April 2010 - Pharo-dev - lists.pharo.org

[Pharo-project] Communication with VM guys (part two)
by StÃ©phane Ducasse April 9, 2010

April 9, 2010

Begin forwarded message: > From: Ian Piumarta <piumarta(a)speakeasy.net> > Date: April 9, 2010 12:35:16 PM GMT+02:00 > To: stephane ducasse <stephane.ducasse(a)gmail.com> > Cc: Squeak Virtual Machine Development Discussion <vm-dev(a)lists.squeakfoundation.org> > Subject: Re: [Vm-dev] Questions, communication and process > > Hi Stef, > > On Apr 9, 2010, at 2:23 AM, stephane ducasse wrote: > >> - How do people report problems? Just sending an email in the mailing-list is enough? > > Here are the options, necessarily biased towards my preference which is to try to avoid having to poll for outstanding issues. > > If it's a general problem of installation or bizarre behaviour that differs between platforms, ask for help on the Squeak VM mailing list. > > http://lists.squeakfoundation.org/mailman/listinfo/vm-dev > > Pharo and Squeak users are EQUALLY WELCOME to report VM issues on that list. Squeak has several people who regularly take the time to forward things from squeak-dev to vm-dev when appropriate. You might want to encourage the same thing on the Pharo list. I know for a fact that there are several vm-dev subscribers that monitor the Pharo list. (You might want to make it clear that Pharo uses the Squeak VM too. This does not necessarily appear to be universally understood.) > > If it's a Unix-specific problem that you think might be a bug, report it to the above mailing list AND send me email. Fixes are best submitted as complete modified files (not 'diffs'). (There might well be a way to tell ediff to use diff output directly, but that does not guarantee that your 'original' and my 'original' are in any way related. Sending whole files is MUCH more robust.) > > If it's Windows-specific, try the above list and Andreas Raab. For Mac-specific, the above list and John McIntosh. If it's common code shared by all platforms, especially code generated by VMMaker, send to Dave Lewis and the above list. > > VM bugs can fed to mantis here: bugs.squeak.org > > The people mentioned above probably look at mantis on a regular basis. Anything with more than four legs gives me the heebie-jeebies, which isn't why I don't regularly trawl mantis but it'll do. > >> - How a fix in one OS is propagated to the others? > > Fixes that affect more than one OS are very often in the common code that is generated automatically and shared between all the platforms. There is no explicit manual propagation required. (That being said, the people mentioned above are often in contact on a daily basis. When co-ordination is required, it will happen.) > > The exception is that some code is shared between Unix and Mac. John is good at spotting commits from me that affect him, and is good at bugging me when necessary into fixing anything Unixy that might affect him. > >> - How the user can know that? > > Watch vm-dev for discussions or commit reports that say the issue was fixed. If it is in mantis, it might even get closed too. > > The above reflects how I like to work, and I can only speak for myself. MPG from the others mentioned above might not be comparable. > > Cheers, > Ian > > (Feel free to forward, re-post, etc., the above as you see fit.) >

1 0

[Pharo-project] Communication with VM guys (part one)
by stephane ducasse April 9, 2010

April 9, 2010

Hi guys I will send a following mail. We should probably synthetize and put them some summary on the web. Stef Begin forwarded message: > From: Ian Piumarta <ian(a)vpri.org> > Date: April 9, 2010 11:43:33 AM GMT+02:00 > To: laurent laffont <laurent.laffont(a)gmail.com> > Cc: stephane ducasse <stephane.ducasse(a)free.fr>, stephane ducasse <stephane.ducasse(a)gmail.com> > Subject: Re: [Vm-dev] Fwd: loading library with FFI fails (was [Pharo-project] Linux vm: assertions and strategy) > > Hi Laurent, > >> I've just seen you reply me in private. Do you mind if I forward your mail to Pharo & VM mailing lists, to see how we can improve communication and collaboration between projects ? I don't want to start a "counterproductive" flame war, I hate this, so maybe we can write another mail which doesn't offense people ? Maybe you have ideas on this ? > > How about we just say this... > >> It seems there is a complete lack of communication within the Pharo community regarding the virtual machine and its development/support. VM development and support (including bug reports and fixes) are discussed on the vm-dev mailing list. Both Pharo users and Squeak users are equally welcome there (and both communities should avoid discussing VM problems on their respective developer lists, since the people who should see those VM bug reports are more likely reading vm-dev than the other lists). Problems that appear to be platform-independent should always be reported to the vm-dev list. Problems specific to a particular platform should be sent to the maintainer of the platform whose e-mail address is included in the release notes or in the download area on squeakvm.org. > > Nobody could be offended by that? :) > > Cheers, > Ian >

