Mac release broken? ClassBinding>>objectForDataStrea
Locked
Mac release broken? ClassBinding>>objectForDataStrea
 
|
For both 4.6 & 5.0 on the mac, saving a project generates the following error: "NotImplemented: ClassBinding or a superclass should implement objectForDataStream."
The comments in the method suggest this shouldn't have been released to a stable version: "It's not yet clear how serialization should work in the presence of environments". Problem 2 is that if I then "abandon" the error messge, the workspace is then flooded with "MessageNotUnderstood" error boxes. Is "environments" something experimental that should be switched off? Best, Huw |
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
Arbitrarily picking an earlier version ... 4.1 does not have this issue/bug. Any quick fixes for this? I'm interested in using this in conjunction with aida. Best, Huw On 3 December 2015 at 01:13, Huw Lloyd <[hidden email]> wrote:
|
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
Huw,
The timing of your question is perfect :-) There are some very recent updates (in the last week or so) to the Squeak trunk development stream that address this problem. This is work in progress, but in a nutshell, the latest trunk fixes should resolve the problem for 4.6 images, but there is work still to be done in order to make it work in 5.0 Spur. We still need to get these latest updates into the standard update stream for squeak46, but as soon as that is complete it should work for those images. I am less certain of the timing for 5.0 images. Dave On Thu, Dec 03, 2015 at 01:25:37AM +0000, Huw Lloyd wrote: > Arbitrarily picking an earlier version ... 4.1 does not have this issue/bug. > > Any quick fixes for this? I'm interested in using this in conjunction with > aida. > > Best, > Huw > > > On 3 December 2015 at 01:13, Huw Lloyd <[hidden email]> wrote: > > > For both 4.6 & 5.0 on the mac, saving a project generates the following > > error: "NotImplemented: ClassBinding or a superclass should implement > > objectForDataStream." > > > > The comments in the method suggest this shouldn't have been released to a > > stable version: "It's not yet clear how serialization should work in the > > presence of environments". > > > > Problem 2 is that if I then "abandon" the error messge, the workspace is > > then flooded with "MessageNotUnderstood" error boxes. > > > > Is "environments" something experimental that should be switched off? > > > > Best, > > Huw > > > > > > > |
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
Thanks Dave. How do I achieve that -- do I swap in a trunk image to a 4.6 application? What about the VM? Best, Huw On 3 December 2015 at 02:00, David T. Lewis <[hidden email]> wrote: Huw, |
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
In reply to this post by David T. Lewis
Huw,
See a proposal for a quick fix below using the work of Colin Putney last week. --Hannes On 12/3/15, David T. Lewis <[hidden email]> wrote: > Huw, > > The timing of your question is perfect :-) > > There are some very recent updates (in the last week or so) to the Squeak > trunk development stream that address this problem. A copy of the mail describing these updates is here http://wiki.squeak.org/squeak/6218 This is work in > progress, > but in a nutshell, the latest trunk fixes should resolve the problem for > 4.6 images, but there is work still to be done in order to make it work > in 5.0 Spur. We still need to get these latest updates into the standard > update stream for squeak46, but as soon as that is complete it should work > for those images. The updates Colin Putney committed to the trunk on Friday, Saturday 27/28th Nov are Name: Environments-cwp.60 Author: cwp Time: 27 November 2015, 5:14:36.376 pm UUID: c8836cde-f21e-4652-a2e9-73f8de6f00e0 Ancestors: Environments-kfr.59 Always serialize Bindings as DiskProxies. Name: System-cwp.781 Author: cwp Time: 27 November 2015, 5:22:38.591 pm UUID: c0ffbab7-f7ee-4966-ba07-d57436ac42f2 Ancestors: System-mt.780 Remove specialized serialization of Associations. Since associations are no longer used as globals, we never need to serialize them using DiskProxy. Name: System-cwp.782 Author: cwp Time: 27 November 2015, 9:00:42.904 pm UUID: ae1f1166-0438-4e78-8ded-46c59c212210 Ancestors: System-cwp.781 Reload DiskProxies using the current environment When looking up objects from the destination image, DiskProxies should use the current environment, rather than referring to to Smalltalk. This lets objects on disk be loaded into arbitrary environments. These updates are committed to the trunk (i.e. the ongoing development after the release of 5.0). However in a trunk image http://build.squeak.org/job/Trunk/default/lastSuccessfulBuild/artifact/target/TrunkImage.zip they do NOT work yet. What you could do as of now ----------------------------------------- - File in the updates above into a 4.6 image. - It should fix the problem (I have not tried yet, so please report if it is successful). > I am less certain of the timing for 5.0 images. The reason is that the new VM crashes, and Colin has reported this to the Virtual Machine List. > Dave > > > > On Thu, Dec 03, 2015 at 01:25:37AM +0000, Huw Lloyd wrote: >> Arbitrarily picking an earlier version ... 4.1 does not have this >> issue/bug. >> >> Any quick fixes for this? I'm interested in using this in conjunction >> with >> aida. >> >> Best, >> Huw >> >> >> On 3 December 2015 at 01:13, Huw Lloyd <[hidden email]> wrote: >> >> > For both 4.6 & 5.0 on the mac, saving a project generates the following >> > error: "NotImplemented: ClassBinding or a superclass should implement >> > objectForDataStream." >> > >> > The comments in the method suggest this shouldn't have been released to >> > a >> > stable version: "It's not yet clear how serialization should work in >> > the >> > presence of environments". >> > >> > Problem 2 is that if I then "abandon" the error messge, the workspace >> > is >> > then flooded with "MessageNotUnderstood" error boxes. >> > >> > Is "environments" something experimental that should be switched off? >> > >> > Best, >> > Huw >> > >> > >> > > >> > > > |
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
P.S.
