But Tim's point still stands, because deployment is the problem you
solve after having solved the "build it" problem :) frank On 7 November 2013 20:13, Chris Muller <[hidden email]> wrote: > Deployment should be regarded a separate problem than building. > > On Thu, Nov 7, 2013 at 2:00 PM, tim Rowledge <[hidden email]> wrote: >> >> >> One important thing to remember is coping with a first build on a new machine. Having a squeak app that can drive cc to build a vm would be very neat, but if you don't yet have a working vm for a device you could be in trouble. >> An advantage of the cmake process is that it bypasses that issue. Actually running the configure on the target machine means you get (hopefully!) accurate results when querying facilities. There is also some benefit in a production environment in being able to pass a 'normal' build job to someone else without having to find a squeak buildomatic cognizant person. >> > |
Well, I don't think Tim's point stands anyway. It assumes a blank OS on any machine can use configure and cmake, but not squeak. Can windows machines run configure and cmake out of the box? Some sort of installation hurdle must be overcome, whether cygwin or new-squeak-based-builder. People who are building VM's are a much smaller, more-technically capable group than those who need to deploy something (e.g., an application). If you design your deployment system for the former, it think it won't be useful to the latter.
On Thu, Nov 7, 2013 at 2:36 PM, Frank Shearar <[hidden email]> wrote: But Tim's point still stands, because deployment is the problem you |
Wait, why are we talking about deploying a Squeak application here?
We're talking about taking a blank machine and being able to build a Squeak VM on it. Tim's saying that you can't drive a C compiler from within the image on a machine for which no VM has ever been built, because there is no VM. ("Machine" here means a machine architecture, rather than, say, a fresh Ubuntu install.) You need a bootstrap. And configure/cmake/whatever happens to provide a very easy way to bootstrap to the first VM that can run an image that _can_ drive a C compiler to build the next gen VM. frank On 7 November 2013 20:45, Chris Muller <[hidden email]> wrote: > Well, I don't think Tim's point stands anyway. It assumes a blank OS on any > machine can use configure and cmake, but not squeak. Can windows machines > run configure and cmake out of the box? > > Some sort of installation hurdle must be overcome, whether cygwin or > new-squeak-based-builder. > > People who are building VM's are a much smaller, more-technically capable > group than those who need to deploy something (e.g., an application). If > you design your deployment system for the former, it think it won't be > useful to the latter. > > > On Thu, Nov 7, 2013 at 2:36 PM, Frank Shearar <[hidden email]> > wrote: >> >> But Tim's point still stands, because deployment is the problem you >> solve after having solved the "build it" problem :) >> >> frank >> >> On 7 November 2013 20:13, Chris Muller <[hidden email]> wrote: >> > Deployment should be regarded a separate problem than building. >> > >> > On Thu, Nov 7, 2013 at 2:00 PM, tim Rowledge <[hidden email]> wrote: >> >> >> >> >> >> One important thing to remember is coping with a first build on a new >> >> machine. Having a squeak app that can drive cc to build a vm would be very >> >> neat, but if you don't yet have a working vm for a device you could be in >> >> trouble. >> >> An advantage of the cmake process is that it bypasses that issue. >> >> Actually running the configure on the target machine means you get >> >> (hopefully!) accurate results when querying facilities. There is also some >> >> benefit in a production environment in being able to pass a 'normal' build >> >> job to someone else without having to find a squeak buildomatic cognizant >> >> person. >> >> >> > > > |
On Thu, Nov 7, 2013 at 3:28 PM, Frank Shearar <[hidden email]> wrote: Wait, why are we talking about deploying a Squeak application here? Hm, ok. That's a pretty rarely-occurring event to be putting a lot thought into a framework for. I'm doubtful such a framework would ever "work the first time" on new platforms.
For building deployable packages on existing platforms, Igors words really resonated with me. I'll butt out now. :)
|
In reply to this post by tty
Hi Timothy,
On Thu, Nov 7, 2013 at 12:33 PM, gettimothy <[hidden email]> wrote:
I mean good low-level programming skills. Knowing how stacks and address spaces are organized, what the machine's calling convention is, what a system call looks like, etc, and what the current processor's instruction set is, it's theory etc (is it a RISC or a CISC; does it have delay slots, etc, etc).
So first things first. Get the stack VM wokring in 64-bits.
OK, but right now you'll want to be in Smalltalk, and for quite a while.
It's not :-).
Read the latest two articles on my blog (A Spur gear for Cog & Lazy Become).
Make a VMMaker image (see e.g. Cog Blog :: Building a Cog Development Image).
Then read SpurMemoryManager, /especially/ its class comment. Then read its subclasses Spur32BitMemoryManager (functional) & Spur64BitMemoryManager (completely untested). Then read SpurBootstrap32.
