- Pharo-dev - lists.pharo.org

Re: [Pharo-project] NonInteractiveTranscript holds reference to closed files

by Sven Van Caekenberghe

Hi Guillermomo, On 27 Jun 2012, at 14:50, Guillermo Polito wrote: > Select the three statements at once and evaluate them: > > NonInteractiveTranscript stdout install. > Smalltalk snapshot: true andQuit: true. > Transcript show: 'hello'. > > Then reopen your image... Bump! stdout is closed. > > Of course, It got closed when the image quitted, but the NonInteractiveTranscript kept the reference to an invalid fileStream... > > What should be a good solution for this? > - catch the error on the transcript and if closed reopen the file? > - subscribe to the shutdown list to void the reference? > - adding a SystemShutdown announcement and subscribe to it? > > The third one seems... cleaner. But I'm not sure. > > Guille Yes, the stream inside NonInteractiveTranscript should be set to nil whenever an image save closes stdout, so either option 2 or 3. Mind you that the current close/open/reopen logic is not good enough: stream is never set to nil . I think that open and reopen should be removed, and that close should become close self critical: [ stream ifNotNil: [ self isStdout ifTrue: [ stream flush ] ifFalse: [ stream close ]. stream := nil ] ] And then you should add some shutdown hook. Sven

June 27, 2012

Re: [Pharo-project] Encodings, Charsets, converters... streams?

by Henrik Sperre Johansen

On 25.06.2012 18:06, Denis Kudriashov wrote: > Hello. > > What is responsibility of environments (russian, japanese) classes. > Why its needed? Why converter classes not enough? > > For example, I can write russian text at workspace without setting any > specific russian environment. I only need specific fonts for this. > Novadays, they can be used to read number/date printing policies, decide which translation package to use, etc. Basically what you'd expect from an OS locale, but more limited in capabilities. The part linked to charsets and encodings seem to me an anachronism from when the OS's used different encodings based on the selected environment though. (so files would be read and written automatically with the correct (non-unicode) encoding, and potentially displayed using a different bitmap font through selection using leadingChar) > 2012/6/25 Guillermo Polito <guillermopolito(a)gmail.com > <mailto:guillermopolito@gmail.com>> > > Hi guys, > > I'm trying to understand the Multilingual packages to get things a > bit more reorganized. This way we can start to have a simpler and > understandable kernel. So, > > - We have all environments(greek, chinese, japanese, korean, > russian...) in the same package. Each environment has it's > associated charset, and text converter, which are in different > packages. Also, each text converter has a table of conversions > between unicode and the encoding. But all converters are together, > and all charsets are together. > > What about putting related things together? I mean, japanese > converter with japanese environment with japanese charset in one > package. Same with russian, korean, chinese... Does it make > sense? This way we should be able to unload one of them if we do > not really need it. > > - MultiByteFileStream and MultiByteBinaryOrTextStream are inside > Multilingual-TextConversion? Even, they do the same on top of a > file or a memory stream of bytes (or they say so in the comment > :^D )... > > What about merging them? using a decorator? They have a lot in > common... :/ Put them in a different package? > They are not neatly composed. Trying to merge them so the methods work with a file and bytearray/string backend equivalently would probably only make things even worse. Using a decorator would be hard, since the MultiByteFileStream method implementations use alot of its inherited behaviour. As for putting them in a different package, sure you can peddle crap around all you want, but that doesn't make it stink any less. > > > - Just curious. ImmPlugin is working or shipped with vms? Not for > macos, because there is no macos implementation. Then, the class > comment says: > > /"On Windows, the plugin is not provided in the shipped VM's, and > its current whereabouts are uncertain. > > On Unix, the only implemented behaviour is setting the position > when over-the-spot precomposition of characters is the current mode. > > In the VM, the mode is chosen to the first available valid mode > returned by the X Server, so whether this is actually relevant at > all depends on the X Server./" > > So, does someone use it and see value on keeping it? > Sure, there is value in neatly integrating OS input methods for composed characters. Whether that belongs in a plugin of it's own, and whether what the plugin currently offers justifies its existence, is a different question, which I guess the removal was the answer to. :) To find any potential actual users, I guess you'd have to find a pharo app localized to the eastern hemisphere and have it ask their linux users if they appreciate it/if it works. That, or a developer from the same area who codes in his native language (which will be hard I think, given arbitrary tool constraints wrt message names etc.) Cheers, Henry