The report for this issue is here http://bugs.squeak.org/view.php?id=7814 (not yet updated with the new development of last week) On 12/3/15, H. Hirzel <[hidden email]> wrote: > Huw, > > See a proposal for a quick fix below using the work of Colin Putney last > week. > > --Hannes > > On 12/3/15, David T. Lewis <[hidden email]> wrote: >> Huw, >> >> The timing of your question is perfect :-) >> >> There are some very recent updates (in the last week or so) to the Squeak >> trunk development stream that address this problem. > > A copy of the mail describing these updates is here > http://wiki.squeak.org/squeak/6218 > > This is work in >> progress, >> but in a nutshell, the latest trunk fixes should resolve the problem for >> 4.6 images, but there is work still to be done in order to make it work >> in 5.0 Spur. We still need to get these latest updates into the standard >> update stream for squeak46, but as soon as that is complete it should >> work >> for those images. > > The updates Colin Putney committed to the trunk on Friday, Saturday > 27/28th Nov are > > Name: Environments-cwp.60 > Author: cwp > Time: 27 November 2015, 5:14:36.376 pm > UUID: c8836cde-f21e-4652-a2e9-73f8de6f00e0 > Ancestors: Environments-kfr.59 > > Always serialize Bindings as DiskProxies. > > > Name: System-cwp.781 > Author: cwp > Time: 27 November 2015, 5:22:38.591 pm > UUID: c0ffbab7-f7ee-4966-ba07-d57436ac42f2 > Ancestors: System-mt.780 > > Remove specialized serialization of Associations. > Since associations are no longer used as globals, we never need to > serialize them using DiskProxy. > > > Name: System-cwp.782 > Author: cwp > Time: 27 November 2015, 9:00:42.904 pm > UUID: ae1f1166-0438-4e78-8ded-46c59c212210 > Ancestors: System-cwp.781 > > Reload DiskProxies using the current environment > > When looking up objects from the destination image, DiskProxies should > use the current environment, rather than referring to to Smalltalk. > This lets objects on disk be loaded into arbitrary environments. > > These updates are committed to the trunk (i.e. the ongoing development > after the release of 5.0). However in a trunk image > http://build.squeak.org/job/Trunk/default/lastSuccessfulBuild/artifact/target/TrunkImage.zip > they do NOT work yet. > > > What you could do as of now > ----------------------------------------- > > - File in the updates above into a 4.6 image. > - It should fix the problem (I have not tried yet, so please report if > it is successful). > > > > >> I am less certain of the timing for 5.0 images. > > The reason is that the new VM crashes, and Colin has reported this to > the Virtual Machine List. > > > >> Dave >> >> >> >> On Thu, Dec 03, 2015 at 01:25:37AM +0000, Huw Lloyd wrote: >>> Arbitrarily picking an earlier version ... 4.1 does not have this >>> issue/bug. >>> >>> Any quick fixes for this? I'm interested in using this in conjunction >>> with >>> aida. >>> >>> Best, >>> Huw >>> >>> >>> On 3 December 2015 at 01:13, Huw Lloyd <[hidden email]> >>> wrote: >>> >>> > For both 4.6 & 5.0 on the mac, saving a project generates the >>> > following >>> > error: "NotImplemented: ClassBinding or a superclass should implement >>> > objectForDataStream." >>> > >>> > The comments in the method suggest this shouldn't have been released >>> > to >>> > a >>> > stable version: "It's not yet clear how serialization should work in >>> > the >>> > presence of environments". >>> > >>> > Problem 2 is that if I then "abandon" the error messge, the workspace >>> > is >>> > then flooded with "MessageNotUnderstood" error boxes. >>> > >>> > Is "environments" something experimental that should be switched off? >>> > >>> > Best, >>> > Huw >>> > >>> > >>> > >> >>> >> >> >> > |
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
In reply to this post by Hannes Hirzel
Hannes, I got as far as a failure to load System-cwp.781 with missing dependency on class UnhandledWarning: "Warning: This package depends on the following classes: UnhandledWarning You must resolve these dependencies before you will be able to load these definitions: UnhandledWarning>>defaultAction" The version of package system for 4.6 is System-topa.753 Best, Huw On 3 December 2015 at 08:57, H. Hirzel <[hidden email]> wrote: Huw, |
