Magma March 2006

magma@lists.squeakfoundation.org

9 participants
21 discussions

10 MB limit?
by Bert Freudenberg 22 Mar '06

22 Mar '06

I just tried to import an existing domain model into Magma, by giving it as root when creating the repository. The models serialized size was 6 MB before (using DataStream). However, the commit failed with a notice that a commit must not be larger than 10 MB. Now I worked around that problem by splitting the model and re-adding parts in separate commits later, but wouldn't it be nice for a one- time import to lift the limit? Why is it there? - Bert -

1 0

A small gotcha and a proposed fix
by goran＠krampe.se 22 Mar '06

22 Mar '06

Hi Chris and folks! With the current implementation of MagmaSession>>commit: you can make an "early exit" and miss the actual commit like this: session commit: [ ^mObject foo: 3] Which will set foo to 3 and then return the result of that - unfortunately missing the actual commit in MagmaSession>>commit: since that code looks like this: commit: aBlock | result | self begin. result := aBlock value. self commit. ^result Since Magma "supports" nested transactions (it just keeps a nest level count and the actual commit occurs when you are all the way back out - so it all turns into a BIG commit - fair enough) the effect is that you now are one level down and subsequent calls to commit: will of course not do anything. And yes - it was a bit painful to figure that stupid thing out. :) Better would probably be: commit: aBlock self begin. ^[aBlock value] ensure: [self commit] If there is an error in "self begin" we don't try to commit but if "aBlock value" returns we don't miss the commit. regards, Göran

1 0

RE: For interest: Magma in 3.9a-7008
by Mart-Mari Breedt 22 Mar '06

22 Mar '06

Hallo all, Sorry for only replying now. We, South Africans, had a public holiday yesterday. Today is my first day back at the office. Marcus: Thank you for the patch in 7013. I got 7015 this morning and it did indeed make a big difference. The Magma Monticello load is still a bit slower than in 3.8, but it is a lot more acceptable now. Chris: Don't worry, I also have no idea how Magma works in 3.9 yet either (I don't even know how 3.9 works :-)). But it is always nice to try something new and there are already a couple of things that I like in 3.9 (or would like to explore). After loading Magma, I couldn't get the unit tests to run completely. The test procedure gets stuck in the middle of 82 tests and when I alt-. , I see that it has something to do with LargeInteger>>+. It looks as if primitive 21 didn't work correctly. I am still very new to the primitive concept, but I will continue to dig around (if and when time permits) and see if I can figure out what actually is going on. (BTW: The call stack originates from MaHashIndexRecord>>byteArrayIndexesFrom:to:do:.) Then there is also the deprecated method, but I am not sure how that affects everything yet. Once again, thanks for everyone's help and comments. Kind regards, Mart-Mari -----Original Message----- From: Petr [mailto:petr@madnetwork.org] Sent: 20 Maart 2006 03:29 To: magma(a)lists.squeakfoundation.org Subject: Re: For interest: Magma in 3.9a-7008 Hi Marcus, I did have a full Monticello installed. Petr Marcus Denker wrote: > Hi, > > Did you have Monticello installed in the 3.8 image? If not, then it uses > MCInstaller for loading the package, which is much faster then Monticello > itself (which is installed in 3.9a). > > Other then that: Yes, it's slow... we are looking into it a bit as > time permits. > Friday we found a huge bottleneck of Monticello to be calling #category > on Classes a lot, which does a linear search... 7013 has a patch that > makes #category ~ 10 times faster on everage (for classes "at the end" > which newly loaded classes are, much much more). This turns out to > double MC loading speed... amazing. > > I'm sure there are a lot of cases like this in Squeak... we only need > to find > and fix them ;-) > > Marcus > > On 17.03.2006, at 13:30, Mart-Mari Breedt wrote: > >> Hi, >> >> Sorry about the re-asking. I only joined the new list today. >> >> I had the same symptons... During most of the load there was no load on >> the CPU as well. >> >> We can continue discussing this on squeak-dev. Perhaps someone on that >> side has an idea of what might be wrong. >> >> Kind regards, >> >> Mart-Mari >> >> -----Original Message----- >> From: Petr [mailto:petr@madnetwork.org] >> Sent: 17 Maart 2006 02:20 nm >> To: magma(a)lists.squeakfoundation.org >> Subject: Re: For interest: Magma in 3.9a-7008 >> >> Hi, >> I asked this question here in my e-mail few days ago. I am mostly using >> the 3.9a images and when I tried to load Magma into 3.8, it was >> lightning fast. During most of the loading time, there is was no load on >> >> CPU. It happens with all packages in 3.9a loaded from Monticello. I >> guess we should discuss this issue in the squeak-dev. >> >> Petr >> >> Mart-Mari Breedt wrote: >>> Hallo all, >>> >>> I am not sure whether this mail belongs on the Magma mailing list or >> the >>> Squeak mailing list. Since my problem only seems to occur when >>> installing Magma, I thought it better to post it here first. >>> >>> This morning I was playing a bit with the new squeak image, 3.9a-7008. >> I >>> found that installing Magma1.0-cmm.4.mcz (from a local copy) takes >> about >>> twenty-five (!) minutes. The installation otherwise progresses fine, >>> except for a warning towards the end of the installation: >>> "CompiledMethod>>who has been deprecated. use #methodClass and >> #selector >>> directly", but you can still proceed in this case. The same >> installation >>> in 3.8-6665 doesn't even take a minute. >>> >>> Maybe someone can venture a guess as to why the Magma installation >> takes >>> so much longer in the new image? >>> >>> Kind regards, >>> >>> Mart-Mari >>> >>> _______________________________________________ >>> Magma mailing list >>> Magma(a)lists.squeakfoundation.org >>> http://lists.squeakfoundation.org/mailman/listinfo/magma >>> >>> >> >> _______________________________________________ >> Magma mailing list >> Magma(a)lists.squeakfoundation.org >> http://lists.squeakfoundation.org/mailman/listinfo/magma >> >> >> > > _______________________________________________ > Magma mailing list > Magma(a)lists.squeakfoundation.org > http://lists.squeakfoundation.org/mailman/listinfo/magma > _______________________________________________ Magma mailing list Magma(a)lists.squeakfoundation.org http://lists.squeakfoundation.org/mailman/listinfo/magma

