[Pharo-project] KernelTests green
I have posted some patches to the inbox to make the Kernel Number tests run, and in the Issues page I propose to delete a bunch of nonsensical testHash tests. With these changes the Kernel tests should all be green. - on
We checked with lukas what is strange is that the same tests are green in 3.10. Stef On Jun 10, 2008, at 3:48 PM, Oscar Nierstrasz wrote:
I have posted some patches to the inbox to make the Kernel Number tests run, and in the Issues page I propose to delete a bunch of nonsensical testHash tests. With these changes the Kernel tests should all be green.
- on
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
OK, I investigated more deeply. It appears that SequenceableCollection>>hash behaves differently in Pharo and Squeak 3.10. Squeak: #(1) hash --> 22516953 Pharo: #(1) hash --> 61541560 Since DateAndTime>>hash is ultimately implemented in terms of Array>>hash (inherited from SequenceableCollection), this test fails. Not yet sure why these two behave differently ... Already the first line (self species hash) is different: Squeak: Array hash --> 259007228 Pharo: Array hash --> 2455 - on On Jun 10, 2008, at 6:34 PM, Stéphane Ducasse wrote:
We checked with lukas what is strange is that the same tests are green in 3.10.
Stef
On Jun 10, 2008, at 3:48 PM, Oscar Nierstrasz wrote:
I have posted some patches to the inbox to make the Kernel Number tests run, and in the Issues page I propose to delete a bunch of nonsensical testHash tests. With these changes the Kernel tests should all be green.
- on
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Excellent! may be this is the hash of class that changed. Marcus retrochanged the hash behavior that we harvested in 3.9. And may be in 3.10 they changed class hash too. Now I do not know if in 3.9 these tests were broken. Stef On Jun 10, 2008, at 9:32 PM, Oscar Nierstrasz wrote:
OK, I investigated more deeply.
It appears that SequenceableCollection>>hash behaves differently in Pharo and Squeak 3.10.
Squeak:
#(1) hash --> 22516953
Pharo:
#(1) hash --> 61541560
Since DateAndTime>>hash is ultimately implemented in terms of Array>>hash (inherited from SequenceableCollection), this test fails.
Not yet sure why these two behave differently ...
Already the first line (self species hash) is different:
Squeak:
Array hash --> 259007228
Pharo:
Array hash --> 2455
- on
On Jun 10, 2008, at 6:34 PM, Stéphane Ducasse wrote:
We checked with lukas what is strange is that the same tests are green in 3.10.
Stef
On Jun 10, 2008, at 3:48 PM, Oscar Nierstrasz wrote:
I have posted some patches to the inbox to make the Kernel Number tests run, and in the Issues page I propose to delete a bunch of nonsensical testHash tests. With these changes the Kernel tests should all be green.
- on
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
These tests are fine in Squeak 3.9 too. Further investigation reveals that Behaviour>>hash is missing in Pharo! So hash reverts to Object>>hash, which sends Object>>identityHash. In Pharo: Array hash --> 2455 Object identityHash --> 2274 In Squeak: Array hash --> 259007228 Object identityHash --> 2274 Looks like someone removed Behaviour>>hash ^ self name hash - on On Jun 10, 2008, at 9:43 PM, Stéphane Ducasse wrote:
Excellent! may be this is the hash of class that changed. Marcus retrochanged the hash behavior that we harvested in 3.9. And may be in 3.10 they changed class hash too.
Now I do not know if in 3.9 these tests were broken.
Stef
On Jun 10, 2008, at 9:32 PM, Oscar Nierstrasz wrote:
OK, I investigated more deeply.
It appears that SequenceableCollection>>hash behaves differently in Pharo and Squeak 3.10.
Squeak:
#(1) hash --> 22516953
Pharo:
#(1) hash --> 61541560
Since DateAndTime>>hash is ultimately implemented in terms of Array>>hash (inherited from SequenceableCollection), this test fails.