Try and create a 32-bit Spur image and run it. Then try and implement SpurBootstrap64. Try and create a 64-bit Spur image and run it. This is /lots/ of work. In particular the last two steps require some significant design.
So start slowly and see how you get on.
best, Eliot
|
In reply to this post by Frank Shearar-3
On 7 November 2013 21:36, Frank Shearar <[hidden email]> wrote: But Tim's point still stands, because deployment is the problem you This is not an issue. You can always run VM on other machine, where it runs to generate config/source files, and then build VM on machine where VM don't runs yet. Not mentioning cross-compiling approach. And that technique already employed in pharo building for number of builds, like for iphone and android platforms. Even building ppa-distro for linux goes same way: its just takes configuration & sources (and all source files) generated by build server and built somewhere on other end of the world in linux ppa build farms. frank -- Best regards, Igor Stasenko. |
In reply to this post by Eliot Miranda-2
Awesome, Thank you. t
|
In reply to this post by Eliot Miranda-2
Eliot, Re "I mean good low-level programming skills. Knowing how stacks and address spaces are organized, what the machine's calling convention is, what a system call looks like, etc, and what the current processor's instruction set is, it's theory etc (is it a RISC or a CISC; does it have delay slots, etc, etc)." Would an updated edition of Patterson and Hennessy be helpful here? My copy (like my c skills) is from 1993. t.
|
In reply to this post by Chris Muller-4
On 7 November 2013 22:50, Chris Muller <[hidden email]> wrote:
Thanks for supportive words :) For android, which does not supported by cmake Dimiry Golubovsky just managed to do it in a way how it can work: he wrote a configuration (in smalltalk)
which generates files which compatible with android build infrastructure. I don't know the details, but you can always look at CMakeVMMaker-Android category in CMakeVMMaker package. That means two things:
- CMakeVMMaker despite its name actually does not means 'cmake only'. - the guy with little knowledge of android nuances (like me), can easily pick up this work and modify it in case of need. What else i can say? If people for some reason don't like it, it would be stupid to force it upon them. The thing is there, open source, and everyone is welcome to use or abuse it. So, if you don't like it, you're free to write own.. I informed that same work been done (and it works, works well). so rather than starting from scratch, i would twink twice, because i could take what's already done as a starting point and improve or change it (if i don't like something). So the decision whether to use it or write own from scratch is up to you, people.
I'm out. -- Best regards, Igor Stasenko. |
In reply to this post by tty
On Thu, Nov 7, 2013 at 3:16 PM, gettimothy <[hidden email]> wrote:
nothing fundamental that affects Cog has changed in that time... it'll do just fine.
best, Eliot
|
Thanks. I took a gander at Spur--very challenging and new, lots to learn. Just what the doctor ordered (: t
|
In reply to this post by Eliot Miranda-2
Eliot First, if these discussion is inappropriate for squeak-dev, I have subscribed to squeak-vm-beginners and we can move it there or elsewhere if I am creating too much noise here. In working the steps at http://www.mirandabanda.org/cogblog/build-image/ and I have found that running all the tests for the Alien install on the SqueakVM that comes with the all-in-one app fails all but one of its tests. (primitives fail) If I start the image using the CogVM then all the tests pass. So, is it correct to say that building a Cog Development Image on Squeak4.4 or greater depends on a pre-built CogVM from http://www.mirandabanda.org/files/Cog/VM/ ? I am assuming yes, but I want to verify with you before I document what I am doing thx. t |
In reply to this post by Igor Stasenko
Hi Igor,
hi list, sorry for replying late,… My reply was intended to stir the conversation a litte, to keep it going, which, I assume, it did. I really like your opinion, Igor :) On 07.11.2013, at 14:28, Igor Stasenko <[hidden email]> wrote: >> >> >> >> On 7 November 2013 08:58, Tobias Pape <[hidden email]> wrote: >> Hey Tim and all, >> On 07.11.2013, at 05:00, Ron Teitelbaum <[hidden email]> wrote: >> >> > Hey Tim, >> > >> > Göran and Eliot have been discussing this. I just talked to Eliot about it >> > yesterday. He said he was happy to support cmake but didn't want to have to >> > do the work himself. I'm hoping that the work Göran is doing on this now >> > can be turned over to Eliot when it is finished. I think there is agreement >> > that combining the excellent work of pharo build with Eliot's new vm work is >> > a very good thing to do. The work to merge the build environments is >> > already being done at 3D ICC. We are working on this for the Mac also but >> > have not discussed it with Ian yet. >> > >> > Ron Teitelbaum >> > >> >> -----Original Message----- >> >> From: [hidden email] [mailto:squeak-dev- >> >> [hidden email]] On Behalf Of tim Rowledge >> >> Sent: Wednesday, November 06, 2013 9:04 PM >> >> To: The general-purpose Squeak developers list >> >> Subject: Re: [squeak-dev] A Bounty for CMake-ifying stack/Cog vm build >> > process >> >> >> >> >> >> Wow, no interest? Is it because make is painful or because nobody needs to >> >> make any more money? >> >> >> I actually have interest but I am currently reluctant, >> for time-capturing reasons in my personal and work life. >> >> However, Here is what I would do: >> >> • Start from Ians CMake work, and try to generalize it for the >> non-Linux platforms >> • Windows would be kind-of straight-forward, >> • For OS X, I would really want to have the Cocoa-based VM stuff >> (aka Squeak VM 5.4.7.x) integrated and proper Xcode files >> generated, especially to be able to have iOS VMs built automatically. >> Here, some information from John McIntosh would be helpfull, especially >> a simple walk-through. >> I have some experience with that[1] >> >> >> Despite all admiration I have for the way the PharoVM is built, there is one >> thing that bothers me: >> • You first somehow generate the VM sources (which is fine, and the automated way is amazing) >> • THEN, the a Script generates CMake-Files >> • CMake then generates Makefiles >> • and then finally a vm is compiled. >> This is one generating step too much, for my taste. >> Personally, I'd prefer CMake to just pick up the C-files generated by VMMaker, >> so that adding a new platform does not need to change the VMMaker… >> >> The main big issue with 'platform-neutral' static source files, that they tend to >> grow with tons of #ifdef-s over time, reducing readability and creating a mess. >> > My point is that one or another way you have to deal with platform differences, > and the way how i prefer to do it is using smalltalk code, > but not .m4 files, or .cmake files, or writing a medium novel long configuration files > using strange (better to say weird) scripting DSL. > I can express all build process logic in smalltalk, which does not lessens the amount > of work to deal with all platform nuances and dependencies etc, but at least it allows me to organize and shape it > into some manageable form, which is easy to change & learn. > > And another difference is, that i prefer working with browser & classes than > with dozens of directories and files lying here and there. I did not want to imply we have to do it my way, not at all. And you are certainly right. Smalltalk is a good way to do things. The only thing that got me thinking is, a lot of people use tools like CMake to drive build processes, and they know the edges of the domain very well. Building software can be quite complex: think Platforms, compiler quirks, SDK restrictions, cross compilations, to just name a few. CMake (even more than autotools) is quite good at these tasks, especially regarding the cross-platform part. If we now start to, effectively, reimplement CMake or autotools, we will run in the issues, those tools all have solved years ago. CMake even has beautiful output. I think it would take an tremendous effort to get a Build tool that (only thinking of the Squeak/Pharo VM) • can build well with different compilers (Clang/Gcc, probably Microsoft) • can handle the platform-discovery (Linux, OSX, Windows, RISC!) • can handle library discovery (finding stuff like libuuid) • can do cross-compilation and the signing-stuff (iOS, perhaps android) If we get this working, this would be amazing. I am a bit reluctant, it would be a lot of work for me… > > But sure thing, it is a question of taste. If you prefer to maintain this: > ------------------------- > [22 lines old CMAKE] > -------------------------- > And/or this: > -------------------------- > [96 lines autoconf/make/whatever] > ---------------------------- > > instead of plain smalltalk code, it is up to you. > > But that's about 'too much' for my taste. I concur. My impression is, we can choose from four “evils”[1]: 1. Accept the status quo. Painful for starters with its 8+ ways of doing essentially the same and every single one has some kind of big drawback. 2. Use CMake completely. Painful, because we have to maintain two bodies of knowledge, VMMaker and CMake. 3. Use cmake through CMakeVMMaker. Painful, because it can not fully benefit from Smalltalk nor from CMake. You have always to make up your mind whether to implement something on the Smalltalk or the CMake level 4. Have a Smalltalk-only build too. Painful, because it is going to take a lot of time. I can accept any of them, but the first :) 2. and 3. have the benefits that they are already there, to certain degrees. (2. only in the Unix-vm part, done by Ian, 3. currently only used by the Pharo community) How shall we as the VM community[2] decide? Best -Tobias [1] not really evil, only each painful to a certain degree [2] Who are we actually? • The Pharo and Squeak people, • The Cuis people • The Newspeak People • The CogVM people • Whom am I missing? signature.asc (1K) Download Attachment |
In reply to this post by tty
On Fri, Nov 08, 2013 at 07:29:42AM -0800, gettimothy wrote:
