- Pharo-dev - lists.pharo.org

Re: [Pharo-project] Delay bug

by Gary Chambers

Issue http://code.google.com/p/pharo/issues/detail?id=3986 Regards, Gary ----- Original Message ----- From: "Adrian Lienhard" <adi(a)netstyle.ch> To: <Pharo-project(a)lists.gforge.inria.fr> Sent: Wednesday, April 06, 2011 5:41 PM Subject: Re: [Pharo-project] Delay bug We managed to resolve our initial problem with the process scheduling [1]. All Cmsbox customers have been migrated to Pharo and it works very reliably now. Gary, maybe you could open a ticket already now and post what you know at this time to get early feedback. [1] http://code.google.com/p/pharo/issues/detail?id=3498 Cheers Adrian On Apr 6, 2011, at 18:24 , StÃ©phane Ducasse wrote: > thanks gary. > Yes adrian got a problem on his server after two weeks of load. I do not > know how we can improve them. > > Stef > > On Apr 6, 2011, at 6:02 PM, Gary Chambers wrote: > >> Well, we have been having problems with applications appearing to hang >> every 6 days or so (sound familiar?). >> >> I have identified a bug in the Delay implementation, rather obscure, but >> highlighted by our applications that make HEAVY use of processes/(tight, >> a few tens of milliseconds)delays/semaphores etc. >> Naturally due to an edge case on millisecond clock rollover... >> >> Affects all Squeak/Pharo versions, as far as I can tell. (It is separate >> from the OSProcess issues I have posted about a few weeks ago). >> >> I've developed a potential fix, though I'll have to let you all know >> whether (hopefully) successful in a week or two... fingers crossed! >> >> I hope it is fixable, last hurdle for business applications, from our >> point of view. >> >> I'll post details when fix confirmed or failed then. >> >> Regards, Gary > >

April 11, 2011

[Pharo-project] Issue 3986 in pharo: Bug with Delay during clock rollover

by pharo＠googlecode.com

Status: FixProposed Owner: gazzagu...(a)googlemail.com CC: adrian.l...(a)gmail.com, marcus.d...(a)gmail.com Labels: Type-ReportDefect New issue 3986 by gazzagu...(a)googlemail.com: Bug with Delay during clock rollover http://code.google.com/p/pharo/issues/detail?id=3986 I have identified a bug with Delays that are scheduled close to the millisecond clock rollover that can occur when other delays expire around the same time. The issue appears to be that when scheduling a delay, the resumption time gets set and, before the AccessProtect critical block can be entered, the TimingSemaphore gets signalled, causing the timer process to handle an expired delay. If the clock rolls over before the resumption of the process that was scheduling the original delay, that delay gets scheduled with the resumption time that was determined prior to the clock rollover, hence a delay of around 6 days. A fix is proposed in the attached change set, tested empirically during a rollover with the result of no "rogue" delays - not conclusive but in our production code it appears the problem is resolved. Regards, Gary

April 11, 2011

[Pharo-project] Brussels Sprint: gathering number of attendes

by Johan Brichau

Hi everyone. To make sure we have enough food and drinks for everyone. Please make sure to let us know you are coming. >From previous emails, I have understood the following people are attending: - Steph - Marcus - Sven - Mariano - Craig Latta - Thomas Holzer - Andy - Johan Someone on here who is not attending? Someone not attending who is listed above? The parking access code will follow shortly. ps: I will bring some Faro beer this time then ;-) Johan

April 11, 2011

Re: [Pharo-project] Issue 3982 in pharo: Error: can''t find EOCD position

by pharo＠googlecode.com

Comment #3 on issue 3982 by marianopeck: Error: can''t find EOCD position http://code.google.com/p/pharo/issues/detail?id=3982 Yes, that would be nice. We have already seen this problem FAR so many times.

