[boost@home] revolutionary solution to create a testing farm

Hi, go ahead and call me crazy :) I couldn't sleep last night thinking about the problems that boost has with the testing process. Although the roots of my idea were different it turns out to be very similar to well known SETI@Home project. Therefore I call it Boost@Home :) Solution is simple in following steps: 1. create a vmware image with operating system and a single compiler installed. One vmware image per compiler to ease bugfix and upgrades in the vmware images. 2. let people download as many images (compilers) as they want. They will use it with freely available vmware-server to launch it. 3. After the virtual machine starts (with the compiler) a small program will be automatically executed to await instructions from boost.org. 4. The boost.org website will list the number of available virtual machines and which tests they are currently running. No interaction from the person who runs the virtual machine is required. Once it is launched it obeys instructions from the central boost.org server. Problems: 1. vmware is free to use, but supports only x86 architectures. Fortunately (to my knowledge) every compiler supported by boost has some version working in x86 environment. 2. some compilers require propertiary windows OS. Giving away a clean install of windows is prohibited. We can use ReactOS intead: http://www.reactos.org/ and install the compilers there. 3. It's possible that not every compiler has a version available for free. I'm not sure - is that true? If yes then it's in their interest to help us: - we ask the to provide a free copy of their compiler for the sole purpose of being used by Boost@Home project. - alternatively we drop support for their compiler. I have several machines lying around that could be used to help boost. If deploying a testing farm will become as easy, as downloading a vmware image and starting it - it's obvious that many people will join this. The capcity of boost testing farm can increase more than tenfold! -- Janek Kozicki |

on Thu Jun 07 2007, Janek Kozicki <janek_listy-AT-wp.pl> wrote:
2. some compilers require propertiary windows OS. Giving away a clean install of windows is prohibited. We can use ReactOS intead: http://www.reactos.org/ and install the compilers there.
Given my experiences trying to install recent versions of Visual Studio under WINE, I have my doubts about the feasibility of doing that: "Please bear in mind that ReactOS 0.3.1 is still in alpha stage, meaning it is not feature complete and is not recommended for everyday use." -- Dave Abrahams Boost Consulting http://www.boost-consulting.com

David Abrahams said: (by the date of Thu, 07 Jun 2007 10:38:56 -0400)
install of windows is prohibited. We can use ReactOS intead: http://www.reactos.org/
Given my experiences trying to install recent versions of Visual Studio under WINE, I have my doubts about the feasibility of doing that:
Sigh, unfortunate showstopper :/ Didn't microsoft announce sth. like "windows95 (or some other, doesn't matter) can be installed without a license" ? I have a vague memory, that this (very old, no longer supported) windows version can be now installed legally without a license? -- Janek Kozicki |

David Abrahams said: (by the date of Thu, 07 Jun 2007 10:38:56 -0400)
on Thu Jun 07 2007, Janek Kozicki <janek_listy-AT-wp.pl> wrote:
2. some compilers require propertiary windows OS. Giving away a clean install of windows is prohibited. We can use ReactOS intead:
Given my experiences trying to install recent versions of Visual Studio under WINE, I have my doubts about the feasibility of doing that
Hold on, I've just found a legal way to have a 4 month boost testing on windows vista :) "Official Way to Install and Use Windows Vista without Activation for Free for 120 Days" http://www.neowin.net/index.php?act=view&id=38289 http://www.howtogeek.com/howto/windows-vista/extend-windows-vista-trial-to-1... http://www.google.com/search?q=windows+vista+120+days "Windows Vista is the first Microsoft operating system to let you have a trial period. One of the questions that I get asked constantly is "Where do I get a copy of Vista?". This answer is less obvious. If you know somebody that has a copy of Vista, you can borrow the CD and use it to install in trial mode. As long as you aren't trying to use their serial number, there shouldn't be a problem with this." Does it mean that we can start the Boost@Home project? And support all compilers and platforms? 4 months is a lot, and during that time we could contact microsoft to grant a license explicitly for this project (and prohibit any other use of those vmware images). In general this idea is supporting microsoft OS, so they might like it... The key point here is to start the project and show how great increase in boost testing farm it provided. With good results it should be easier to get permission from them. -- Janek Kozicki |

I think the idea is good but my experience with running regressions/compilations on vmware is that it is SLOOOOOOOOOOOOOOOOOOOOOOOOOOW. If instead you have a per-platform "installer package" that does the same thing, only runs natively, that would be good. -----Original Message----- From: boost-bounces@lists.boost.org on behalf of Janek Kozicki Sent: Thu 6/7/2007 5:24 AM To: boost@lists.boost.org Subject: [boost] [boost@home] revolutionary solution to create a testing farm Hi, go ahead and call me crazy :) I couldn't sleep last night thinking about the problems that boost has with the testing process. Although the roots of my idea were different it turns out to be very similar to well known SETI@Home project. Therefore I call it Boost@Home :) Solution is simple in following steps: 1. create a vmware image with operating system and a single compiler installed. One vmware image per compiler to ease bugfix and upgrades in the vmware images.