Not yet sure why these two behave differently ...
Already the first line (self species hash) is different:
Squeak:
Array hash --> 259007228
Pharo:
Array hash --> 2455
- on
On Jun 10, 2008, at 6:34 PM, Stéphane Ducasse wrote:
We checked with lukas what is strange is that the same tests are green in 3.10.
Stef
On Jun 10, 2008, at 3:48 PM, Oscar Nierstrasz wrote:
I have posted some patches to the inbox to make the Kernel Number tests run, and in the Issues page I propose to delete a bunch of nonsensical testHash tests. With these changes the Kernel tests should all be green.
- on
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Marcus should have a look since this is fix he reverted. Stef On Jun 10, 2008, at 10:04 PM, Oscar Nierstrasz wrote:
These tests are fine in Squeak 3.9 too.
Further investigation reveals that Behaviour>>hash is missing in Pharo!
So hash reverts to Object>>hash, which sends Object>>identityHash.
In Pharo:
Array hash --> 2455 Object identityHash --> 2274
In Squeak:
Array hash --> 259007228 Object identityHash --> 2274
Looks like someone removed
Behaviour>>hash ^ self name hash
- on
On Jun 10, 2008, at 9:43 PM, Stéphane Ducasse wrote:
Excellent! may be this is the hash of class that changed. Marcus retrochanged the hash behavior that we harvested in 3.9. And may be in 3.10 they changed class hash too.
Now I do not know if in 3.9 these tests were broken.
Stef
On Jun 10, 2008, at 9:32 PM, Oscar Nierstrasz wrote:
OK, I investigated more deeply.
It appears that SequenceableCollection>>hash behaves differently in Pharo and Squeak 3.10.
Squeak:
#(1) hash --> 22516953
Pharo:
#(1) hash --> 61541560
Since DateAndTime>>hash is ultimately implemented in terms of Array>>hash (inherited from SequenceableCollection), this test fails.
Not yet sure why these two behave differently ...
Already the first line (self species hash) is different:
Squeak:
Array hash --> 259007228
Pharo:
Array hash --> 2455
- on
On Jun 10, 2008, at 6:34 PM, Stéphane Ducasse wrote:
We checked with lukas what is strange is that the same tests are green in 3.10.
Stef
On Jun 10, 2008, at 3:48 PM, Oscar Nierstrasz wrote:
I have posted some patches to the inbox to make the Kernel Number tests run, and in the Issues page I propose to delete a bunch of nonsensical testHash tests. With these changes the Kernel tests should all be green.
- on
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo- project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
What was his fix? How can I tell what was changed? Is Behaviour>>hash missing in your image too? I don't trust anything anymore. - on On Jun 10, 2008, at 10:13 PM, Stéphane Ducasse wrote:
Marcus should have a look since this is fix he reverted.
Stef
On Jun 10, 2008, at 10:04 PM, Oscar Nierstrasz wrote:
These tests are fine in Squeak 3.9 too.
Further investigation reveals that Behaviour>>hash is missing in Pharo!
So hash reverts to Object>>hash, which sends Object>>identityHash.
In Pharo:
Array hash --> 2455 Object identityHash --> 2274
In Squeak:
Array hash --> 259007228 Object identityHash --> 2274
Looks like someone removed
Behaviour>>hash ^ self name hash
- on
On Jun 10, 2008, at 9:43 PM, Stéphane Ducasse wrote:
Excellent! may be this is the hash of class that changed. Marcus retrochanged the hash behavior that we harvested in 3.9. And may be in 3.10 they changed class hash too.
Now I do not know if in 3.9 these tests were broken.
Stef
On Jun 10, 2008, at 9:32 PM, Oscar Nierstrasz wrote:
OK, I investigated more deeply.
It appears that SequenceableCollection>>hash behaves differently in Pharo and Squeak 3.10.