> Eliot > > First, if these discussion is inappropriate for squeak-dev, I have subscribed to squeak-vm-beginners and we can move it there or elsewhere if I am creating too much noise here. > Hi Timothy, The vm-dev list is here: http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/vm-dev It is not a beginner's list, and your discussion would be welcome there. Either way, you are not creating "noise" and it is great to see you getting involved. > In working the steps at http://www.mirandabanda.org/cogblog/build-image/ and I have found that running all the tests for the Alien install on the SqueakVM that comes with the all-in-one app fails all but one of its tests. (primitives fail) > > If I start the image using the CogVM then all the tests pass. > > So, is it correct to say that building a Cog Development Image on Squeak4.4 or greater depends on a pre-built CogVM from http://www.mirandabanda.org/files/Cog/VM/ ? > > I am assuming yes, but I want to verify with you before I document what I am doing > It is a good idea to use a VM from mirandaband.org, because different VMs may have different runtime behaviour, and using a runtime environment that matches Eliot's will prevent confusion. Strictly speaking, you can build a Cog VM using pretty much any VM that works with your Squeak4.4 image. So no it is not necessary to use a specific pre-built image, but it is a very good idea to do so, especially during the time when you are first learning how to do VM building. I think Eliot will give you a similar answer, so hopefully he won't mind me stepping in to answer some of your questions. Dave > thx. > > t > |
David, Thank you. I have joined that list and will continue the conversation there. Cordially, t. On Fri, Nov 08, 2013 at 07:29:42AM -0800, gettimothy wrote: |
In reply to this post by Tobias Pape
On 8 November 2013 16:49, Tobias Pape <[hidden email]> wrote: Hi Igor, hehe, do you think i felt something else when started work on trying make VM to be able to build automatically? i doubt anyone can find it fun to grok through dozens of dusty old source & configuration files in attempt to put some order there. i did what i did, and now we using it to build pharo. And you free to use hours we spent on what we did. as well as free to not, of course.
i don't wanna advertise or convince anyone because you can just download it and see by yourself.
-- Best regards, Igor Stasenko. |
In reply to this post by Eliot Miranda-2
Part 4 of the Blue Book--chapter 26 The implementation--in the intro there is this example:
With these operations
In a Workspace, when I doit on 3 + 7/2 I get 5, left, right, unary, binary....rules apply. What I see here--and please check my reasoning--is a RPN that places things onto the stack via Smalltalk's order of operation rules and then does conventional RPN processing. In other words, Smalltalk expressions are translated to RPN and then placed on the stack. A simple 'banish the thought' or 'yep' or 'sometimes' would be helpful as I move forward in my study. I am just trying to avoid carrying forward an improper idea that I will have to revise later when/if the facts turn out otherwise. thx. t |
On 19 November 2013 16:33, gettimothy <[hidden email]> wrote:
> Part 4 of the Blue Book--chapter 26 The implementation--in the intro there > is this example: > > > > center > ^ origin + corner / 2 > > > > With these operations > > > Rectangle center > 0 push the value of the receiver's first instance variable (origin) > onto the > stack > 1 push the value of the receiver's second instance variable > (corner) onto the > stack > 176 send a binary message with the selector + > 119 push the Smalllnteger 2 onto the stack > 185 send a binary message with the selector / > 124 return the object on top of the stack as the value of the message > (center) > > > In a Workspace, when I doit on 3 + 7/2 I get 5, left, right, unary, > binary....rules apply. Yep, because "message sending is all there is", which equates here to a lack of operator precedence. So "3 + 7/2" means "(3 + 7) / 2". + and / here are known as binary selectors. > What I see here--and please check my reasoning--is a RPN that places things > onto the stack via Smalltalk's order of operation rules and then does > conventional RPN processing. More generally, keyword sends take precedence over unary sends which take precedence over binary selectors. So in foo something: 1 + 2 negated with: bar the #negated message will be sent to the object 2, that will be the argument of the #+ message, sent to the object 1, and then that result and the object bar will be the arguments of the #something:with: message to the foo object. frank > In other words, Smalltalk expressions are translated to RPN and then placed > on the stack. > > A simple 'banish the thought' or 'yep' or 'sometimes' would be helpful as I > move forward in my study. I am just trying to avoid carrying forward an > improper idea that I will have to revise later when/if the facts turn out > otherwise. > > thx. > > t > > > |
In reply to this post by tty
Hi Timothy,
On Tue, Nov 19, 2013 at 8:33 AM, gettimothy <[hidden email]> wrote:
Yes, and so they are in most stack-based implementations of languages. Bytecode for a stack machine is a form of RPN.
best, Eliot
|
Eliot and Frank, Thank you.
|
Free forum by Nabble | Edit this page |