[Pharo-project] A newly discovered Metacello bug (yet to be characterized)
In the mail to Lukas below, I was able to identify a previously unreported bug in Metacello1.0-beta.28.3.3 when loading Seaside3.0.5 into Pharo1.2-12339. In the process of characterizing the bug, I discovered that the Seaside load was correct when using Metacello 1.0-beta.29 (an unreleased version). However, when I tested Seaside3.0.5 and Metacello1.0-beta.29 using Pharo1.2.2-12353, I discovered more odd behavior and Seaside3.0.5 hit errors during the load. There is something fishy going on and at this point in time the root cause is not obvious. The Metacello unit tests are passing in all three versions: Pharo1.2-12339, Pharo1.2.1-12345 and Pharo1.2.2-12353. I am now in the process of attempting to characterize the bugs. I will keep you posted on my progress. Dale ----- Original Message -----
Lukas,
I have looked specifically at Seaside3.0.5 and Pharo1.2-12339/PharoCore1.2-12336 and have found two things:
1. #stable version for ConfigurationOfOmniBrowser for 1.2 is 1.2.3 and it should be 1.2.4. 2. if OB 1.2.4 is already loaded then the version is downgraded to 1.2.3
Item #1 is a configuration issue and item #2 is a bug ...
I had not heard that "many projects could no longer be loaded into Pharo1.2", so I was (and still am) blissfully unaware of the problems that folks are having.
Seaside3.0.5 loads into Pharo1.2/PharoCore1.2 and passes all of the unit tests ... so there are no "load issues." The Seaside control panel works fine and I have not noticed any other OmniBrowser-related issues ... but then I don't live in the images day in and day out so there could be subtle issues.
I am currently trying to characterized the "downgrade version" bug and will push out a fix as soon as I have one.
Dale
----- Original Message -----
Here's the current state of 3.0.5: I saved the images from Jenkins and uploaded them to the web site. The change log has also been updated [1]. Everything but the Metacello configuration seems to work fine. If I load the Metacello configuration into a Pharo 1.2 it tries to load a version of OB that does not work in Pharo 1.2.
As far as I know most configurations have been updated to Pharo 1.3 already and many projects (I learned that with Glamour, Mondrian, Moose, Spy, ...) cannot be loaded into latest stable (Pharo 1.2) anymore. What surprised me more is that Metacello does not allow to load old versions (at least this is true for the configurations I had problems with). So I guess we have the same problem here.
I suggest that you load OB with this script: https://github.com/renggli/builder/blob/master/scripts/omnibrowser.st
Lukas
-- Lukas Renggli www.lukas-renggli.ch _______________________________________________ seaside-dev mailing list seaside-dev@lists.squeakfoundation.org http://lists.squeakfoundation.org/mailman/listinfo/seaside-dev
_______________________________________________ seaside-dev mailing list seaside-dev@lists.squeakfoundation.org http://lists.squeakfoundation.org/mailman/listinfo/seaside-dev
Okay the odd behavior between Pharo1.2-12339, Pharo1.2.1-12345 and Pharo1.2.2-12353 can be attributed to a known Metacello issue: http://code.google.com/p/metacello/issues/detail?id=112 So we are only facing two bugs: - the issue reported by Lukas, where version OmniBrowser 1.2.3 is loaded on top of version 1.2.4 (when nothing should have been loaded). Which appears to be fixed by Metacello 1.0-beta.29 (still need a test case to reproduce the bug) - the Seaside3.0.5 load failure Pharo1.2.2-12353 that doesn't show up in earlier versions. This error has to do with OcompletionPharoCompatibility and is likely due to a configuration error somewhere. I'll keep on digging... Dale ----- Original Message -----
In the mail to Lukas below, I was able to identify a previously unreported bug in Metacello1.0-beta.28.3.3 when loading Seaside3.0.5 into Pharo1.2-12339. In the process of characterizing the bug, I discovered that the Seaside load was correct when using Metacello 1.0-beta.29 (an unreleased version).