Squeak:
#(1) hash --> 22516953
Pharo:
#(1) hash --> 61541560
Since DateAndTime>>hash is ultimately implemented in terms of Array>>hash (inherited from SequenceableCollection), this test fails.
Not yet sure why these two behave differently ...
Already the first line (self species hash) is different:
Squeak:
Array hash --> 259007228
Pharo:
Array hash --> 2455
- on
On Jun 10, 2008, at 6:34 PM, Stéphane Ducasse wrote:
We checked with lukas what is strange is that the same tests are green in 3.10.
Stef
On Jun 10, 2008, at 3:48 PM, Oscar Nierstrasz wrote:
I have posted some patches to the inbox to make the Kernel Number tests run, and in the Issues page I propose to delete a bunch of nonsensical testHash tests. With these changes the Kernel tests should all be green.
- on
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo- project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
IIRC Marcus introduced Behavior>>hash in 3.9 and later realized that this is not a good idea. The reason is that names of behaviors are not constant because when a class is deleted its name is changed to AnAbsolete..., which makes the name not suitable for hashing. Looking through ScriptLoader, #update10020 is where Behavior>>hash is removed again. Unfortunately the changes log is a bit unspecific as it mentions only: "changes from Reflectivity"... Cheers, Adrian On Jun 10, 2008, at 22:54 , Oscar Nierstrasz wrote:
What was his fix?
How can I tell what was changed?
Is Behaviour>>hash missing in your image too? I don't trust anything anymore.
- on
On Jun 10, 2008, at 10:13 PM, Stéphane Ducasse wrote:
Marcus should have a look since this is fix he reverted.
Stef
On Jun 10, 2008, at 10:04 PM, Oscar Nierstrasz wrote:
These tests are fine in Squeak 3.9 too.
Further investigation reveals that Behaviour>>hash is missing in Pharo!
So hash reverts to Object>>hash, which sends Object>>identityHash.
In Pharo:
Array hash --> 2455 Object identityHash --> 2274
In Squeak:
Array hash --> 259007228 Object identityHash --> 2274
Looks like someone removed
Behaviour>>hash ^ self name hash
- on
On Jun 10, 2008, at 9:43 PM, Stéphane Ducasse wrote:
Excellent! may be this is the hash of class that changed. Marcus retrochanged the hash behavior that we harvested in 3.9. And may be in 3.10 they changed class hash too.
Now I do not know if in 3.9 these tests were broken.
Stef
On Jun 10, 2008, at 9:32 PM, Oscar Nierstrasz wrote:
OK, I investigated more deeply.
It appears that SequenceableCollection>>hash behaves differently in Pharo and Squeak 3.10.
Squeak:
#(1) hash --> 22516953
Pharo:
#(1) hash --> 61541560
Since DateAndTime>>hash is ultimately implemented in terms of Array>>hash (inherited from SequenceableCollection), this test fails.
Not yet sure why these two behave differently ...
Already the first line (self species hash) is different:
Squeak:
Array hash --> 259007228
Pharo:
Array hash --> 2455
- on
On Jun 10, 2008, at 6:34 PM, Stéphane Ducasse wrote:
We checked with lukas what is strange is that the same tests are green in 3.10.
Stef
On Jun 10, 2008, at 3:48 PM, Oscar Nierstrasz wrote:
I have posted some patches to the inbox to make the Kernel Number tests run, and in the Issues page I propose to delete a bunch of nonsensical testHash tests. With these changes the Kernel tests should all be green.
- on
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo- project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
OK, so that is why all these silly testHash tests are broken. They depend on Behaviour>>#hash to return a specific hash value. I suggest trashing them as the tests are not robust. - on On Jun 11, 2008, at 6:03 PM, Adrian Lienhard wrote:
IIRC Marcus introduced Behavior>>hash in 3.9 and later realized that this is not a good idea. The reason is that names of behaviors are not constant because when a class is deleted its name is changed to AnAbsolete..., which makes the name not suitable for hashing.