1 0

RE: For interest: Magma in 3.9a-7008
by Mart-Mari Breedt 20 Mar '06

20 Mar '06

Hi, Sorry about the re-asking. I only joined the new list today. I had the same symptons... During most of the load there was no load on the CPU as well. We can continue discussing this on squeak-dev. Perhaps someone on that side has an idea of what might be wrong. Kind regards, Mart-Mari -----Original Message----- From: Petr [mailto:petr@madnetwork.org] Sent: 17 Maart 2006 02:20 nm To: magma(a)lists.squeakfoundation.org Subject: Re: For interest: Magma in 3.9a-7008 Hi, I asked this question here in my e-mail few days ago. I am mostly using the 3.9a images and when I tried to load Magma into 3.8, it was lightning fast. During most of the loading time, there is was no load on CPU. It happens with all packages in 3.9a loaded from Monticello. I guess we should discuss this issue in the squeak-dev. Petr Mart-Mari Breedt wrote: > Hallo all, > > I am not sure whether this mail belongs on the Magma mailing list or the > Squeak mailing list. Since my problem only seems to occur when > installing Magma, I thought it better to post it here first. > > This morning I was playing a bit with the new squeak image, 3.9a-7008. I > found that installing Magma1.0-cmm.4.mcz (from a local copy) takes about > twenty-five (!) minutes. The installation otherwise progresses fine, > except for a warning towards the end of the installation: > "CompiledMethod>>who has been deprecated. use #methodClass and #selector > directly", but you can still proceed in this case. The same installation > in 3.8-6665 doesn't even take a minute. > > Maybe someone can venture a guess as to why the Magma installation takes > so much longer in the new image? > > Kind regards, > > Mart-Mari > > _______________________________________________ > Magma mailing list > Magma(a)lists.squeakfoundation.org > http://lists.squeakfoundation.org/mailman/listinfo/magma > > _______________________________________________ Magma mailing list Magma(a)lists.squeakfoundation.org http://lists.squeakfoundation.org/mailman/listinfo/magma

3 2

magma install troubles
by Marcelo Cortez 17 Mar '06

17 Mar '06

hi all why is dificult to install magma from monticello package? the last version don´t install at all , for me the version MagmaServerLoader-cmm.11 Process to install: from http://minnow.cc.gatech.edu/squeak/2657 MCHttpRepository location: 'http://squeaksource.com/MagmaTester' user: '' password: '' the version MagmaServerLoader-cmm.13,MagmaServerLoader-cmm.12 don't install at all. ther version MagmaServerLoader-cmm.11 isntall if only if first install cryptographic package at hand or form sm , then install. the other versions is working? into versions MagmaServerLoader-cmm.13,MagmaServerLoader-cmm.12 this script fail MakoSecurityManager current createKeyringFor: 'sume' MakoSecurityManager current beMute: true MakoSecurityManager current saveKeyring MagmaRepositoryController create: 'c:\magma' root: Dictionary new fail. any clue? best regards MDC