June 27, 2012

Re: [Pharo-project] Encodings, Charsets, converters... streams?

by Guillermo Polito

Hi Denis, On Mon, Jun 25, 2012 at 6:06 PM, Denis Kudriashov <dionisiydk(a)gmail.com>wrote: > Hello. > > What is responsibility of environments (russian, japanese) classes. > Why its needed? Why converter classes not enough? > > For example, I can write russian text at workspace without setting any > specific russian environment. I only need specific fonts for this. > > Without being an expert, what I can see is that LanguageEnvironments are a kind of facade/factory of Charset, Encoding, Locales, Fonts... So, Greek environment should provide you all the correct configuration to do greek stuff... > > 2012/6/25 Guillermo Polito <guillermopolito(a)gmail.com> > >> Hi guys, >> >> I'm trying to understand the Multilingual packages to get things a bit >> more reorganized. This way we can start to have a simpler and >> understandable kernel. So, >> >> - We have all environments(greek, chinese, japanese, korean, russian...) >> in the same package. Each environment has it's associated charset, and >> text converter, which are in different packages. Also, each text converter >> has a table of conversions between unicode and the encoding. But all >> converters are together, and all charsets are together. >> >> What about putting related things together? I mean, japanese converter >> with japanese environment with japanese charset in one package. Same with >> russian, korean, chinese... Does it make sense? This way we should be >> able to unload one of them if we do not really need it. >> >> - MultiByteFileStream and MultiByteBinaryOrTextStream are inside >> Multilingual-TextConversion? Even, they do the same on top of a file or a >> memory stream of bytes (or they say so in the comment :^D )... >> >> What about merging them? using a decorator? They have a lot in common... >> :/ Put them in a different package? >> >> - Just curious. ImmPlugin is working or shipped with vms? Not for macos, >> because there is no macos implementation. Then, the class comment says: >> >> *"On Windows, the plugin is not provided in the shipped VM's, and its >> current whereabouts are uncertain. >> >> On Unix, the only implemented behaviour is setting the position when >> over-the-spot precomposition of characters is the current mode. >> >> In the VM, the mode is chosen to the first available valid mode returned >> by the X Server, so whether this is actually relevant at all depends on the >> X Server.*" >> >> So, does someone use it and see value on keeping it? >> >> >> Guille >> > >

June 27, 2012

[Pharo-project] NonInteractiveTranscript holds reference to closed files

by Guillermo Polito

Select the three statements at once and evaluate them: NonInteractiveTranscript stdout install. Smalltalk snapshot: true andQuit: true. Transcript show: 'hello'. Then reopen your image... Bump! stdout is closed. Of course, It got closed when the image quitted, but the NonInteractiveTranscript kept the reference to an invalid fileStream... What should be a good solution for this? - catch the error on the transcript and if closed reopen the file? - subscribe to the shutdown list to void the reference? - adding a SystemShutdown announcement and subscribe to it? The third one seems... cleaner. But I'm not sure. Guille

