Attending: Chris Muller, Bert Freudenberg, Colin Putney, Randal Schwartz, Levente Uzonyi, Jecel Assumpacao Jr., Chris Cunnington
Notes on the 4.3 release:
- this won’t be the Cog release because of the needs of Linux which might not be aware of a sudden change
- Jecel is going to look at Linux distros and the vms they use
- Colin will talk to Esteban about multiple versions of the vm and get the OmniBrowser to run on 4.3
- Chris C. will go through the Daily Commit Logs back to the last release to compile a list of changes to the 4.3 release
Notes on Package Management ( Or, what would need to happen for Nicolas Cellier to take Complex out of the image?):
- The question was discussed if Squeak Map would be used for package management or not in future
- In the future Squeak Map may be the tool that lists packages that conform to a series of tests which are a de facto contract between developers and core contributors
- If that’s the case, Squeak Map needs some work. But owing to some low profile changes that have already been made to Squeak Map, perhaps not as much as one might expect
On 19/10/11 14:21, Chris Cunnington wrote:
Attending: Chris Muller, Bert Freudenberg, Colin Putney, Randal Schwartz, Levente Uzonyi, Jecel Assumpacao Jr., Chris Cunnington
Notes on the 4.3 release:
- this won’t be the Cog release because of the needs of Linux which
might not be aware of a sudden change
- Jecel is going to look at Linux distros and the vms they use
What's wrong with Linux? Squeak 4.2 + CogVM running fine here on Ubuntu 9.10
Image ----- /home/doc/std/dev/squeak/squeak4/Squeak4.2/Squeak4.2-10966.image Squeak4.2 latest update: #10966 Current Change Set: webcam-fixes
Virtual Machine --------------- /home/doc/std/dev/squeak/squeak4/Squeak4.2/squeak Croquet Closure Cog MT VM [CoInterpreterMT VMMaker.oscog-eem.114] Unix built on Aug 8 2011 08:24:13 Compiler: 4.1.2 20080704 (Red Hat 4.1.2-48)
- Colin will talk to Esteban about multiple versions of the vm and get
the OmniBrowser to run on 4.3
- Chris C. will go through the Daily Commit Logs back to the last
release to compile a list of changes to the 4.3 release
Notes on Package Management ( Or, what would need to happen for Nicolas Cellier to take Complex out of the image?):
- The question was discussed if Squeak Map would be used for package
management or not in future
- In the future Squeak Map may be the tool that lists packages that
conform to a series of tests which are a de facto contract between developers and core contributors
- If that’s the case, Squeak Map needs some work. But owing to some
low profile changes that have already been made to Squeak Map, perhaps not as much as one might expect
This body part will be downloaded on demand.
"Derek" == Derek O'Connell doc@doconnel.f9.co.uk writes:
- Jecel is going to look at Linux distros and the vms they use
Derek> What's wrong with Linux? Squeak 4.2 + CogVM running fine here on Derek> Ubuntu 9.10
The problem we discussed is that the packages with most linux distributions are *not* the CogVM yet (rightfully so). So we are trying to address the issue of "get vm from packages, get image from squeak.org".
If that's a smaller issue than we should worry about... as in, everyone who would be downloading 4.3 is smart enough to also pick up the CogVM as well, then maybe we can just blast forward. Thoughts?
On Wed, Oct 19, 2011 at 4:53 AM, Randal L. Schwartz merlyn@stonehenge.com wrote:
"Derek" == Derek O'Connell doc@doconnel.f9.co.uk writes:
- Jecel is going to look at Linux distros and the vms they use
Derek> What's wrong with Linux? Squeak 4.2 + CogVM running fine here on Derek> Ubuntu 9.10
The problem we discussed is that the packages with most linux distributions are *not* the CogVM yet (rightfully so). So we are trying to address the issue of "get vm from packages, get image from squeak.org".