April 11, 2011

Re: [Pharo-project] Issue 3982 in pharo: Error: can''t find EOCD position

by pharo＠googlecode.com

Comment #2 on issue 3982 by nicolas....(a)gmail.com: Error: can''t find EOCD position http://code.google.com/p/pharo/issues/detail?id=3982 Couldn't the exception be handled with an automated or manual action via a more appropriated dialog like: Error can't find EOCD blah blah The package appear to be corrupted, maybe due to a communication error What would you like to do ? - flush the cache and retry - abort operation

April 11, 2011

[Pharo-project] Issue 3985 in pharo: Create temporary in debugger does not work for new keyword method

by pharo＠googlecode.com

Status: Accepted Owner: laurent....(a)gmail.com New issue 3985 by laurent....(a)gmail.com: Create temporary in debugger does not work for new keyword method http://code.google.com/p/pharo/issues/detail?id=3985 Reproductible in Pharo 1.2 (Full) - looks ok in PharoCore 1.3 Create a class in the browser like Foo >> bar self doSomethingWith: 2. save and evaluate Foo new bar. Debugger pops up. Click create method doSomethingWith: Foo>>doSomethingWith: aSmallInteger a := aSmallInteger ^ a. Accept. Dialog popus up, select declare temp => does not work.

April 11, 2011

[Pharo-project] GitFS and Pharogenesis

by Max Leske

Hi folks I promised a while ago that I would keep you up to date with GitFS and Pharogenesis. Well, now that I've handed in my thesis you can read it if you like: http://scg.unibe.ch/archive/projects/Lesk11a.pdf I also wrote a short How-To on Filesystem (alas, the "old" version) and GitFS: http://scg.unibe.ch/archive/projects/Lesk11b.pdf You can find a summary of important information (loading the code etc.) on this page: http://scg.unibe.ch/research/Pharogenesis Although I'm sort of "releasing" the code for both now, please note that Pharo has progressed a bit too fast lately for me to keep up. GitFS was built on the original implementation of Filesystem and I will have to put some work into updating the implementation. I'm even thinking of starting over from scratch. Pharogenesis also depends on the old version of Filesystem. The Metacello configurations load the old Filesystem version into the image, just so you know. If you don't know what this is all about, here's the abstract of my thesis: In the Smalltalk dialects Squeak and Pharo it is difficult to view changes to source code that occurred accross major releases. This poses a problem for development as well as for code analysis: Developers often need to be able to revert code to a previous version and for code analysis there should ideally be a single source of information. We solve the problem by building a Git repository with the complete source code history of Pharo. To build the repository we propose two tools: GitFS and Pharogenesis. GitFS is an implementation of Git written entirely in Smalltalk. With GitFS we can create, read and manipulate Git repositories. Pharogenesis uses GitFS to build a Git repository from the source code that Pharogenesis extracts from the source files. We also show how Git can be used to provide a single point of access for all source code from Squeak 1 through Squeak 4 and Pharo 1 using GitFS. The source code database we built with Pharogenesis is available on github (http://www.github. com/pharogenesis/pharogenesis.git). If you have any questions, feel free to ask. Cheers, Max

April 11, 2011

[Pharo-project] Issue 3984 in pharo: Announcements announced when a test is runned

by pharo＠googlecode.com

Status: FixProposed Owner: Benjamin...(a)gmail.com Labels: Milestone-1.3 Nautilus New issue 3984 by Benjamin...(a)gmail.com: Announcements announced when a test is runned http://code.google.com/p/pharo/issues/detail?id=3984 More precisely, when the history of the test case is updated Attachments: AnnouncementsInTestCase.1.cs 2.8 KB

April 11, 2011

[Pharo-project] Issue 3983 in pharo: TestCase assertions are in accessing protocol

by pharo＠googlecode.com

Status: Accepted Owner: laurent....(a)gmail.com Labels: Type-RequestForEnhancement Difficulty-Easy New issue 3983 by laurent....(a)gmail.com: TestCase assertions are in accessing protocol http://code.google.com/p/pharo/issues/detail?id=3983 Pharo 1.2 & 1.3 Put them in asserting protocol like TestAsserter

April 11, 2011

Re: [Pharo-project] Issue 3977 in pharo: Undeclared vars in MorphTreeMorph

by pharo＠googlecode.com

Updates: Status: FixProposed Comment #3 on issue 3977 by Benjamin...(a)gmail.com: Undeclared vars in MorphTreeMorph http://code.google.com/p/pharo/issues/detail?id=3977 It should fix the issue :) Attachments: MorphTreeMorph.BenjaminVanRyseghem.1.cs 5.5 KB

April 11, 2011