June 27, 2012

Re: [Pharo-project] Zinc intermittently returning nil

by Sven Van Caekenberghe

This will be hard to debug. I known that sometimes there are odd failures using Zdc, but I can't put my finger on them. Any help would be greatly appreciated. I just added an extra test to the Zinc-Zodiac functional test package: --- Name: Zinc-Zodiac-SvenVanCaekenberghe.13 Author: SvenVanCaekenberghe Time: 27 June 2012, 1:57:30.506 pm UUID: dcc9a44f-f484-434f-bf7d-c44a1cb9c18e Ancestors: Zinc-Zodiac-SvenVanCaekenberghe.12 added an HTTPS transfer stress test reading odd sized files from Amazon S3 in size between 2 KB and 100 KB checking the full contents to make sure the whole transfer was OK --- Like all other tests, this one works for me, all the time. Here is the code: testTransfers (Integer primesUpTo: 100) do: [ :each | | size client | size := 1024 * each + each. (client := ZnClient new) https; host: 's3-eu-west-1.amazonaws.com'; addPath: 'public-stfx-eu'; addPath: ('test-', size asString, '.txt'); get; close. self assert: client isSuccess. self assert: client response contentType equals: ZnMimeType textPlain. self assert: client response contentLength equals: size. self assert: client contents equals: (self class generateTestData: size) ] The test is 'special' because it actually tests the full contents of the response. It also uses odd sizes to try to stress test secure stream internal buffering and TLS framing issues, as well as interactions with the SSL pluging. But your URL does indeed fail, sometimes. I need a reliable failure so that I can debug it manually ;-) Sven On 27 Jun 2012, at 09:00, Sven Van Caekenberghe wrote: > Yes, that is *very* strange (latest 2.0, latest VM). > > Thanks a lot for reporting an intermittent bug, Sean ;-) > > On 27 Jun 2012, at 05:33, Sean P. DeNigris wrote: > >> The following all intermittently return nil. When I do the same thing in >> Safari, I always seem to get the correct answer (plain text). >> >> (ZnEasy get: >> 'https://ci.lille.inria.fr/pharo/job/Pharo-2.0/lastSuccessfulBuild/api/xml?x…') >> contents. >> >> (ZnClient new get: >> 'https://ci.lille.inria.fr/pharo/job/Pharo-2.0/lastSuccessfulBuild/api/xml?x…') >> contents. >> >> (ZnUrl fromString: >> 'https://ci.lille.inria.fr/pharo/job/Pharo-2.0/lastSuccessfulBuild/api/xml?x…') >> retrieveContents. >> >> Even this returns nil intermittently (maybe less frequently, not sure): >> client := ZnClient new. >> client >> systemPolicy; >> url: >> 'https://ci.lille.inria.fr/pharo/job/Pharo-2.0/lastSuccessfulBuild/api/xml?x…'; >> accept: ZnMimeType textPlain. >> >> client get contents. >> >> Sean >> >> -- >> View this message in context: http://forum.world.st/Zinc-intermittently-returning-nil-tp4636820.html >> Sent from the Pharo Smalltalk mailing list archive at Nabble.com. >> > >

June 27, 2012

Re: [Pharo-project] Google Code labels/templates

by Guillermo Polito

Yes, I think also that the Type (bug, enh, feature, cleanup) of the issue must be orthogonal to the tool/app/subsystem in question :) On Tue, Jun 26, 2012 at 8:48 PM, Sean P. DeNigris <sean(a)clipperadams.com>wrote: > Someone had suggested to remove the "Pharo 1.4 Enhancement" template, since > we're not accepting any. That makes sense. > > Also, the Nautilus template labels the issue "Nautilus", but the label in > the list is "Type-Nautilus". Which brings me to my next question... > > The other Types are Bug, Enhancement, Cleanup, Feature [1]. Nautilus and > Jenkins [2] really don't fit into this framework. Maybe because there will > be many fewer issues, we don't need the level of granularity of [1] for > [2]. > But when we get close to a release, it'd be nice to know that there is 1 > Nautilus bug and 10 Enhancements pending (i.e. probably only one critical > thing) vs. 11 Type-Nautilus. > > So, now that I said all that, it seems the most powerful/flexible approach > would be to leave the "nautilus" tag, remove the leading "Type-" from > "Type-[Jenkins|Spec|Seed]", and then to make templates for > Jenkins/Spec/Seed > like the one for Nautilus. > > What do you think? > > Sean > > -- > View this message in context: > http://forum.world.st/Google-Code-labels-templates-tp4636767.html > Sent from the Pharo Smalltalk mailing list archive at Nabble.com. > >

June 27, 2012

[Pharo-project] is PackageDependencyTest used?

by Guillermo Polito

It's a testcase without tests :). Seems to have a template to validate package dependencies... but no subclasses either... And it's in SUnit-Core-Tests :/ For now I'll submit an issue to recategorize it.

June 27, 2012

[Pharo-project] pharo-2.0-tests Â» 32,mac - Build # 164 - Failure!

by jenkins-pharo.ci.inria.fr＠jenkins-server.lille.inria.fr

June 27, 2012

[Pharo-project] pharo-2.0-tests Â» 32,win - Build # 164 - Failure!

by jenkins-pharo.ci.inria.fr＠jenkins-server.lille.inria.fr

June 27, 2012

[Pharo-project] pharo-2.0-tests Â» 32,linux - Build # 164 - Failure!

by jenkins-pharo.ci.inria.fr＠jenkins-server.lille.inria.fr

June 27, 2012