El 10/10/07 9:40 AM, "[hidden email]" <[hidden email]> escribió: > > On 10-10-07 12:40, Damien Cassou sent you this reminder about: > > http://bugs.squeak.org/view.php?id=5641 > > Don't forget this bug! No, but seems is not ready > The most recent "RecentChanges" seem to be in the parent report. > > Also as this upgrade has already caused problems once, I highly recommend > there be tests to prove it does not still offend. > > I would not even recommend attempting this fix without the tests. We have a requirement of bugs MUST have test. Or some other Squeaker say is a good fix. Here some say this is problematic. Should I do the update ? Sorry by answer in list, but this way all working on this could see we are working. Edgar |
Some changes from Keith
- made stringVersion inst var lazy initialized - provided a default value via #stringVersionDefault - provided a #asStringOrText accessor to replace #stringVersion The error from Keith was to let the #stringVersion accessor in place, letting it answer nil, and letting #stringVersion senders in trouble... Then, you have basically 2 solutions 1) revert to full initialization 2) remove the stringVersion accessor and make the ivar private jorsaria proposed 1) MethodReference-setStandardClassmethodSymbol.st http://bugs.squeak.org/view.php?id=6691 I proposed 2). StringVersion-Patch-M6691.1.cs http://bugs.squeak.org/view.php?id=6691 ---------------- If you choose 2), then some cosmetic changes might be good to do: The #stringVersion: accessor should be rename to setStringVersion: and classified in protocol 'setting', this would be more homogeneous with the rest of this protocol and enforce that the ivar is kind of private. For these reasons I would push StringVersion-Patch-M6691.2.cs http://bugs.squeak.org/view.php?id=6691 ------------------ RecentMessageSet-updateListsAndCodeIn.st http://bugs.squeak.org/view.php?id=5641 is a reversion to use #stringVersioninstead instead of #asStringOrText It will not work if ivar is lazy initialized. So it must go with 1) ------------------ RecentChanges.2.cs http://bugs.squeak.org/view.php?id=5641 seems a duplication of RecentChanges.2.cs http://bugs.squeak.org/view.php?id=6402 What a mess! Jerome said at http://bugs.squeak.org/view.php?id=6402 that this change is not viable as refering to classes not in base image ------------------ RecentChanges.3.cs http://bugs.squeak.org/view.php?id=6402 does use #asStringOrtext but does not rename #stringVersion: to #setStringVersion:. It seems a good candidate too. But it does some more work that i did not analyze. Up to Keith to document his work. Generally, most changes come from him, so you should ask to him. Nicolas Edgar J. De Cleene a écrit : > > > El 10/10/07 9:40 AM, "[hidden email]" > <[hidden email]> escribió: > >> On 10-10-07 12:40, Damien Cassou sent you this reminder about: >> >> http://bugs.squeak.org/view.php?id=5641 >> >> Don't forget this bug! > > No, but seems is not ready > >> The most recent "RecentChanges" seem to be in the parent report. >> >> Also as this upgrade has already caused problems once, I highly recommend >> there be tests to prove it does not still offend. >> >> I would not even recommend attempting this fix without the tests. > > > We have a requirement of bugs MUST have test. > Or some other Squeaker say is a good fix. > Here some say this is problematic. > Should I do the update ? > Sorry by answer in list, but this way all working on this could see we are > working. > > Edgar > > > > > |
El 10/10/07 5:05 PM, "nicolas cellier" <[hidden email]> escribió: > Some changes from Keith > - made stringVersion inst var lazy initialized > - provided a default value via #stringVersionDefault > - provided a #asStringOrText accessor to replace #stringVersion > > The error from Keith was to let the #stringVersion accessor in place, > letting it answer nil, and letting #stringVersion senders in trouble... > > Then, you have basically 2 solutions > 1) revert to full initialization > 2) remove the stringVersion accessor and make the ivar private > > jorsaria proposed 1) > MethodReference-setStandardClassmethodSymbol.st > http://bugs.squeak.org/view.php?id=6691 > > I proposed 2). > StringVersion-Patch-M6691.1.cs > http://bugs.squeak.org/view.php?id=6691 > > ---------------- > > If you choose 2), then some cosmetic changes might be good to do: > The #stringVersion: accessor should be rename to setStringVersion: and > classified in protocol 'setting', this would be more homogeneous with > the rest of this protocol and enforce that the ivar is kind of private. > > For these reasons I would push > StringVersion-Patch-M6691.2.cs > http://bugs.squeak.org/view.php?id=6691 > > ------------------ > > RecentMessageSet-updateListsAndCodeIn.st > http://bugs.squeak.org/view.php?id=5641 > > is a reversion to use #stringVersioninstead instead of #asStringOrText > It will not work if ivar is lazy initialized. > So it must go with 1) > > ------------------ > > RecentChanges.2.cs > http://bugs.squeak.org/view.php?id=5641 > > seems a duplication of > > RecentChanges.2.cs > http://bugs.squeak.org/view.php?id=6402 > > What a mess! > Jerome said at http://bugs.squeak.org/view.php?id=6402 that this change > is not viable as refering to classes not in base image > > ------------------ > > RecentChanges.3.cs > http://bugs.squeak.org/view.php?id=6402 > > does use #asStringOrtext but does not rename #stringVersion: to > #setStringVersion:. It seems a good candidate too. > But it does some more work that i did not analyze. > Up to Keith to document his work. > Generally, most changes come from him, so you should ask to him. > > Nicolas Well, I don't wish scare all with inside release affairs :=) Thanks to all. Keith ? Edgar |
Free forum by Nabble | Edit this page |