Hi Levente --

> What you described sounds like a bug in the update mechanism.

Maybe. But loading new methods first can also break the system if those new methods override existing ones in the super class.

Best,
Marcel
Am 21.09.2021 13:59:33 schrieb Levente Uzonyi <leves at caesar.elte.hu>:
Hi Marcel,

On Tue, 21 Sep 2021, Marcel Taeumel wrote:

> Hi all --
> I just moved Kernel-eem.1415 from Trunk to Inbox because it breaks the update stream due to missing checkpoints in the update map.

The update mechanism ought to
- load all new methods first
- then apply method changes
- and finally delete methods
What you described sounds like a bug in the update mechanism.

>
> I am in favor of Levente's fix (i.e., Kernel-ul.1415) because it does not need any checkpoints but directly loads into Trunk.

I tend to prefer my version because it keeps the trigger for #grow in a
single place (HashedCollection >> #atNewIndex:put:) in the whole
HashedCollection hierarchy, and it does not duplicate code.


Levente

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20210921/4a0f2932/attachment.html>