Looking through ScriptLoader, #update10020 is where Behavior>>hash is removed again. Unfortunately the changes log is a bit unspecific as it mentions only: "changes from Reflectivity"...
Cheers, Adrian
On Jun 10, 2008, at 22:54 , Oscar Nierstrasz wrote:
What was his fix?
How can I tell what was changed?
Is Behaviour>>hash missing in your image too? I don't trust anything anymore.
- on
On Jun 10, 2008, at 10:13 PM, Stéphane Ducasse wrote:
Marcus should have a look since this is fix he reverted.
Stef
On Jun 10, 2008, at 10:04 PM, Oscar Nierstrasz wrote:
These tests are fine in Squeak 3.9 too.
Further investigation reveals that Behaviour>>hash is missing in Pharo!
So hash reverts to Object>>hash, which sends Object>>identityHash.
In Pharo:
Array hash --> 2455 Object identityHash --> 2274
In Squeak:
Array hash --> 259007228 Object identityHash --> 2274
Looks like someone removed
Behaviour>>hash ^ self name hash
- on
On Jun 10, 2008, at 9:43 PM, Stéphane Ducasse wrote:
Excellent! may be this is the hash of class that changed. Marcus retrochanged the hash behavior that we harvested in 3.9. And may be in 3.10 they changed class hash too.
Now I do not know if in 3.9 these tests were broken.
Stef
On Jun 10, 2008, at 9:32 PM, Oscar Nierstrasz wrote:
OK, I investigated more deeply.
It appears that SequenceableCollection>>hash behaves differently in Pharo and Squeak 3.10.
Squeak:
#(1) hash --> 22516953
Pharo:
#(1) hash --> 61541560
Since DateAndTime>>hash is ultimately implemented in terms of Array>>hash (inherited from SequenceableCollection), this test fails.
Not yet sure why these two behave differently ...
Already the first line (self species hash) is different:
Squeak:
Array hash --> 259007228
Pharo:
Array hash --> 2455
- on
On Jun 10, 2008, at 6:34 PM, Stéphane Ducasse wrote:
We checked with lukas what is strange is that the same tests are green in 3.10.
Stef
On Jun 10, 2008, at 3:48 PM, Oscar Nierstrasz wrote:
I have posted some patches to the inbox to make the Kernel Number tests run, and in the Issues page I propose to delete a bunch of nonsensical testHash tests. With these changes the Kernel tests should all be green.
- on
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo- project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Yes I have the impression that this would be a good things to do because they do not have add that much important information. On Jun 11, 2008, at 6:11 PM, Oscar Nierstrasz wrote:
OK, so that is why all these silly testHash tests are broken. They depend on Behaviour>>#hash to return a specific hash value. I suggest trashing them as the tests are not robust.
- on
On Jun 11, 2008, at 6:03 PM, Adrian Lienhard wrote:
IIRC Marcus introduced Behavior>>hash in 3.9 and later realized that this is not a good idea. The reason is that names of behaviors are not constant because when a class is deleted its name is changed to AnAbsolete..., which makes the name not suitable for hashing.
Looking through ScriptLoader, #update10020 is where Behavior>>hash is removed again. Unfortunately the changes log is a bit unspecific as it mentions only: "changes from Reflectivity"...
Cheers, Adrian
On Jun 10, 2008, at 22:54 , Oscar Nierstrasz wrote:
What was his fix?
How can I tell what was changed?
Is Behaviour>>hash missing in your image too? I don't trust anything anymore.
- on
On Jun 10, 2008, at 10:13 PM, Stéphane Ducasse wrote:
Marcus should have a look since this is fix he reverted.
Stef
On Jun 10, 2008, at 10:04 PM, Oscar Nierstrasz wrote:
These tests are fine in Squeak 3.9 too.
Further investigation reveals that Behaviour>>hash is missing in Pharo!
So hash reverts to Object>>hash, which sends Object>>identityHash.