If that's a smaller issue than we should worry about... as in, everyone who would be downloading 4.3 is smart enough to also pick up the CogVM as well, then maybe we can just blast forward. Thoughts?
If we choose to not include cog then we should either include a modest, useful image that explains how to get more packages, what cog is and how to get it, OR we should include a text file that explains how to get an image, additional packages,and what cog is and how to get it.
If we assume that everyone ... is smart enough we may end up losing a lot of new folks.
On Wed, Oct 19, 2011 at 07:53:32AM -0700, Randal L. Schwartz wrote:
The problem we discussed is that the packages with most linux distributions are *not* the CogVM yet (rightfully so). So we are trying to address the issue of "get vm from packages, get image from squeak.org".
The challenge is finding people with the knowledge and long term commitment to support the various Linux package distributions. This is not an easy thing to do, and I suspect that it requires a significant amount of work from anyone who might volunteer for one of those roles.
Assuming that we can find people willing and able to take the package maintainer roles, then we would want to encourage those people to build and distribute from sources that are recent enough to support the Cog image format (specifically, that would be October 2010 or later in the case of a standard VM). The sources should come from an upstream provider (either Ian's tarball or equivalent from Eliot or Igor).
If that can be achieved, then we will have no problem with running the latest Squeak image on any of these distros. But getting people to serve in the distro maintainer role may be a big challenge, as it is a lot of work that probably does not come with a lot of appreciation.
So my summary would be that there are no real techical impediments at all, but maybe an opportunity to define a distro maintainer role and to make sure that our community gives some appreciation to anyone willing to step forward and take on the hard work of building, packaging, testing, answering questions, soothing license anxieties, etc.
For someone taking on a distro maintainer role, all we can offer by way of compensation is fame, glory, and the appreciation of a few million users. It ain't much, but it ain't nothing either.
Dave
On 19/10/11 15:53, Randal L. Schwartz wrote:
Derek> What's wrong with Linux? Squeak 4.2 + CogVM running fine here Derek> on Ubuntu 9.10
The problem we discussed is that the packages with most linux distributions are *not* the CogVM yet (rightfully so). So we are trying to address the issue of "get vm from packages, get image from squeak.org".
If that's a smaller issue than we should worry about... as in, everyone who would be downloading 4.3 is smart enough to also pick up the CogVM as well, then maybe we can just blast forward. Thoughts?
Anyone using Squeak is ipso facto "smart enough" :-) The real issue is are they sufficiently informed and can safe-guards put in place if required (hence the suggestion in another email to auto-back-up an image).
My initial response was to the implication that Linux presented some sort of barrier to progress. Linux packaging is not a prerequisite for installing other software, it's an option that helps manage the process and not always a gaurentee of stability. Packaging is definitely not a restriction on what users can do or what software creators can supply. Official package maintainers do a great job and should be fully supported but the reality is that packages will often lag latest developments. The curious end-user, who may be new to Squeak, should have an easy way to explore regardless (even if just a zipped vm, image and start-up script for dumping in a single folder). So, while I appreciate the problem/s mentioned here and elsewhere I don't think they should hold up progress and nothing should be delayed due to packaging requirements. I'm not a packager but isn't it also true that between major releases, aprt from bug fixes, packages typically get updated when end-users request it?
Another reason to take this approach is the general perception and/or lack of current knowledge of people who "know" Smalltalk. Here's a recent quote from the Raspberry-Pi forums (links below) from a general discussion of the pro's and con's of various programming languages...
"Smalltalk has been covered by MarkSmith and about the only other "con" I can think of is the relative slowness of the extra indirection caused by message passing rather than direct function calling."
Anyone reading that will think all Smalltalks are slow and therefore not worthy of consideration. Best way to overcome misconceptions is to provide concrete evidence to the contrary.
Forums: http://www.raspberrypi.org/?page_id=43 Post: http://www.raspberrypi.org/?page_id=43&mingleforumaction=viewtopic&t...
