[Pharo-project] Test Server
Hi! I've been told that Keith and maybe few others are working on a test server. In a 10 minutes seaside coding session I did a small application that displays a result of a SUnit test. Is there anything like this around? Cheers, Alexandre -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;.
On Oct 27, 2008, at 5:13 PM, Alexandre Bergel wrote:
Hi!
I've been told that Keith and maybe few others are working on a test server.
this is called Bob and use Sake now I do not know more than that. But would be good to invest to see and possibly help. The refactoring of matthew of installer is really nice.
In a 10 minutes seaside coding session I did a small application that displays a result of a SUnit test. Is there anything like this around?
Cheers, Alexandre
-- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;.
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Alexandre Bergel wrote:
Hi!
I've been told that Keith and maybe few others are working on a test server. In a 10 minutes seaside coding session I did a small application that displays a result of a SUnit test. Is there anything like this around?
Cheers, Alexandre
Yes... 'SUnit improved' has a TestReporter, since Jan/Feb 2007. It is designed to be invoked from the command line on any image, and to work headless if desired. It generates results to files, the idea being that the progress can be observed remotely via ftp or http. Failures and Errors are generated in a file form that can be viewed remotely, but selected failures/errors can be run by doiting the line of the output file. There is one file which has a byte written per test so that you can see progress remotely just by looking at the file size in a web /dir/ftp browser. There are also custom icons for apache to show pass/fail/errors etc though I am not sure where I have put them. === Please please please, you pharo guys, actually look at SUnit improved, before doing other things with SUnit, I put a lot of effort in to it, over the past 2 years. The latest suite defining and building architecture is really cool. It uses less code than the old system and is better factored to achieve more. For those of us that are interested in developing packages that everyone can use, we need a coherant story accross the board so that we can write tests and flag which images we expect them to work in. 'SUnit improved' has this in TestCaseVersioned. ClassClonerTestResource is another case in point, I wrote it so that we can test the class side of DateAndTime, etc, but still the wheel gets reinvented again and again. A really useful addition which I have been planning for ages, but you would be welcome to have a go at, would be excellent if SSpec specs could be run in the same ui. cheers Keith p.s. Bob the Builder update to follow
Dear All, Bob is slowly taking shape, I am perhaps a day away from alpha, but I don't have much spare time this week, so I expect that this will not be available until next week. If you would like to contribute, I am hanging around in squeak irc most (uk) evenings. some docs follow, cheers Keith .... ======= * Bob the Builder* */Your Build is defined as a Class/* To add your own 'Build' process to Bob, create a subclass of BobBuild, with your process defined as a SakeTask returned from the class method #build. When Bob starts his work in BobBuild-c-#taskMain, he collects all of these tasks, and sorts them according to their dependencies. /*Your build can depend upon other builds*/ So your build can specify which other builds it depends upon, for example: BobBuildMyWebServer-#build ^ self define: [ :task | task dependsOn: { BobBuildSeasideOneClick build }. .... ... ... ]. /*Repositories are watched and builds run when changes are spotted*/ Sake-Scheduler is used to run periodic tasks. These tasks can do many useful things, as well as running #taskMain. Scheduled tasks are used to: 1. Periodically collect data that either Bob/users may need in their work (e.g. BobPeriodicallyReadMantis ) 2. Periodically update the Bob server image with the latest code from contributors (e.g. BobPeriodicallyWatchRepositories ) You can add your own periodic tasks by subclassing ScheduledTask, however some tasks are designed to be extended, by adding (extension) instance methods. e.g. BobPeriodicallyWatchRepositories-#taskWatchPackages (watches the squeaksource Packages repository for updates) if you wish to invoke bob following an update to your repository this example shows how. BobPeriodicallyWatchRepositories >> #taskWatchTasks ^ SakeTask checkUrl: 'http://www.squeaksource.com/Tasks/feed.rss' onChanged: [ Installer sake install: 'Tasks'. BobBuild taskMain run. ]. /*Configuration*/ Your build process 'task' #build is defined on the class side of your Build class. If your task has a number of "const" parameters that configure the task, these should also be defined on the class side. For example, the location of a working directory would be returned from #configWorkingDir /*Metadata - build numbers etc*/ The metadata for each Build task is specified on the instance side of the class, e.g. #build0001, #build0002. Your task definition in #build automatically uses the most recent metadata method. example metadata: MyBuild>>#build0001 info name: 'Squeak3.11-basic'. info moniker: '311'. info codeName: 'Kipper'. info by: 'kph'. info image: 'bzr://bzr.warwick.st/squeak/3.10'. info candidate: false. info release: false. info overwrite: true. info comment: 'testing'. /*Only build if necessary*/ It is very important that your build only runs when absolutely necessary. If your build is manually invoked by an increment in build number (i.e. info overwrite: false) then the existence of the output directory with that build number is enough to determine that the build has already been run. If your build is automatically invoked by updates to particular repositories, then the versions of those packages need to be recorded in the metadata. The metadata is written to a file in the output directory. Subsequent builds should compare the current metadata with that in the file to see whether the build needs to be re-run. If your build is a derivative of a released product, then you only need to build if the dependant has #info release = true.