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
Looks earliest System package to not load to 4.6 is: Name: System-ul.757 Author: ul Time: 14 August 2015, 7:19:00.474 pm UUID: fca51b99-736e-4666-acac-e5c5b6a1d397 Ancestors: System-cmm.756 #includesSubString: -> #includesSubstring: #subStrings -> #substrings Best, Huw On 3 December 2015 at 14:43, Huw Lloyd <[hidden email]> wrote:
|
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
UnhandledWarning was added in Best, Karl On Thu, Dec 3, 2015 at 3:58 PM, Huw Lloyd <[hidden email]> wrote:
|
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
In reply to this post by Huw Lloyd-2
As direct loading does not work I suggest something like copy/paste
updated methods from a Squeak 5 trunk image one by one. back to the 4.6 image. I.e. open the mcz file in a code browser but do not load it. If necessary the code has to be adapted to cater for updates which happened elsewhere in the meantime as you write #includesSubString: -> #includesSubstring: #subStrings -> #substrings It seems that there not all that many fixes. --Hannes On 12/3/15, Huw Lloyd <[hidden email]> wrote: > Looks earliest System package to not load to 4.6 is: > > Name: System-ul.757 > Author: ul > Time: 14 August 2015, 7:19:00.474 pm > UUID: fca51b99-736e-4666-acac-e5c5b6a1d397 > Ancestors: System-cmm.756 > > #includesSubString: -> #includesSubstring: > #subStrings -> #substrings > > Best, > Huw > > > On 3 December 2015 at 14:43, Huw Lloyd <[hidden email]> wrote: > >> Hannes, >> >> I got as far as a failure to load System-cwp.781 with missing dependency >> on class UnhandledWarning: >> >> "Warning: This package depends on the following classes: >> UnhandledWarning >> You must resolve these dependencies before you will be able to load these >> definitions: >> UnhandledWarning>>defaultAction" >> >> The version of package system for 4.6 is System-topa.753 >> >> Best, >> Huw >> >> >> On 3 December 2015 at 08:57, H. Hirzel <[hidden email]> wrote: >> >>> Huw, >>> >>> See a proposal for a quick fix below using the work of Colin Putney last >>> week. >>> >>> --Hannes >>> >>> On 12/3/15, David T. Lewis <[hidden email]> wrote: >>> > Huw, >>> > >>> > The timing of your question is perfect :-) >>> > >>> > There are some very recent updates (in the last week or so) to the >>> Squeak >>> > trunk development stream that address this problem. >>> >>> A copy of the mail describing these updates is here >>> http://wiki.squeak.org/squeak/6218 >>> >>> This is work in >>> > progress, >>> > but in a nutshell, the latest trunk fixes should resolve the problem >>> > for >>> > 4.6 images, but there is work still to be done in order to make it >>> > work >>> > in 5.0 Spur. We still need to get these latest updates into the >>> > standard >>> > update stream for squeak46, but as soon as that is complete it should >>> work >>> > for those images. >>> >>> The updates Colin Putney committed to the trunk on Friday, Saturday >>> 27/28th Nov are >>> >>> Name: Environments-cwp.60 >>> Author: cwp >>> Time: 27 November 2015, 5:14:36.376 pm >>> UUID: c8836cde-f21e-4652-a2e9-73f8de6f00e0 >>> Ancestors: Environments-kfr.59 >>> >>> Always serialize Bindings as DiskProxies. >>> >>> >>> Name: System-cwp.781 >>> Author: cwp >>> Time: 27 November 2015, 5:22:38.591 pm >>> UUID: c0ffbab7-f7ee-4966-ba07-d57436ac42f2 >>> Ancestors: System-mt.780 >>> >>> Remove specialized serialization of Associations. >>> Since associations are no longer used as globals, we never need to >>> serialize them using DiskProxy. >>> >>> >>> Name: System-cwp.782 >>> Author: cwp >>> Time: 27 November 2015, 9:00:42.904 pm >>> UUID: ae1f1166-0438-4e78-8ded-46c59c212210 >>> Ancestors: System-cwp.781 >>> >>> Reload DiskProxies using the current environment >>> >>> When looking up objects from the destination image, DiskProxies should >>> use the current environment, rather than