1 0

Re: [Pharo-project] [Vm-dev] Fwd: loading library with FFI fails (was Linux vm: assertions and strategy)
by Schwab,Wilhelm K April 9, 2010

April 9, 2010

Ian, I will try what you want. -----Original Message----- From: Ian Piumarta [mailto:ian@vpri.org] Sent: Friday, April 09, 2010 5:42 AM To: Schwab,Wilhelm K Cc: laurent laffont; Pharo-project(a)lists.gforge.inria.fr Subject: Re: [Vm-dev] Fwd: loading library with FFI fails (was [Pharo-project] Linux vm: assertions and strategy) Hi Bill, On Apr 9, 2010, at 3:24 AM, Schwab,Wilhelm K wrote: > I cannot ignore what was never delivered to me - you might want to > reconsider your logic in this connection. My initial assumption appears to be incorrect and so the suggestion I sent not to be relevant to debugging your problem. But the original is attached in case you care. It seems something between you and me is unreliable in delivery or in spam detection. > You might also note that #moduleName can check the OS, so "it will > never work on Windows" is just plain wrong. My mistake. The Unix moduleName should be 'odbc' unless there's a good reason for it not to be. > Later in the day I will revert to your vm and try 'odb' Recompile from source. My best attempt at a fix for your issue is in the repo. > - I will admit to being less than optimistic, especially given what I > "ignored" - make that never received. Well now that your mailer likes me, why not try to be happy to work with me to resolve this rather than assuming my attempts are futile and in vain? Cheers, Ian From: Ian Piumarta Date: April 7, 2010 11:56:14 AM PDT To: "Schwab,Wilhelm K" <bschwab(a)anest.ufl.edu> Subject: Re: [squeak-dev] RE: Linux 3.11-3 vm and shared libraries Received: (qmail 20182 invoked from network); 7 Apr 2010 18:56:15 -0000 Received: from dsl092-038-093.lax1.dsl.speakeasy.net (HELO [10.0.1.2]) by mail7.sea5.speakeasy.net (qmail-ldap-1.03) with SMTP for <bschwab(a)anest.ufl.edu >; 7 Apr 2010 18:56:15 -0000 Message-Id: <70021DC8-9A49-48C1-B26F-6E75820EDC38(a)speakeasy.net> In-Reply-To: <A93938CC72687245835AC0BE10199F7C7B73C7B4(a)HSC-CMS01.ad.ufl.edu > Content-Type: text/plain; charset=US-ASCII; format=flowed; delsp=yes Content-Transfer-Encoding: 7bit Mime-Version: 1.0 (Apple Message framework v936) References: <A93938CC72687245835AC0BE10199F7C7B73C767(a)HSC-CMS01.ad.ufl.edu > <A93938CC72687245835AC0BE10199F7C7B73C7B4(a)HSC-CMS01.ad.ufl.edu> Bill, On Apr 7, 2010, at 10:34 AM, Schwab,Wilhelm K wrote: It is one thing for my hacked together .so to avoid detection due to some mistake on my part If your .so is in the library search path then it should be found. If your .so contains unresolved references then they should be reported on the console when it fails to load. has anyone tried ODBC on Linux with the new vm? It's sitting in /usr/ lib like it has always been; the vm should find it, right? Yes it should, but try setting your LD_LIBRARY_PATH=/lib:/usr/lib and running the VM. I think there is something wrong with the library searches in 3.11-3. Just different, and a lot more predictable: From: Ian Piumarta <piumarta(a)gmail.com> Date: September 16, 2009 7:36:01 PM PDT To: The general-purpose Squeak developers list <squeak-dev(a)lists.squeakfoundation.org > Subject: 3.11.3.2135 for Unix [...] NOTE: FFI libraries are no longer searched for in every single nook and cranny, which means you might have to set LD_LIBRARY_PATH (or your local equivalent) for FFI-based programs to work properly. Cheers, Ian

