Squeak-dev April 2008

squeak-dev@lists.squeakfoundation.org

146 participants
187 discussions

Thread Safe Collections
by Guillermo Adrián Molina 24 Apr '08

24 Apr '08

Hi! I would like to discuss an implementation of thread safe collections (and streams, and others). I can think on three possible approaches, the first one is to convert every class to be thread safe, the other is to implement new thread safe classes. I think that the first approach is interesting because the user wouldn't have to care about any thing to be thread safe. We would just use the class that is guaranteed to be thread safe by the system. But there would be an overhead imposed by the synchronization mechanism needed by a thread safe implementation. The second approach would let you choose between fast thread unsafe collections, and a slower thread safe collections. But you would have to choose it yourself. It would not be automatic. This could be done by subclassing every collection class like this: Object +Collection +ThreadSafeCollection +Bag |+ThreadSafeBag +Set +ThreadSafeSet Or doing an entirely new hierarchy like this (with Traits to implement similar behavior): Object +Collection |+Bag |+Set +ThreadSafeCollection +ThreadSafeBag +ThreadSafeSet What do you think? Cheers Guille

4 3

Re: [squeak-dev] Thread Safe Collections
by Bill Schwab 24 Apr '08

24 Apr '08

Guille, It took me a while to get comfortable with this position, but having done things this way for a long time, I now think it is correct: the explicitly thread-safe collection (SharedQueue, weak collections, thread-safe dictionary-like classes) are a natural part of the thread synchronization tools that happen to be useful as collections. They are thread-safe because they need to be. In general, I doubt you will help yourself very much by creating other thread-safe collections, as there will still be operations that should be atomic, and in the name of performance and avoiding a false sense of security, "do it the hard way." Actually, it is not all that hard (most of the time<g>). Over-protect with critical sections and then test: deadlocks will reveal themselves particularly well in call-stacks for all non-dead threads; dirty reads and writes are more trouble to find, so when in doubt, protect; it is easy to let it deadlock and then back off as appropriate. IMHO, trying to create a universal solution to synchronization is a prescription for headaches. Bill Wilhelm K. Schwab, Ph.D. University of Florida Department of Anesthesiology PO Box 100254 Gainesville, FL 32610-0254 Email: bschwab(a)anest.ufl.edu Tel: (352) 846-1285 FAX: (352) 392-7029

1 0

About Etoys and another Morphic creatures
by Edgar J. De Cleene 24 Apr '08

24 Apr '08

In #IRC with some people we discuss about future directions of Squeak. A point of disagree with any "MinimalMorphic" is which Morphs should be into. For me , I wish a image which could save and load Projects, Morphs and have Flaps. For Etoys, I think they should go out, to a his own package and maintained by OLPC people or others who know more about. But should be loadable as option. I hear about Etoys troubles in 3.10. I glad to show with attached picture , Etoys are alive and projects created in 3.10 could be loaded without any trouble in FunSqueak and also in old 3.9 dirty image I also use. If any wish, I could create a TestEtoys page in swiki with this very single project for any wishing test Project/Etoys loading and work in shrinked images. Edgar

1 0

Where has my memory gone
by Keith Hodges 24 Apr '08

24 Apr '08

Dear All, I have written a very useful little utility that tells me how much memory each Monticello package is using up. This was going to end up in a package called "Senility", however I added it as a utility to Monticello 1.5. In the very latest Monticello, the left hand panel, now has a menu "package sizes". enjoy Keith

1 0

Re: [squeak-dev] Mac OSX ID
by Norbert Hartl 24 Apr '08

24 Apr '08

On Wed, 2008-04-23 at 11:28 -0700, tim Rowledge wrote: > On 23-Apr-08, at 10:35 AM, Norbert Hartl wrote: > > > On Wed, 2008-04-23 at 15:00 +0200, Javier Reyes wrote: > >> Hi Edgar, > >> > >> It's for a licensing server. I think I've found a way to do it using > >> the ethernet physical ID trough: > >> > >> Applescript doIt: 'do shell script "ifconfig en0"'. > >> > >> And then filtering the physical address. > >> > > This won't work. Usually the MAC address shown is only a copy > > of the number stored on the device. You can alter this address > > easily. > I assume simply replacing the ethernet card - or even just adding one > and reassigning which is en0 - would mess it up pretty badly. > > Many CPus have an actual unique ID that is findable somewhere. Even > there you're in trouble if the cpu has to be replaced due to failure > of upgrade. > > Yes, that is hardly avoidable. I think that's the reason so many companies still use some sort of dongle. Norbert

2 1

Mac OSX ID
by Javier Reyes 24 Apr '08

24 Apr '08

Hello ! I need to obtain a unique hardware ID from Apple OSX machines using squeak. ¿Any hint? Thanks in advance, Javier