In Pharo:
Array hash --> 2455 Object identityHash --> 2274
In Squeak:
Array hash --> 259007228 Object identityHash --> 2274
Looks like someone removed
Behaviour>>hash ^ self name hash
- on
On Jun 10, 2008, at 9:43 PM, Stéphane Ducasse wrote:
Excellent! may be this is the hash of class that changed. Marcus retrochanged the hash behavior that we harvested in 3.9. And may be in 3.10 they changed class hash too.
Now I do not know if in 3.9 these tests were broken.
Stef
On Jun 10, 2008, at 9:32 PM, Oscar Nierstrasz wrote:
OK, I investigated more deeply.
It appears that SequenceableCollection>>hash behaves differently in Pharo and Squeak 3.10.
Squeak:
#(1) hash --> 22516953
Pharo:
#(1) hash --> 61541560
Since DateAndTime>>hash is ultimately implemented in terms of Array>>hash (inherited from SequenceableCollection), this test fails.
Not yet sure why these two behave differently ...
Already the first line (self species hash) is different:
Squeak:
Array hash --> 259007228
Pharo:
Array hash --> 2455
- on
On Jun 10, 2008, at 6:34 PM, Stéphane Ducasse wrote:
We checked with lukas what is strange is that the same tests are green in 3.10.
Stef
On Jun 10, 2008, at 3:48 PM, Oscar Nierstrasz wrote:
I have posted some patches to the inbox to make the Kernel Number tests run, and in the Issues page I propose to delete a bunch of nonsensical testHash tests. With these changes the Kernel tests should all be green.
- on
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo- project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
On 11.06.2008, at 19:11, Stéphane Ducasse wrote:
Yes I have the impression that this would be a good things to do because they do not have add that much important information.
Yes, that is my impression too.
On Jun 11, 2008, at 6:11 PM, Oscar Nierstrasz wrote:
OK, so that is why all these silly testHash tests are broken. They depend on Behaviour>>#hash to return a specific hash value. I suggest trashing them as the tests are not robust.
- on
On Jun 11, 2008, at 6:03 PM, Adrian Lienhard wrote:
IIRC Marcus introduced Behavior>>hash in 3.9 and later realized that this is not a good idea. The reason is that names of behaviors are not constant because when a class is deleted its name is changed to AnAbsolete..., which makes the name not suitable for hashing.
Looking through ScriptLoader, #update10020 is where Behavior>>hash is removed again. Unfortunately the changes log is a bit unspecific as it mentions only: "changes from Reflectivity"...
Cheers, Adrian
On Jun 10, 2008, at 22:54 , Oscar Nierstrasz wrote:
What was his fix?
How can I tell what was changed?
Is Behaviour>>hash missing in your image too? I don't trust anything anymore.
- on
On Jun 10, 2008, at 10:13 PM, Stéphane Ducasse wrote:
Marcus should have a look since this is fix he reverted.
Stef
On Jun 10, 2008, at 10:04 PM, Oscar Nierstrasz wrote:
These tests are fine in Squeak 3.9 too.
Further investigation reveals that Behaviour>>hash is missing in Pharo!
So hash reverts to Object>>hash, which sends Object>>identityHash.
In Pharo:
Array hash --> 2455 Object identityHash --> 2274
In Squeak:
Array hash --> 259007228 Object identityHash --> 2274
Looks like someone removed
Behaviour>>hash ^ self name hash
- on
On Jun 10, 2008, at 9:43 PM, Stéphane Ducasse wrote:
Excellent! may be this is the hash of class that changed. Marcus retrochanged the hash behavior that we harvested in 3.9. And may be in 3.10 they changed class hash too.
Now I do not know if in 3.9 these tests were broken.
Stef
On Jun 10, 2008, at 9:32 PM, Oscar Nierstrasz wrote:
OK, I investigated more deeply.
It appears that SequenceableCollection>>hash behaves differently in Pharo and Squeak 3.10.