1 0

Re: [Pharo-project] [Vm-dev] Fwd: loading library with FFI fails (was Linux vm: assertions and strategy)
by Schwab,Wilhelm K April 9, 2010

April 9, 2010

Ian, I cannot ignore what was never delivered to me - you might want to reconsider your logic in this connection. You might also note that #moduleName can check the OS, so "it will never work on Windows" is just plain wrong. Later in the day I will revert to your vm and try 'odb' - I will admit to being less than optimistic, especially given what I "ignored" - make that never received. Bill -----Original Message----- From: Ian Piumarta [mailto:ian@vpri.org] Sent: Friday, April 09, 2010 2:25 AM To: Schwab,Wilhelm K Cc: laurent laffont Subject: Re: [Vm-dev] Fwd: loading library with FFI fails (was [Pharo-project] Linux vm: assertions and strategy) On Apr 8, 2010, at 11:40 PM, Ian Piumarta wrote: > I have no idea. I sent email to Bill with a suggestion but he ignored > me. I therefore have no information about whether there is something > funny in his Linux installation or whether there is a real problem in > the VM (which appears to work fine for me). I could be completely wrong about this. If so, please disregard. But... From the patch I am inferring that the odbc library specifies "module: '/usr/lib/libodbc.so'". This would indeed be a problem for three reasons. 1. Mine might be in /usr/local/lib, or might be a personally-compiled version to fix some issue with the system version. In either case I'm screwed. 2. I might be using Darwin, in which case '/usr/lib/libodbc.dylib' might be the only thing that would work -- or even '/System/Library/ Frameworks/ODBCFramework/Libraries/libodbc.dylib'. Same comment as before. 3. It will never work on Windows, even if some day Windows supplies an 'odbc.dll' that is 100% compatible with the Linux one. The usual way for the plugin to specify this library would be "module: 'odbc'" which should work perfectly in all the above cases using the latest VM (and explains why I could not reproduce the problem using a simpler test case). I'm happy to modify the VM to cater for this specific situation of an over-specified library name, but the real solution would be fix the primitive declarations in the ODBC plugin. Cheers, Ian

1 0

[Pharo-project] [Preview] Help System with new model
by Torsten Bergmann April 9, 2010