2 1

For interest: Magma in 3.9a-7008
by Mart-Mari Breedt 17 Mar '06

17 Mar '06

Hallo all, I am not sure whether this mail belongs on the Magma mailing list or the Squeak mailing list. Since my problem only seems to occur when installing Magma, I thought it better to post it here first. This morning I was playing a bit with the new squeak image, 3.9a-7008. I found that installing Magma1.0-cmm.4.mcz (from a local copy) takes about twenty-five (!) minutes. The installation otherwise progresses fine, except for a warning towards the end of the installation: "CompiledMethod>>who has been deprecated. use #methodClass and #selector directly", but you can still proceed in this case. The same installation in 3.8-6665 doesn't even take a minute. Maybe someone can venture a guess as to why the Magma installation takes so much longer in the new image? Kind regards, Mart-Mari

3 2

RE: For interest: Magma in 3.9a-7008
by Mart-Mari Breedt 17 Mar '06

17 Mar '06

Hi, You are right, it probadly has nothing to do with Magma, it just happened to be the only package I had this problem with. I first thought that it might be because of the dependancies of Magma1.0, but I loaded some other packages that also had depandacies, and they seemed to be working fine. Maybe it is related to a new version of Monticello? I can load the newest version of Monticello in my 3.8 image and check it out. Kind regards, Mart-Mari -----Original Message----- From: goran(a)krampe.se [mailto:goran@krampe.se] Sent: 17 Maart 2006 09:33 To: Mart-Mari Breedt Cc: magma(a)lists.squeakfoundation.org Subject: Re: For interest: Magma in 3.9a-7008 Hi! This has nothing to do with Magma I am pretty sure. But I haven't looked at what it is - my guess is that it is Traits related - or related to other changes made to the core system related to that. regards, Göran

1 0

some more questions
by Petr 14 Mar '06

14 Mar '06

Hello, I have some more questions about Magma. The first one is, about objects which can be stored. Here http://minnow.cc.gatech.edu/squeak/5722 is written, that any subclasses of Set and Dictionary are not supported and they are not expected to work. Why? Does it mean, if I subclass a Dictionary, I will not be able to save it in Magma? My second question is about about the time required to load Magma into my image. This is probably more a problem to be addressed to the squeak-dev list. When I load Magma into fresh 3.8 image, it is pretty fast, under 2 minutes. When I do the same in 3.9, it takes ages. It happens with many other packages loaded from Monticello. Third, is this list going to be archived by Gmane? I prefer it over mailing, as I can easily look up things in older posts. gmane.comp.lang.smalltalk.squeak.magma contains only information about moving the list to mailman and there are no old messages and new are not recorded. Petr

2 1

unable to get Magma working
by Petr 13 Mar '06

13 Mar '06

Hello, after 3 hours of try-and-fail, I have decided to ask someone more clever. I am trying to install Magma into virgin 3.8basic image. Loading order 1) Monticello 2) SqueakMap 3) Cryptography 0.3 from SqueakMap (I also tried lates MC version from SqueakSource, but it somehow didn't work for me with next step) 4) latest MagmaServerLoader from SqueakSource up to this step everything is fine, loads without errors. Now comes the problem. I tried to do this: MagmaRepositoryController create: 'c:\myMagmaFolder' root: Dictionary new. MagmaServerConsole new open: 'c:\myMagmaFolder' ; processOn: 1010 Then I found out, that I have to do something first with the KryptOn. I tried steps from http://minnow.cc.gatech.edu/squeak/5785, but everytime I save or load the image it asks for password twice and then it gives me some error. Any hints or some step by step guide for a braindead? Thanks, Petr

2 2

Attention: this Magma list is moving!
by Chris Muller 08 Mar '06

08 Mar '06

Ken Causey has offered to move our mailing list discussion managed by the ezmlm program at http://discuss.squeakfoundation.org/cgi-bin/ezmlm-browse? to the mailman program at http://lists.squeakfoundation.org/mailman/listinfo The existing message archive will move too, however everyone will need to resubscribe to the new list at http://list.squeakfoundation.org/mailman/listinfo/magma which should be available by the time you read this or very soon afterward. I want to personally thank Ken, Marcus and the squeakfoundation for their support and hosting our discussion. See you at the new list! - Chris

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

Magma March 2006