Em 19-10-2011 11:21, Chris Cunnington escreveu:
Attending: Chris Muller, Bert Freudenberg, Colin Putney, Randal Schwartz, Levente Uzonyi, Jecel Assumpacao Jr., Chris Cunnington
Notes on the 4.3 release:
- this won’t be the Cog release because of the needs of Linux which
might not be aware of a sudden change
Cog in (Fedora 14/15) Linux still have some problems:
a) Keyboard input differs from "traditional VM" and makes it almost impossible to enter accented characters like áéíóúãõâêô etc... b) FFI is not completely compatible with "traditional VM" and certain things like CroquetGL just won't work like they do with "traditional VM" c) Compiler recompileAll crashes CogVM
Jecel is going to look at Linux distros and the vms they use
Colin will talk to Esteban about multiple versions of the vm and get
the OmniBrowser to run on 4.3
- Chris C. will go through the Daily Commit Logs back to the last
release to compile a list of changes to the 4.3 release
Notes on Package Management ( Or, what would need to happen for Nicolas Cellier to take Complex out of the image?):
- The question was discussed if Squeak Map would be used for package
management or not in future
- In the future Squeak Map may be the tool that lists packages that
conform to a series of tests which are a de facto contract between developers and core contributors
- If that’s the case, Squeak Map needs some work. But owing to some
low profile changes that have already been made to Squeak Map, perhaps not as much as one might expect
On Wed, Oct 19, 2011 at 1:55 PM, Casimiro de Almeida Barreto < casimiro.barreto@gmail.com> wrote:
Em 19-10-2011 11:21, Chris Cunnington escreveu:
Attending: Chris Muller, Bert Freudenberg, Colin Putney, Randal Schwartz, Levente Uzonyi, Jecel Assumpacao Jr., Chris Cunnington
Notes on the 4.3 release:
- this won’t be the Cog release because of the needs of Linux which might
not be aware of a sudden change
Cog in (Fedora 14/15) Linux still have some problems:
You need to be more specific. Which Cog? Mine or Igor's?
a) Keyboard input differs from "traditional VM" and makes it almost impossible to enter accented characters like áéíóúãõâêô etc... b) FFI is not completely compatible with "traditional VM" and certain things like CroquetGL just won't work like they do with "traditional VM" c) Compiler recompileAll crashes CogVM
Not so. My VMs on linux haven't been crashing in recompileAll for several months now. So which specific Cog? All the version info is in recent VMs:
Smalltalk getSystemAttribute: 1009 'r2502 http://www.squeakvm.org/svn/squeak/branches/Cog'
Jecel is going to look at Linux distros and the vms they use
Colin will talk to Esteban about multiple versions of the vm and get the
OmniBrowser to run on 4.3
- Chris C. will go through the Daily Commit Logs back to the last release
to compile a list of changes to the 4.3 release
Notes on Package Management ( Or, what would need to happen for Nicolas Cellier to take Complex out of the image?):
- The question was discussed if Squeak Map would be used for package
management or not in future
- In the future Squeak Map may be the tool that lists packages that conform
to a series of tests which are a de facto contract between developers and core contributors
- If that’s the case, Squeak Map needs some work. But owing to some low
profile changes that have already been made to Squeak Map, perhaps not as much as one might expect
Em 19-10-2011 19:27, Eliot Miranda escreveu:
Smalltalk getSystemAttribute: 1009 'r2502 http://www.squeakvm.org/svn/squeak/branches/Cog'
Here what I got: r2496 http://www.squeakvm.org/svn/squeak/branches/Cog
I'll try again downloading cog from squeakvm, recompiling & checking. To do so:
$ svn co http://www.squeakvm.org/svn/squeak/branches/Cog/
Then I follow the how to build. Only change is that UUID is made an internal plugin.