Ok we have to digest that. Jannik started to expand what I wrote on Installer and the next step is Sake than Bob I want to have all that three package well reviewed, documented, advertized. Now personally I should learn Sake. On Oct 27, 2008, at 10:50 PM, Keith Hodges wrote:
Dear All,
Bob is slowly taking shape, I am perhaps a day away from alpha, but I don't have much spare time this week, so I expect that this will not be available until next week.
If you would like to contribute, I am hanging around in squeak irc most (uk) evenings.
some docs follow,
cheers
Keith
....
======= * Bob the Builder*
*/Your Build is defined as a Class/*
To add your own 'Build' process to Bob, create a subclass of BobBuild, with your process defined as a SakeTask returned from the class method #build. When Bob starts his work in BobBuild-c-#taskMain, he collects all of these tasks, and sorts them according to their dependencies.
/*Your build can depend upon other builds*/
So your build can specify which other builds it depends upon, for example:
BobBuildMyWebServer-#build
^ self define: [ :task | task dependsOn: { BobBuildSeasideOneClick build }.
.... ... ... ].
/*Repositories are watched and builds run when changes are spotted*/
Sake-Scheduler is used to run periodic tasks. These tasks can do many useful things, as well as running #taskMain. Scheduled tasks are used to:
1. Periodically collect data that either Bob/users may need in their work (e.g. BobPeriodicallyReadMantis ) 2. Periodically update the Bob server image with the latest code from contributors (e.g. BobPeriodicallyWatchRepositories )
You can add your own periodic tasks by subclassing ScheduledTask, however some tasks are designed to be extended, by adding (extension) instance methods.
e.g. BobPeriodicallyWatchRepositories-#taskWatchPackages (watches the squeaksource Packages repository for updates) if you wish to invoke bob following an update to your repository this example shows how.
BobPeriodicallyWatchRepositories >> #taskWatchTasks
^ SakeTask checkUrl: 'http://www.squeaksource.com/Tasks/feed.rss' onChanged: [ Installer sake install: 'Tasks'. BobBuild taskMain run. ].
/*Configuration*/
Your build process 'task' #build is defined on the class side of your Build class. If your task has a number of "const" parameters that configure the task, these should also be defined on the class side. For example, the location of a working directory would be returned from #configWorkingDir
/*Metadata - build numbers etc*/
The metadata for each Build task is specified on the instance side of the class, e.g. #build0001, #build0002. Your task definition in #build automatically uses the most recent metadata method.
example metadata:
MyBuild>>#build0001
info name: 'Squeak3.11-basic'. info moniker: '311'. info codeName: 'Kipper'. info by: 'kph'. info image: 'bzr://bzr.warwick.st/squeak/3.10'. info candidate: false. info release: false. info overwrite: true. info comment: 'testing'.
/*Only build if necessary*/
It is very important that your build only runs when absolutely necessary. If your build is manually invoked by an increment in build number (i.e. info overwrite: false) then the existence of the output directory with that build number is enough to determine that the build has already been run.
If your build is automatically invoked by updates to particular repositories, then the versions of those packages need to be recorded in the metadata. The metadata is written to a file in the output directory. Subsequent builds should compare the current metadata with that in the file to see whether the build needs to be re-run.
If your build is a derivative of a released product, then you only need to build if the dependant has #info release = true.
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
participants (3)
-
Alexandre Bergel -
Keith Hodges -
Stéphane Ducasse