April 9, 2010

Hi Stef and Michael, your responses have shown me that I may have not clearly communicated what is possible and what my goals are. It has also shown me that you mainly talk about the help system as to provide "API help" or reference documentation, which is possible but was not my main goal. So lets explain it in more detail: We have two separate cases/kind of help: A. Help for the system in front of you ("SystemHelp") Documentation of a tool or a process (General Tool Help/Tutorials/F1 Help/Written Book) B. Help on how the system in front of you is implemented ("SystemReference") Documentation of classes and methods, examples as in any other Smalltalk/Dev system ("API Help"/Class comments/Method comments) Regading A - Help for the system in front of you ("SystemHelp"): ================================================================ Just evaluate HelpBrowser openOn: SystemHelp or just HelpBrowser open My initial goal was to do have "F1 Help" as in most other IDE's so a help browser opens to display a complete help book with a full written Tutorial on how the Monticello Browser works. Or how one is able to work with Metacello or what object oriented programming is all about, or how one is able to create a project on Squeaksource. We could even have an ASCII version of "Pharo by Example", ... This kind of help doesnt even have to be related to stuff that is in the image. You could provide books on gardening and read them in the image if you like. ;) This kind of help could have been external in files or on a webpage, - but I since my goal was that one can read/display it also within the image I JUST (AB)USE classes and class hierarchies to store the help book contents. This also allows to manage the books and contents with usual Smalltalk tools and (if required) also to provide dynamic content. But typically these classes just provide the help contents and it is intended that they have no behavior! I dont want this to be limited to tutorials or tool help. Think of a case where one writes a card playing game with Pharo and want to provide a help book to the commercial end user explaining the game. I know it is much more effort to write such documentation - but for the dev-image/pharo tools we may provide at least simple books to explain basic things like the project itself, license, how to get started with the tools, ... Since this kind of help is typically structured like a real book there is the "CustomHelp" class allowing you to map your books structure directly to subclasses. If you write such a "class mapped" book you want to keep the book/subbooks/pages together and not spread around the image. And I call this "System help" since it shows you help to explain the system that is in front of you (either Pharo or the commercial card game). When you run HelpBrowser open this "SystemHelp" (see the according class SystemHelp) just collects all direct subclasses of CustomHelp as root books. Nothing more. So if you want to have your book appear in this "system wide help" you: 1. either store your book contents in the CustomHelp hierarchy (this is the typical case where you map the book to subclasses, see class "HelpOnHelp" or "PharoHelp" for example) 2. or provide a direct/indirect CustomHelp subclass to just define the book to appear in the system help but get the contents from other sources (see class "PharoAPIHelp" which allows us to provide the whole class API in the Pharo book) Especially the second option allows you to mix dynamic content like API Help into these "system help books". Just open the chapter "API Reference" in the "Pharo" book when you have the package "Pharo-Project-Help" loaded. More on this later. Dont get me wrong, it is alway possible to open the Help browser on any object that is able to convert itself into a HelpTopic instance using #asHelpTopic HelpBrowser openOn: myObjectWhoCanBeConverted but it is necessary to have a subclass of CustomHelp it appears in the default "system wide help" (SystemHelp) when you open the browser using the menu or by "HelpBrowser open". So CustomHelp is the entry point here, nothing more! To summarize: - A. is intended for whole books and a system wide help system - A. requires the class "CustomHelp" - to define at least the "root books" appearing in this "System help" - to allow mapping a book structure directly to a class hierarchy (and keep this book with its subbooks together) Providing such a help was the original goal of the help system which is now fully possible. Still plain ASCII and without support for markup, images, ... yet - but usable. Regarding B - Help on how the system in front of you is implemented ("SystemReference"): ====================================================================================== Just evaluate HelpBrowser openOn: SystemReference Beside A. there is also "Documentation of classes and methods" as we have in any other programming languages or Smalltalk systems. Maybe also enhanced with provided examples. I think this is what you both want. But this is "help" oriented towards developers since you wont need that for the end user if you ship your commercial card playing game ;) This kind of "browsable developer help" is what you both seem to understand when you speak about the "Help system". To me this is nothing more than documentation on how the system is implemented. It is a reference of the system - we already have it in the image with class comments, method comments. I call this "API-Help" or "System reference" (see the class SystemReference) and is only a subpart of a help system that I want. Anyway - the general help system is able to work with it too since you can build books from any source. Dont be afraid: the developer typically uses the Smalltalk browser to define and look at the comments as before. There is no dependency on the help system here. But if the help system is loaded he is able to get a better summary using the help tools: HelpBrowser openOn: SystemReference or HelpBrowser openOn: Integer or HelpBrowser openOn: (SystemReference hierarchyFor: Integer) since the info is just extracted by a builder or an object that converts the contents into help topics with structure. So the place where the help contents for B. is defined (class comments, method comments, ...) is not bound to the help system and the builder decides what is extracted and from what source. See HelpBuilder subclasses for example. So when Stef wrote: "The point is to avoid duplicating code between a class and its help or between a testcase and the help". I can just answer: there is no duplication, nothing has changed here, you document your classes and methods as before and you dont need a subclass of CustomHelp or the help system to do that!!! But if me may want to enhance the informations that can be extracted by the builder here or spice up the "API help" browsing experience we can define pragmas that can later be used by the help builders (like the current ClassAPIHelpBuilder). If we want to drive this forward then we should discuss about having not only Pragmas for methods but also for classes. For instance in Java I'm able to add annotations to a class comment: @author: Tut Ench Amun @version: 1333-1324 BC that is later used by the JavaDoc system. AFAIK Pragmas in Pharo are currently limited to method level. If we would have this HelpBrowser openOn: Integer could show the author, version, whatever you like. You just have to define what you need and enhance the builder to extract it. To summarize: - the more general help system (A) is able to display contents from any source including help contents extracted from usual Smalltalk classes/methods (B) - there is no dependency between the documented classes/methods and the help system itself in any case - if we want to enhance this part we can define pragmas that can be used by a help builder to provider a better "system reference" experience Long story, but in general: please dont see the help system as a better "documentation" browser to document your classes. It's more. Michael wrote: >I just wonder if the definition of the help (the raw content) should also >be bound up with the definition of the overall structure? For A. if you write a full book as described in A. I would answer YES since it helps to see the books structure also in the class hierarchy browser For B. the structure is already separated since it is defined by the builder depending on your needs, compare: HelpBrowser openOn: (SystemReference forClass: Integer) HelpBrowser openOn: (SystemReference hierarchyFor: Integer) HelpBrowser openOn: (SystemReference hierarchyWithMethodsFor: Integer) >So that's why i think of it in terms of annotations. >Having the help be in its own help class, on the class side, without >any real behavior is a design smell to me. You didnt understand why I provide "CustomHelp". Having a subclass is only necessary when you map a book structure to a class hierarchy. It is intended that these books typically have no behavior since they are just used to store book contents. You can see it as a vehicle to bring text content into the image (similar to WAFileLibrary subclasses in Seaside to bring in resources like images, stylesheets). This has nothing to do with a design smell. You seem to be too much focused on API like help. And I always said that we can use annotations for B. here to keep it independent from the help system. If you want to reference documentation for your classes you dont need to use CustomHelp! If you read you will see that I said it is harder to describe the structure of a full book using pragmas when you want A. I would prefer the CustomHelp subclassing here since you find your books structure also in the class hierarchy. Stef wrote: >Yes but it can also be out of sync. This is more related to us all beeing lazy writing and maintaining docu ;) >I like literate programming just because the comments are in the code. You can continue to have that, I see no conflict here >Writing separate and maintaining doc is costly and I was thinking that >I wanted to reuse as much as possible as the existing information (class >comments, unit tests...). Yes, but there is little value in this since you can already access this using the usual Smalltalk tools. Its just a simple way to fill the help browser with contents, but we have it now. ;) Convinced for inclusion? Bye T. -- GRATIS fÃ¼r alle GMX-Mitglieder: Die maxdome Movie-FLAT! Jetzt freischalten unter http://portal.gmx.net/de/go/maxdome01