Besides, the problem with accented characters will be tested again. Also the problem with FFI... which FFI I'm supposed to use? Currently loaded latest FFI-Kernel, FFI-Pools, FFI-Tests, FFI-Unix but CroquetGL just won't work...
Cheers,
CdAB
On Thu, Oct 20, 2011 at 10:22 AM, Casimiro de Almeida Barreto < casimiro.barreto@gmail.com> wrote:
Em 19-10-2011 19:27, Eliot Miranda escreveu:
Smalltalk getSystemAttribute: 1009 'r2502 http://www.squeakvm.org/svn/squeak/branches/Cog'
Here what I got: r2496 http://www.squeakvm.org/svn/squeak/branches/Cog
I'll try again downloading cog from squeakvm, recompiling & checking. To do so:
$ svn co http://www.squeakvm.org/svn/squeak/branches/Cog/
Then I follow the how to build. Only change is that UUID is made an internal plugin.
Besides, the problem with accented characters will be tested again. Also the problem with FFI... which FFI I'm supposed to use? Currently loaded latest FFI-Kernel, FFI-Pools, FFI-Tests, FFI-Unix but CroquetGL just won't work...
What happens when you run a VM downloaded from http://www.mirandabanda.org/files/Cog/VM/VM.r2502/, e.g. coglinux.tgzhttp://www.mirandabanda.org/files/Cog/VM/VM.r2502/coglinux.tgz instead of one you build yourself?
Cheers,
CdAB
Em 20-10-2011 15:45, Eliot Miranda escreveu:
What happens when you run a VM downloaded from http://www.mirandabanda.org/files/Cog/VM/VM.r2502/, e.g. coglinux.tgz http://www.mirandabanda.org/files/Cog/VM/VM.r2502/coglinux.tgz instead of one you build yourself?
-- best, Eliot
Ok, then I have:
Smalltalk getSystemAttribute: 1009 Now returns correct version (r2502).
Accented characters wont work (as expected from previous e-mails).
Compiler recompileAll wont crash VM (as expected)
OpenGL example still wont work: external module not found:
OGLUnixLE(OpenGL)>>glPixelStorei:with: glPixelStorei: pname with: param "This method was automatically generated." "void glPixelStorei(GLenum pname, GLint param);" <apicall: void 'glPixelStorei' (ulong long) module: 'opengl32.dll'> ^self externalCallFailed
Interesting thing is that it's trying to open a opengl32.dll (ms win32) while in the "standard VM version" FFI works correctly... I'll investigate more about this...
On Thu, 20 Oct 2011, Casimiro de Almeida Barreto wrote:
Em 20-10-2011 15:45, Eliot Miranda escreveu:
What happens when you run a VM downloaded from http://www.mirandabanda.org/files/Cog/VM/VM.r2502/, e.g. coglinux.tgz http://www.mirandabanda.org/files/Cog/VM/VM.r2502/coglinux.tgz instead of one you build yourself?
-- best, Eliot
Ok, then I have:
Smalltalk getSystemAttribute: 1009 Now returns correct version (r2502).
Accented characters wont work (as expected from previous e-mails).
Compiler recompileAll wont crash VM (as expected)
OpenGL example still wont work: external module not found:
OGLUnixLE(OpenGL)>>glPixelStorei:with: glPixelStorei: pname with: param "This method was automatically generated." "void glPixelStorei(GLenum pname, GLint param);" <apicall: void 'glPixelStorei' (ulong long) module: 'opengl32.dll'> ^self externalCallFailed
Interesting thing is that it's trying to open a opengl32.dll (ms win32) while in the "standard VM version" FFI works correctly... I'll investigate more about this...
It's not trying to open opengl32.dll. The first line of OpenGL >> #initialize is a hack, which rewrites all methods behind the scene to use the platform's library name. The library names are defined in the subclasses of OpenGL. Look for #openGLLibraryName.
Levente
squeak-dev@lists.squeakfoundation.org