However, when I tested Seaside3.0.5 and Metacello1.0-beta.29 using Pharo1.2.2-12353, I discovered more odd behavior and Seaside3.0.5 hit errors during the load. There is something fishy going on and at this point in time the root cause is not obvious.
The Metacello unit tests are passing in all three versions: Pharo1.2-12339, Pharo1.2.1-12345 and Pharo1.2.2-12353.
I am now in the process of attempting to characterize the bugs.
I will keep you posted on my progress.
Dale
----- Original Message -----
Lukas,
I have looked specifically at Seaside3.0.5 and Pharo1.2-12339/PharoCore1.2-12336 and have found two things:
1. #stable version for ConfigurationOfOmniBrowser for 1.2 is 1.2.3 and it should be 1.2.4. 2. if OB 1.2.4 is already loaded then the version is downgraded to 1.2.3
Item #1 is a configuration issue and item #2 is a bug ...
I had not heard that "many projects could no longer be loaded into Pharo1.2", so I was (and still am) blissfully unaware of the problems that folks are having.
Seaside3.0.5 loads into Pharo1.2/PharoCore1.2 and passes all of the unit tests ... so there are no "load issues." The Seaside control panel works fine and I have not noticed any other OmniBrowser-related issues ... but then I don't live in the images day in and day out so there could be subtle issues.
I am currently trying to characterized the "downgrade version" bug and will push out a fix as soon as I have one.
Dale
----- Original Message -----
Here's the current state of 3.0.5: I saved the images from Jenkins and uploaded them to the web site. The change log has also been updated [1]. Everything but the Metacello configuration seems to work fine. If I load the Metacello configuration into a Pharo 1.2 it tries to load a version of OB that does not work in Pharo 1.2.
As far as I know most configurations have been updated to Pharo 1.3 already and many projects (I learned that with Glamour, Mondrian, Moose, Spy, ...) cannot be loaded into latest stable (Pharo 1.2) anymore. What surprised me more is that Metacello does not allow to load old versions (at least this is true for the configurations I had problems with). So I guess we have the same problem here.
I suggest that you load OB with this script: https://github.com/renggli/builder/blob/master/scripts/omnibrowser.st
Lukas
-- Lukas Renggli www.lukas-renggli.ch _______________________________________________ seaside-dev mailing list seaside-dev@lists.squeakfoundation.org http://lists.squeakfoundation.org/mailman/listinfo/seaside-dev
_______________________________________________ seaside-dev mailing list seaside-dev@lists.squeakfoundation.org http://lists.squeakfoundation.org/mailman/listinfo/seaside-dev
The Seaside3.0.5 load failure in Pharo1.2.2-12353 is due to the fact that Ocompletion-ul.102 is loaded in the image which is a package that is part of ConfigurationOfOCompletion 1.4 (presumably targetted at Pharo1.3 not Pharo1.2.2) ... In the earlier version of Pharo 1.2, Ocompletion-ul.92 is loaded and it is part of ConfigurationOfOCompletion 1.2.2 ... Not sure why Ocompletion-ul.102 was loaded as ConfigurationOfPharo points to the correct version of ConfigurationOfOCompletion, so I'll have to do more detective work... Dale ----- Original Message -----
Okay the odd behavior between Pharo1.2-12339, Pharo1.2.1-12345 and Pharo1.2.2-12353 can be attributed to a known Metacello issue:
http://code.google.com/p/metacello/issues/detail?id=112
So we are only facing two bugs:
- the issue reported by Lukas, where version OmniBrowser 1.2.3 is loaded on top of version 1.2.4 (when nothing should have been loaded). Which appears to be fixed by Metacello 1.0-beta.29 (still need a test case to reproduce the bug) - the Seaside3.0.5 load failure Pharo1.2.2-12353 that doesn't show up in earlier versions. This error has to do with OcompletionPharoCompatibility and is likely due to a configuration error somewhere.
I'll keep on digging...