Squeak:
#(1) hash --> 22516953
Pharo:
#(1) hash --> 61541560
Since DateAndTime>>hash is ultimately implemented in terms of Array>>hash (inherited from SequenceableCollection), this test fails.
Not yet sure why these two behave differently ...
Already the first line (self species hash) is different:
Squeak:
Array hash --> 259007228
Pharo:
Array hash --> 2455
- on
On Jun 10, 2008, at 6:34 PM, Stéphane Ducasse wrote:
We checked with lukas what is strange is that the same tests are green in 3.10.
Stef
On Jun 10, 2008, at 3:48 PM, Oscar Nierstrasz wrote:
I have posted some patches to the inbox to make the Kernel Number tests run, and in the Issues page I propose to delete a bunch of nonsensical testHash tests. With these changes the Kernel tests should all be green.
- on
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo- project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Marcus -- Marcus Denker -- denker@iam.unibe.ch http://www.iam.unibe.ch/~denker
On 11.06.2008, at 18:03, Adrian Lienhard wrote:
IIRC Marcus introduced Behavior>>hash in 3.9 and later realized that this is not a good idea. The reason is that names of behaviors are not constant because when a class is deleted its name is changed to AnAbsolete..., which makes the name not suitable for hashing.
Yes, that was the reason.
Looking through ScriptLoader, #update10020 is where Behavior>>hash is removed again. Unfortunately the changes log is a bit unspecific as it mentions only: "changes from Reflectivity"...
I will try to be more specific next time...
Cheers, Adrian
On Jun 10, 2008, at 22:54 , Oscar Nierstrasz wrote:
What was his fix?
How can I tell what was changed?
Is Behaviour>>hash missing in your image too? I don't trust anything anymore.
- on
On Jun 10, 2008, at 10:13 PM, Stéphane Ducasse wrote:
Marcus should have a look since this is fix he reverted.
Stef
On Jun 10, 2008, at 10:04 PM, Oscar Nierstrasz wrote:
These tests are fine in Squeak 3.9 too.
Further investigation reveals that Behaviour>>hash is missing in Pharo!
So hash reverts to Object>>hash, which sends Object>>identityHash.
In Pharo:
Array hash --> 2455 Object identityHash --> 2274
In Squeak:
Array hash --> 259007228 Object identityHash --> 2274
Looks like someone removed
Behaviour>>hash ^ self name hash
- on
On Jun 10, 2008, at 9:43 PM, Stéphane Ducasse wrote:
Excellent! may be this is the hash of class that changed. Marcus retrochanged the hash behavior that we harvested in 3.9. And may be in 3.10 they changed class hash too.
Now I do not know if in 3.9 these tests were broken.
Stef
On Jun 10, 2008, at 9:32 PM, Oscar Nierstrasz wrote:
OK, I investigated more deeply.
It appears that SequenceableCollection>>hash behaves differently in Pharo and Squeak 3.10.
Squeak:
#(1) hash --> 22516953
Pharo:
#(1) hash --> 61541560
Since DateAndTime>>hash is ultimately implemented in terms of Array>>hash (inherited from SequenceableCollection), this test fails.
Not yet sure why these two behave differently ...
Already the first line (self species hash) is different:
Squeak:
Array hash --> 259007228
Pharo:
Array hash --> 2455
- on
On Jun 10, 2008, at 6:34 PM, Stéphane Ducasse wrote:
We checked with lukas what is strange is that the same tests are green in 3.10.
Stef
On Jun 10, 2008, at 3:48 PM, Oscar Nierstrasz wrote:
I have posted some patches to the inbox to make the Kernel Number tests run, and in the Issues page I propose to delete a bunch of nonsensical testHash tests. With these changes the Kernel tests should all be green.
- on
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo- project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Marcus -- Marcus Denker -- denker@iam.unibe.ch http://www.iam.unibe.ch/~denker
participants (4)
-
Adrian Lienhard -
Marcus Denker -
Oscar Nierstrasz -
Stéphane Ducasse