7 9

gstreamer on the mac?
by John M McIntosh 24 Apr '08

24 Apr '08

Various folks have asked can't we have GStreamer on os-x? Well yes if you care to install a hundred packages in a terminal session. But what if you statically link everything? Ok I tried that, I get a binary that links fine, but sadly after gstreamer is initialized, when we try to load a GStreamer plugin we get (<unknown>:11870): GLib-GObject-CRITICAL **: gtype.c:2242: initialization assertion failed, use g_type_init() prior to this function The obvious thing would be toput g_type_init() in the GStreamer initialization, not that it's needed when it's dynamically linked, well yes I've done that, and still no go. I'm wondering if anyone wants to help, I can supply the XCode package, or of course you can use the GStreamer makefile with the needed linker statements to build a binary. -- = = = ======================================================================== John M. McIntosh <johnmci(a)smalltalkconsulting.com> Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com = = = ========================================================================

2 1

printShowingDecimalPlaces: is broken in 3.10
by bryce＠kampjes.demon.co.uk 23 Apr '08

23 Apr '08

Execute the following in a workspace to reproduce the bug. 1.009 printShowingDecimalPlaces: 3 '1.900' It looks like a performance improvement went wrong. I've created Mantis entry 7028 for this which includes a fix along with some SUnit tests. I've set it to severity minor, but it's a showstopper if you depend on that functionality. The fix was trivial though. Bryce

1 0

Sake/Packages declared open!
by Keith Hodges 23 Apr '08

23 Apr '08

Dear All, Sake/Packages is now at a point where it is a useful resource for the community to contribute to. To help this along I have set the squeaksource.com commit emails to go to packages(a)lists.squeakfoundation.com Discussion as to what should go where and whether or not specific configurations work or not can be directed there also. I believe that the Sake/Packages package definitions will be a useful resource for all of the Squeak communities. Please join me in in this knowledge gathering exercise, please bring what you know about what loads where and feel free to fill in the package definitions. Its very easy, I updated all of Seaside/Scriptaculous in 3.7 through to 3.10 in about 10 mins. To load Sake/Packages into a LevelPlayingField (LPF) image: Installer install: 'Packages'. If you want to maintain all of the published versions of your package you may prefer to load. Installer install: 'PackageAllVersions'. If you are using squeak 3.10 the following hierarchy of classes will load. 1. Packages - Abstract Base Class 2. PackagesAllVersions - Non-specific package definitions that load the latest version, if that version is likely (but not 100% guarenteed) to work in all versions. You don't have to wait for your package to be perfect, to put a generic definition in here, enabling your users to try it out in their favourite squeak version, and provide you with feedback. 3. PackagesSqueak310beta - Non-specific package definitions that load the very latest version, specific to 3.10. Again it is expected to work, but the author may not have actually tested your configuration. 4. PackagesSqueak310U - The specific exact published version in Universes that is supposed to have been tested in that image version. 5. PackagesSqueak310 - The specific exact version that 'you'/'we' have found 'actually' works, overriding the universes definition. Make it easy for Authors and Users. ========================== Benefits for Authors - You do not have to constantly update your published packages all the time, you can publish a "load latest" script in the generic (2. 3.) levels of the hierarchy. Benefits for Users ============= Quality The community can clearly identify and publish what works where, and this provides a shared collective whiteboard in which active collaboration can take place. In addition, if a specific version in a specific image needs a patch to work, then it can actually be included in the load script for that package. Thus for the first time patches can be included in published packages, and released for testing, without the bottleneck of waiting for the Package author to get around to fixing the problem for your squeak-fork. Those who submit patches will actually see their work get used immediately. Automation All installation scripts should run without interaction, for automated installation. See 'Seaside' for an example. Deterministic Builds (potentially*) All of the package definitions, for the image you are using are version controlled in monticello. So you can go back to a previous configuration from a previous date and load it. *The packages you are loading have to be listed to nominate specific packages for this to work. Flexibility Have your own package definitions for your own squeak fork, subclass to specialize/override etc etc. I declare Sake/Packages for squeak open! enjoy Keith === btw... LevelPlayingField is currently broken in 3.8.1 and I dont have time to fix it, the SqueakMap update process chokes it. to do: Sake/Packages-SqueakMap the same scheme for p.s. In case anyone wonders... the packages mailing list has not been used for ~ 3 years. I emailed the packages mailing over a year ago and discussed its use for this purpose with anyone who cared.

8 14

GraphViz and saltypickle.com
by bryce＠kampjes.demon.co.uk 23 Apr '08

23 Apr '08

squeak.saltypicle.com is down and so GraphViz is unavailable on the Universes and SqueakMap. I'm updating Exupery pre-release and use it for graph layout in some custom inspectors. If John's not interested in maintaining it, I could create a project on SqueakSource and update the universes entry. Bryce

5 6

← Newer
1
2
3
4
5
6
7
8
9
...
19
Older →

Jump to page: