Concurrent Futures

>>> Maybe it's time to think about Squeak 3.9.1?
>>>
>>> Giovanni
>>>
>> I did think about it for a long while, there are/were scripts  
>> available
>> with bugs to be harvested for this...
>> but to be honest I conceded that 3.10 is as good a 3.9.1 as you are
>> likely to get.
>
> That is not true for someone who uses 3.9 in production settings.  
> For anyone who is using a particular version of Squeak in  
> production the step to the next version needs to be very carefully  
> evaluated. Providing a small set of fixes addressing the most  
> important problems without fear "what else it may break" is a very  
> valuable service. Unfortunately, few people in the Squeak community  
> seem to understand that.

> Adrian Lienhard wrote:
>
>> At netstyle.ch we are using 3.9 for many projects now and we will  
>> not move to 3.10 soon. We're using some of the critical fixes, so  
>> basically that means we already have our own 3.9.1 image...
>> I assume, quite some people are using 3.9 in production as well.  
>> So, if somebody is interested we could join forces and release a  
>> 3.9.x from time to time.
>
> Which fixes are you using? It might make sense to also include them  
> in the 3.8.2, if not already.
> The 3.8.2 updates should be working in 3.9 as well, but need to  
> check. So we would have at least a starting point.
>
> Michael
>
>

>>>> Maybe it's time to think about Squeak 3.9.1?
>>>>
>>>> Giovanni
>>>>
>>> I did think about it for a long while, there are/were scripts  
>>> available
>>> with bugs to be harvested for this...
>>> but to be honest I conceded that 3.10 is as good a 3.9.1 as you are
>>> likely to get.
>>
>> That is not true for someone who uses 3.9 in production settings.  
>> For anyone who is using a particular version of Squeak in  
>> production the step to the next version needs to be very carefully  
>> evaluated. Providing a small set of fixes addressing the most  
>> important problems without fear "what else it may break" is a very  
>> valuable service. Unfortunately, few people in the Squeak  
>> community seem to understand that.
>
> Indeed.
>
> At netstyle.ch we are using 3.9 for many projects now and we will  
> not move to 3.10 soon. We're using some of the critical fixes, so  
> basically that means we already have our own 3.9.1 image...
> I assume, quite some people are using 3.9 in production as well.  
> So, if somebody is interested we could join forces and release a  
> 3.9.x from time to time.

>
> Norbert
>
>> 0006581: Image freezes (background processes like Seaside make no
>> progress) and Squeak hoggs CPU
>>
>> -----
>> 0006588: Broken Semaphore>>critical: leads to frozen processes in  
>> Delay
>>
>
>> -----
>> 0006576: Delay is not thread-safe
>>
>
>
>

>>
>> At netstyle.ch we are using 3.9 for many projects now and we will  
>> not move to 3.10 soon. We're using some of the critical fixes, so  
>> basically that means we already have our own 3.9.1 image...
>> I assume, quite some people are using 3.9 in production as well.  
>> So, if somebody is interested we could join forces and release a  
>> 3.9.x from time to time.
>
> I imagine a lot of reason not moving to 3.10 but can you let us  
> know yours? :)

>
> On Nov 15, 2007, at 10:32 , stephane ducasse wrote:
>>>
>>> At netstyle.ch we are using 3.9 for many projects now and we will
>>> not move to 3.10 soon. We're using some of the critical fixes, so
>>> basically that means we already have our own 3.9.1 image...
>>> I assume, quite some people are using 3.9 in production as well. So,
>>> if somebody is interested we could join forces and release a 3.9.x
>>> from time to time.
>>
>> I imagine a lot of reason not moving to 3.10 but can you let us know
>> yours? :)
>
> Well, in general, upgrading can be a major effort and you always have
> the risk that stuff will break in unexpected ways. Obviously, problems
> and the effort to fix them are potentially expensive and customers may
> loose their confidence.
>
> In particular, I haven't looked in detail what happened in 3.10, but
> since 3.9 has proven to be a very good release (excluding the recently
> addressed production problems, which have been around for a long
> time), we will stick to for now. In summary, I trust in 3.9 more than
> in 3.10 and I don't know a reason that would justify the potential
> hassle to upgrade.
>
> Adrian
>

>
> Hi folks!
>
> A few things for possible inclusion, I don't have time right now to
> check if some of these are already included but I think not.
>
> Two of them are nice speedups (time and printing SmallIntegers), the
> other one is a simple addition of behavior (parsing dates) that I wrote
> as a reaction to a post on the beginners list a while back. Oh, and I
> also added a fix for SMTPClient - well, two rolled into one changeset
> actually - harvested from the Gjallar project.
>
> I consider 6513 to be trivial to add (just added behavior - can't break
> stuff), 6512 seems trivial too (easy to verify), 4669 should be green
> with all Date/Time tests and has been recommended for inclusion earlier,
> but perhaps someone else did something in that area that has made it
> obsolete (Keith?), I don't recall. 6768 also seems easy to include, and
> it has been battle tested in Gjallar.
>
> regards, Göran
>
> -----------------------------------------
> http://bugs.squeak.org/view.php?id=6512
>
> Much faster printString (and thus also asString) for SmallIntegers:
> One override of #printString for SmallInteger and a new method for
> counting digits in base 10. This makes #printString for SmallIntegers
> more than 4 times faster than before.
> -----------------------------------------
> http://bugs.squeak.org/view.php?id=6513
>
>          Read a Date from the stream based on the pattern which can include
> the tokens:
>
>         y = A year with 1-n digits
>         yy = A year with 2 digits
>         yyyy = A year with 4 digits
>         m = A month with 1-n digits
>         mm = A month with 2 digits
>         d = A day with 1-n digits
>         dd = A day with 2 digits
>
>     ...and any other Strings inbetween. Representing $y, $m and $d is
> done using
>     \y, \m and \d and slash itself with \\. Simple example patterns:
>
>         'yyyy-mm-dd'
>         'yyyymmdd'
>         'yy.mm.dd'
>         'y-m-d'
>
>     A year given using only two decimals is considered to be >2000.
> --------------------------------------------
> http://bugs.squeak.org/view.php?id=4669
>
> Ok, so we were trying to figure out why our logger in Gjallar spent so
> much time in creating DateAndTime etc... after looking, tweaking,
> profiling and even some thinking I came up with this "SpeedPack" of
> changes.
>
> I might have broken some subtleties (I don't think I have though) but
> the tests are still green. Glad for any indepth review.
>
> And? Well, the speed differences are quite huge. :) Lots of it comes
> from better Duration instance creation methods and from not going back
> and forth too much between nanos and seconds etc.
>
> "DateAndTime now" is about 6 times faster and "Date today" 4 times. The
> full chronology test suite runs almost 2 times faster. See preamble for
> a do-it that shows this (run before and after installing cs).
>
> So far only tested in 3.8.
> --------------------------------------------
> http://bugs.squeak.org/view.php?id=6768
>
> Added support for EHLO (sending SMTP using Exchange) and fixed a problem
> with authentication - #initiateSession needs to be sent before login
> occurs.
>
>