereabouts on his web site. All systems with my closure compiler support ifNotNil: with an argument. So there are lots of sources for the enhancement. I could get this done for you tomorrow night if you're leery of doing it yourself.

best

Eliot

I agree with you that some packages should be kept compatible with older image versions. In particular those packages that are necessary to upgrade older code bases.

Does anyone have a snippet to find in a package all ifNotNil: sends that have a 1-argument block?

- Bert -

Bert Freudenberg

Re: ifNotNilDo: (was Re: [squeak-dev] SqueakMap in 4.1)

On 12.04.2010, at 20:47, Eliot Miranda wrote:

Hi Bert,

On Mon, Apr 12, 2010 at 2:18 AM, Bert Freudenberg <[hidden email]> wrote:
On 12.04.2010, at 00:27, Göran Krampe wrote:
>
> I have tried to maintain SM outside of the image earlier - but since the code sits in the image I understand that people start hacking on it. :) The effect is that there have been some changes to SMBase in trunk that only work in *later images*. For example, changing ifNotNilDo: to ifNotNil:. That does NOT work in Squeak 3.8 for example.
>
> So now I have backed out a few of these fixes so that the code still works in 3.8. For the moment - since I don't want to "disturb" the trunk way of things too much - just keep in mind that if you fix stuff in SMBase, the code is CURRENTLY meant to work in 3.8 and higher!

I ran into exactly the same problem this weekend. Wanted to work on packetizing the Etoys image, using Karl's script as a start. But I got stuck when filing in recent Monticello versions gave a *syntax* error because ifNotNilDo: was replaced by ifNotNil:.

Why not upgrade the compiler with Vassili's ifNotNil: enhancement? I believe versions of his changes go back to 3.8. He has a change set for 3.8 or thereabouts on his web site. All systems with my closure compiler support ifNotNil: with an argument. So there are lots of sources for the enhancement. I could get this done for you tomorrow night if you're leery of doing it yourself.

best
Eliot

That would be wonderful, thanks! :)

- Bert -

I agree with you that some packages should be kept compatible with older image versions. In particular those packages that are necessary to upgrade older code bases.

Does anyone have a snippet to find in a package all ifNotNil: sends that have a 1-argument block?

- Bert -

Bert Freudenberg

Re: ifNotNilDo: (was Re: [squeak-dev] SqueakMap in 4.1)

On 13.04.2010, at 02:29, Bert Freudenberg wrote:

On 12.04.2010, at 20:47, Eliot Miranda wrote:
Hi Bert,

On Mon, Apr 12, 2010 at 2:18 AM, Bert Freudenberg <[hidden email]> wrote:
On 12.04.2010, at 00:27, Göran Krampe wrote:
>
> I have tried to maintain SM outside of the image earlier - but since the code sits in the image I understand that people start hacking on it. :) The effect is that there have been some changes to SMBase in trunk that only work in *later images*. For example, changing ifNotNilDo: to ifNotNil:. That does NOT work in Squeak 3.8 for example.
>
> So now I have backed out a few of these fixes so that the code still works in 3.8. For the moment - since I don't want to "disturb" the trunk way of things too much - just keep in mind that if you fix stuff in SMBase, the code is CURRENTLY meant to work in 3.8 and higher!

I ran into exactly the same problem this weekend. Wanted to work on packetizing the Etoys image, using Karl's script as a start. But I got stuck when filing in recent Monticello versions gave a *syntax* error because ifNotNilDo: was replaced by ifNotNil:.

Why not upgrade the compiler with Vassili's ifNotNil: enhancement? I believe versions of his changes go back to 3.8. He has a change set for 3.8 or thereabouts on his web site. All systems with my closure compiler support ifNotNil: with an argument. So there are lots of sources for the enhancement. I could get this done for you tomorrow night if you're leery of doing it yourself.

best
Eliot

That would be wonderful, thanks! :)

- Bert -

No need to anymore, did it myself. Vassili's changeset from 3.9 worked.

- Bert -