Dale
----- Original Message -----
In the mail to Lukas below, I was able to identify a previously unreported bug in Metacello1.0-beta.28.3.3 when loading Seaside3.0.5 into Pharo1.2-12339. In the process of characterizing the bug, I discovered that the Seaside load was correct when using Metacello 1.0-beta.29 (an unreleased version).
However, when I tested Seaside3.0.5 and Metacello1.0-beta.29 using Pharo1.2.2-12353, I discovered more odd behavior and Seaside3.0.5 hit errors during the load. There is something fishy going on and at this point in time the root cause is not obvious.
The Metacello unit tests are passing in all three versions: Pharo1.2-12339, Pharo1.2.1-12345 and Pharo1.2.2-12353.
I am now in the process of attempting to characterize the bugs.
I will keep you posted on my progress.
Dale
----- Original Message -----
Lukas,
I have looked specifically at Seaside3.0.5 and Pharo1.2-12339/PharoCore1.2-12336 and have found two things:
1. #stable version for ConfigurationOfOmniBrowser for 1.2 is 1.2.3 and it should be 1.2.4. 2. if OB 1.2.4 is already loaded then the version is downgraded to 1.2.3
Item #1 is a configuration issue and item #2 is a bug ...
I had not heard that "many projects could no longer be loaded into Pharo1.2", so I was (and still am) blissfully unaware of the problems that folks are having.
Seaside3.0.5 loads into Pharo1.2/PharoCore1.2 and passes all of the unit tests ... so there are no "load issues." The Seaside control panel works fine and I have not noticed any other OmniBrowser-related issues ... but then I don't live in the images day in and day out so there could be subtle issues.
I am currently trying to characterized the "downgrade version" bug and will push out a fix as soon as I have one.
Dale
----- Original Message -----
Here's the current state of 3.0.5: I saved the images from Jenkins and uploaded them to the web site. The change log has also been updated [1]. Everything but the Metacello configuration seems to work fine. If I load the Metacello configuration into a Pharo 1.2 it tries to load a version of OB that does not work in Pharo 1.2.
As far as I know most configurations have been updated to Pharo 1.3 already and many projects (I learned that with Glamour, Mondrian, Moose, Spy, ...) cannot be loaded into latest stable (Pharo 1.2) anymore. What surprised me more is that Metacello does not allow to load old versions (at least this is true for the configurations I had problems with). So I guess we have the same problem here.
I suggest that you load OB with this script: https://github.com/renggli/builder/blob/master/scripts/omnibrowser.st
Lukas
-- Lukas Renggli www.lukas-renggli.ch _______________________________________________ seaside-dev mailing list seaside-dev@lists.squeakfoundation.org http://lists.squeakfoundation.org/mailman/listinfo/seaside-dev
_______________________________________________ seaside-dev mailing list seaside-dev@lists.squeakfoundation.org http://lists.squeakfoundation.org/mailman/listinfo/seaside-dev
Ocompletion-ul.102 was loaded because ConfigurationOfPharo 1.2.2 loads ConfigurationOfOmniBrowser 1.2.5 which loads ConfigurationOfOCompletion 1.3 which is not correct for Pharo 1.2.2 ... It looks like ConfigurationOfPharo 1.2.2 was introduced for Pharo 1.2.2 and previous versions of Pharo-1.2 were using ConfigurationOfPharo 1.2-beta2 which calls for ConfigurationOfOmniBrowser 1.2.4, which is presumably the version that should be used in Pharo 1.2. ConfigurationOfOmniBrowser 1.2.5 looks like a direct copy of ConfigurationOfOmniBrowser 1.3 and 1.3 _is_ targetted at Pharo1.3 ... With the way things are configured, any attempt to load a configuration that either directly or indirectly references ConfigurationOfOmniBrowser will likely fail... ....Now on to characterizing the Metacello bug:) Dale ----- Original Message -----
The Seaside3.0.5 load failure in Pharo1.2.2-12353 is due to the fact that Ocompletion-ul.102 is loaded in the image which is a package that is part of ConfigurationOfOCompletion 1.4 (presumably targetted at Pharo1.3 not Pharo1.2.2) ... In the earlier version of Pharo 1.2, Ocompletion-ul.92 is loaded and it is part of ConfigurationOfOCompletion 1.2.2 ...
