If you're overwriting the update, then you might as well put the code in
Socket class>>initialize, instead of a do it, so that
unloading/reloading it will do the right thing.

Daniel

Doug Way <dway at riskmetrics.com> wrote:
> Andreas Raab wrote:
> 
> >...
> >Secondly, it seems that the Network rewrite updates lack some initialization
> >code. Socket>>deadServer responds with nil which makes it impossible to load
> >further updates without manually resetting the dead server, e.g.,
> >	Socket deadServer: ''
> >does the trick. I think some additional update just containing the above
> >should solve that problem for anyone who updates through the entire sequence
> >but people having stopped after the network rewrite will have the do the
> >above manually.
> >  
> >
> 
> I just tried adding a new update 5260 on the internal (testing) stream, 
> and that doesn't work because the dead server stops any updates after 
> 5259 from being loaded, even if they are done in sequence.
> 
> So, it looks like I'll need to just overwrite update 5259 so that it 
> also includes the " Socket deadServer: '' " reset in its postscript.  
> Overwriting updates has been done on rare occasions in the past, I 
> believe. :-)  Either way, people who have already updated and saved a 
> 5259 image will need to execute the do-it themselves.
> 
> (Speaking of not being out of the woods yet, the update broadcasting 
> mechanism seems to be broken now too.  I ftp-ed these updates directly 
> to the server to test them.  I'll have to look at this problem tonight...)
> 
> - Doug Way