1 0

[Pharo-project] [Preview] Help System with new model
by Torsten Bergmann April 9, 2010

April 9, 2010

Michael Roberts wrote: > I think it should be possible for help to appear in the global help > browser without having to define a subclass of CustomHelp. This would > make the entry barrier really low for people to be able to document > parts of the system using this help system. Yes, I had a pragma solution using <helpTopic: 'Title goes here'> in one of the first versions - but it gets complicated if you want to define structured help contents with nested books and with pages in a specific order. The implementation is easy ... I mean with pragmas it would be complicated for the "help writer" to define these nested book structures and to remember the correct pragmas. And you have to solve the following scenario: you may have a class A defining a help book using pragmas and a class B defining another help book which should be displayed under the book defined by A. Since you do not rely on a "help class hierarchy" and are decoupled using pragmas both classes can live under Object. Good. Everything works fine when A and B are in the image. But you can get a problem when class A is not loaded - since the parent book of B is not there. I would not see a problem in having the help hierarchy clearly defined as a subclass hierarchy of a predefined "CustomHelp" class. Providing CustomHelp subclasses is easy to understand for the "content writer" and system help is not spread around in the image. You can see the structure in the class hierarchy browser and in the help browser. Thats wonderfull. For all these reasons I would prefer a CustomHelp for the standard help. (Independent from that the help browser itself is still capable to use anything with an #asHelpTopic interface). Providing the CustomHelp class is the same as with TestCase - you know that when you subclass you will later find it in the UI tool (TestRunner/help browser) and you can peek in other TestCase subclasses to see what is possible. And yes, that means that minimal help support (in minimum 2 classes HelpTopic and CustomHelp) is already in the image similar to SUnit. Dont know if the name CustomHelp is good (maybe HelpBook/SystemHelpBook?). Stef wrote: >do you have a new version that we could integrate to 1.1? see http://www.squeaksource.com/HelpSystem The baseline works fine in 1.1.#11309, but I would remove a few methods after integration (for Squeak and Pharo 1.0 compatibility, see #initWindowInSqueak #isSqueak #isPharo11 and senders). That's easy. However, it is already working in 1.1. and integration would just mean to load it: - the minimum support is in the Core package HelpSystem-Core-tbn.35 - the according tests (HelpSystem-Tests-tbn.9) are green in 1.1 too. Thats anything you need to have help support, the browser ("Tools"->"Help Browser") and the full API help stuff (just run "HelpBrowser openOn: SystemReference" or any other sample I provided). One thing: if you just have the "HelpSystem-Core" loaded the standard help browser for the system would only display the help for the help system itself. If you want to additionally provide infos/content on the Pharo project you need other packages like "Pharo-Project-Help". We can enhance "Pharo-Project-Help" and use it either for the dev image or core image. We can also keep "Pharo-Project-Help" minimal for the pharo core image with basic infos on the project and create a "Pharo-Project-Dev-Help" or others for the pharo developer image. I see three options: (1) integrate it in 1.1. core as you wrote (2) no help in pharo core to keep it minimal and have the help system as one of the first packages loaded in the pharo dev image (3) wait until we can enhance it with Scamper code, markup, ... (which takes more time than I have now) Another option (4) which I dislike: If you find the help-core package (which currently includes the UI browser) to much for a pharo-core image we can also split the package so that: - the core model is in the image (only classes HelpTopic and CustomHelp are really necessary) so one can define and write help content - the help browser and builders are loadable/exchangable afterwards I would keep them together since you need minimal support to visually control what you wrote and why have help contents in the image when you cant browse it ;) If you would follow that path I can argue that we have to keep the settings framework but provide the settings browser only in the dev-image. Same issue. Thats why I would either follow (1) or (2), but that is just MHO. Bye T. -- GRATIS fÃ¼r alle GMX-Mitglieder: Die maxdome Movie-FLAT! Jetzt freischalten unter http://portal.gmx.net/de/go/maxdome01