Sohail Somani said: (by the date of Thu, 7 Jun 2007 08:42:59 -0700)
I think the idea is good but my experience with running regressions/compilations on vmware is that it is SLOOOOOOOOOOOOOOOOOOOOOOOOOOW.
I use vmware on daily basis. It is not slow. It has about 95% of computer power. And even if it was only 50% of computer power - we still get at least a tenfold increase in the number of computers available for testing. A good deal nonetheless. -- Janek Kozicki |

Hi Janek! On 07/06/07, Janek Kozicki <janek_listy@wp.pl> wrote:
Hi,
go ahead and call me crazy :) I couldn't sleep last night thinking about the problems that boost has with the testing process.
Although the roots of my idea were different it turns out to be very similar to well known SETI@Home project. Therefore I call it Boost@Home :)
<snip> FYI, BOINC (http://boinc.berkeley.edu/) is what SETI@Home has evolved into, which could be used to make a boostable testing farm. There's a lot less to develop/maintain that way, although it's less flexible than your idea. I've been looking at it in the last few days and it seems ideal, but it's probably beyond me to implement. Regards, Darren

Janek Kozicki wrote:
Hi,
go ahead and call me crazy :) I couldn't sleep last night thinking about the problems that boost has with the testing process.
Although the roots of my idea were different it turns out to be very similar to well known SETI@Home project. Therefore I call it Boost@Home :)
:-) I have made a similar proposal: http://lists.boost.org/Archives/boost/2007/05/120834.php Though, unfortunately, nobody seems to have picked up the topic. I don't think anything as radical as a VM is needed, really. Instead, the harness needs to be (incrementally) refined to 1) control as much of the development environment to which the build process is exposed 2) report all free parameters that influence the outcome. I think this is achievable, using a buildbot setup. Regards, Stefan -- ...ich hab' noch einen Koffer in Berlin...

-----Original Message----- From: boost-bounces@lists.boost.org [mailto:boost-bounces@lists.boost.org] On Behalf Of Stefan Seefeld Sent: Thursday, June 07, 2007 6:10 PM Subject: Re: [boost] [boost@home] revolutionary solution to create a testing farm
Janek Kozicki wrote:
Hi,
go ahead and call me crazy :) I couldn't sleep last night thinking about the problems that boost has with the testing process.
Although the roots of my idea were different it turns out to be very similar to well known SETI@Home project. Therefore I call it Boost@Home :)
:-)
I have made a similar proposal:
http://lists.boost.org/Archives/boost/2007/05/120834.php
Though, unfortunately, nobody seems to have picked up the topic. I don't think anything as radical as a VM is needed, really. Instead, the harness needs to be (incrementally) refined to
1) control as much of the development environment to which the build process is exposed
2) report all free parameters that influence the outcome.
I think this is achievable, using a buildbot setup.
Seconded. I wonder why you don't bring your own QMTest to the table. Wouldn't it also nicely fit into the boost picture? cheers, aa -- Andreas Ames | Programmer | Comergo GmbH | ames AT avaya DOT com Sitz der Gesellschaft: Stuttgart Registergericht: Amtsgericht Stuttgart - HRB 22107 Geschäftsführer: Andreas von Meyer zu Knonow, Udo Bühler, Thomas Kreikemeier

Ames, Andreas (Andreas) wrote:
I don't think anything as radical as a VM is needed, really. Instead, the harness needs to be (incrementally) refined to
1) control as much of the development environment to which the build process is exposed
2) report all free parameters that influence the outcome.
I think this is achievable, using a buildbot setup.
Seconded.
I wonder why you don't bring your own QMTest to the table. Wouldn't it also nicely fit into the boost picture?
Oh, I did. And yes, I think it fits. However, it isn't relevant to this particular discussion about achieving test coverage and automation. Both, Rene, as well as Vladimir are aware of QMTest. Thanks, Stefan -- ...ich hab' noch einen Koffer in Berlin...
participants (6)
-
Ames, Andreas (Andreas)
-
Darren Garvey
-
David Abrahams
-
Janek Kozicki
-
Sohail Somani
-
Stefan Seefeld