Not sure why Ocompletion-ul.102 was loaded as ConfigurationOfPharo points to the correct version of ConfigurationOfOCompletion, so I'll have to do more detective work...
Dale
----- Original Message -----
Okay the odd behavior between Pharo1.2-12339, Pharo1.2.1-12345 and Pharo1.2.2-12353 can be attributed to a known Metacello issue:
http://code.google.com/p/metacello/issues/detail?id=112
So we are only facing two bugs:
- the issue reported by Lukas, where version OmniBrowser 1.2.3 is loaded on top of version 1.2.4 (when nothing should have been loaded). Which appears to be fixed by Metacello 1.0-beta.29 (still need a test case to reproduce the bug) - the Seaside3.0.5 load failure Pharo1.2.2-12353 that doesn't show up in earlier versions. This error has to do with OcompletionPharoCompatibility and is likely due to a configuration error somewhere.
I'll keep on digging...
Dale
----- Original Message -----
In the mail to Lukas below, I was able to identify a previously unreported bug in Metacello1.0-beta.28.3.3 when loading Seaside3.0.5 into Pharo1.2-12339. In the process of characterizing the bug, I discovered that the Seaside load was correct when using Metacello 1.0-beta.29 (an unreleased version).
However, when I tested Seaside3.0.5 and Metacello1.0-beta.29 using Pharo1.2.2-12353, I discovered more odd behavior and Seaside3.0.5 hit errors during the load. There is something fishy going on and at this point in time the root cause is not obvious.
The Metacello unit tests are passing in all three versions: Pharo1.2-12339, Pharo1.2.1-12345 and Pharo1.2.2-12353.
I am now in the process of attempting to characterize the bugs.
I will keep you posted on my progress.
Dale
----- Original Message -----
Lukas,
I have looked specifically at Seaside3.0.5 and Pharo1.2-12339/PharoCore1.2-12336 and have found two things:
1. #stable version for ConfigurationOfOmniBrowser for 1.2 is 1.2.3 and it should be 1.2.4. 2. if OB 1.2.4 is already loaded then the version is downgraded to 1.2.3
Item #1 is a configuration issue and item #2 is a bug ...
I had not heard that "many projects could no longer be loaded into Pharo1.2", so I was (and still am) blissfully unaware of the problems that folks are having.
Seaside3.0.5 loads into Pharo1.2/PharoCore1.2 and passes all of the unit tests ... so there are no "load issues." The Seaside control panel works fine and I have not noticed any other OmniBrowser-related issues ... but then I don't live in the images day in and day out so there could be subtle issues.
I am currently trying to characterized the "downgrade version" bug and will push out a fix as soon as I have one.
Dale
----- Original Message -----
Here's the current state of 3.0.5: I saved the images from Jenkins and uploaded them to the web site. The change log has also been updated [1]. Everything but the Metacello configuration seems to work fine. If I load the Metacello configuration into a Pharo 1.2 it tries to load a version of OB that does not work in Pharo 1.2.
As far as I know most configurations have been updated to Pharo 1.3 already and many projects (I learned that with Glamour, Mondrian, Moose, Spy, ...) cannot be loaded into latest stable (Pharo 1.2) anymore. What surprised me more is that Metacello does not allow to load old versions (at least this is true for the configurations I had problems with). So I guess we have the same problem here.
I suggest that you load OB with this script: https://github.com/renggli/builder/blob/master/scripts/omnibrowser.st
Lukas
-- Lukas Renggli www.lukas-renggli.ch _______________________________________________ seaside-dev mailing list seaside-dev@lists.squeakfoundation.org http://lists.squeakfoundation.org/mailman/listinfo/seaside-dev
_______________________________________________ seaside-dev mailing list seaside-dev@lists.squeakfoundation.org http://lists.squeakfoundation.org/mailman/listinfo/seaside-dev
participants (1)
-
Dale Henrichs