3 3

[Pharo-project] Linux vm: assertions and strategy
by Schwab,Wilhelm K April 8, 2010

April 8, 2010

Laurent, all, If I may jump to conclusions a little, the 3.11-3 vm's changes break the loading of libraries, certainly to the point of being arkane. Beyond that, LinuxODBC is transparent to it, so I am left wondering how it is suppose to work at all?? The RC3 one-click image and vm produce very nice fonts on my Ubuntu laptop, but it retains the FFI/double defect, and from the naming of the plugins, must be based on a pre 3.11-3 vm. Is the plan to use 3.11-3 for the one-click? If so, we need to do something about external libraries; at a minimum, LinuxODBC should work, whether by fully understanding how to use environment variables, or (as I suspect) getting the dlopen() call to work again. Another option is to identify the FPU related fixes and apply them to the current one-click vm. Any preferences? Any success with 3.11-3 and LinuxODBC? I'd be happy to be wrong about this. Bill

2 6

Re: [Pharo-project] Phari collaborActive book
by Tudor Girba April 8, 2010

April 8, 2010

Hi Geert, I created the user. Will send the password separately. Cheers, Doru On 8 Apr 2010, at 09:33, Geert Claes wrote: > Hi Doru, > > I would like to put my hand up to help a bit with the collaborActive > book. Not sure yet how much I can contribute but I have good > intentions, do I ask you for login details? :) > > Cheers, > > Geert -- www.tudorgirba.com "Every thing has its own flow."

2 1

April 8, 2010

Hi Pharoers We put in place (big big thanks doru, anton, lukas and gemstone for that) a pier book http://book.pharo-project.org/ The idea is that we can collect/edit.... information about pharo on this collaborative book. Now we would like that **you**/**us** control the quality of the contents. The book will be open: **you**= anybody from the pharo community can be editor (you should ask probably doru because I have to really make sure that I do not mess it up) but not everybody can change text, everybody will be able to add notes following the http://book.seaside.st principle. As an editor you can register to the rss stream and includes the comments. Now the idea is that after a while the ready to get publish parts may be used for Pharo by example 2 since we can generate automatically latex from the web. Now I would love to have chapter on - how to compile my vm on Mac OSX - how to compile my vm on linux - how to compile my vm on Windows - Understanding Announcement - Settings - Building UI with Polymorph.... - Help - ... So if you want to be an editor you are more than welcome! Just ask Stef

9 28

April 8, 2010

Hi guys We have bug entries and now I have one question. Do we want to mix or not todo within them? I deleted all the old tasks from the inria web site. Now we could use it. Do you have experience managing todo (which may be orthogonal to issue and bug trcking)? Stef

8 15