referring to to Smalltalk. >>> This lets objects on disk be loaded into arbitrary environments. >>> >>> These updates are committed to the trunk (i.e. the ongoing development >>> after the release of 5.0). However in a trunk image >>> >>> http://build.squeak.org/job/Trunk/default/lastSuccessfulBuild/artifact/target/TrunkImage.zip >>> they do NOT work yet. >>> >>> >>> What you could do as of now >>> ----------------------------------------- >>> >>> - File in the updates above into a 4.6 image. >>> - It should fix the problem (I have not tried yet, so please report if >>> it is successful). >>> >>> >>> >>> >>> > I am less certain of the timing for 5.0 images. >>> >>> The reason is that the new VM crashes, and Colin has reported this to >>> the Virtual Machine List. >>> >>> >>> >>> > Dave >>> > >>> > >>> > >>> > On Thu, Dec 03, 2015 at 01:25:37AM +0000, Huw Lloyd wrote: >>> >> Arbitrarily picking an earlier version ... 4.1 does not have this >>> >> issue/bug. >>> >> >>> >> Any quick fixes for this? I'm interested in using this in >>> >> conjunction >>> >> with >>> >> aida. >>> >> >>> >> Best, >>> >> Huw >>> >> >>> >> >>> >> On 3 December 2015 at 01:13, Huw Lloyd <[hidden email]> >>> wrote: >>> >> >>> >> > For both 4.6 & 5.0 on the mac, saving a project generates the >>> following >>> >> > error: "NotImplemented: ClassBinding or a superclass should >>> >> > implement >>> >> > objectForDataStream." >>> >> > >>> >> > The comments in the method suggest this shouldn't have been >>> >> > released >>> to >>> >> > a >>> >> > stable version: "It's not yet clear how serialization should work >>> >> > in >>> >> > the >>> >> > presence of environments". >>> >> > >>> >> > Problem 2 is that if I then "abandon" the error messge, the >>> >> > workspace >>> >> > is >>> >> > then flooded with "MessageNotUnderstood" error boxes. >>> >> > >>> >> > Is "environments" something experimental that should be switched >>> >> > off? >>> >> > >>> >> > Best, >>> >> > Huw >>> >> > >>> >> > >>> >> > >>> > >>> >> >>> > >>> > >>> > >>> >>> >> > |
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
I'll backport the trunk changes to 4.6 tonight. On Thu, Dec 3, 2015 at 7:06 AM, H. Hirzel <[hidden email]> wrote: As direct loading does not work I suggest something like copy/paste |
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
In reply to this post by Hannes Hirzel
Best, Huw On 3 December 2015 at 15:06, H. Hirzel <[hidden email]> wrote: As direct loading does not work I suggest something like copy/paste |
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
In reply to this post by Colin Putney-3
On 3 December 2015 at 17:08, Colin Putney <[hidden email]> wrote:
Thanks, Colin. That will also enable new users to save projects. :) Best, Huw
|
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
In reply to this post by David T. Lewis
> There are some very recent updates (in the last week or so) to the Squeak
> trunk development stream that address this problem. This is work in progress, > but in a nutshell, the latest trunk fixes should resolve the problem for > 4.6 images, but there is work still to be done in order to make it work > in 5.0 Spur. We still need to get these latest updates into the standard > update stream for squeak46, but as soon as that is complete it should work > for those images. I am less certain of the timing for 5.0 images. I must've missed the conversation; what is the problem with 5.0 that Colin's Environments fix won't work? |
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
Here is Colin's report of last Saturday
http://wiki.squeak.org/squeak/6218 The new VM with the trunk images crashes with the environment fixes of Colin whereas the fixes work in 4.6 --Hannes On 12/3/15, Chris Muller <[hidden email]> wrote: >> There are some very recent updates (in the last week or so) to the Squeak >> trunk development stream that address this problem. This is work in >> progress, >> but in a nutshell, the latest trunk fixes should resolve the problem for >> 4.6 images, but there is work still to be done in order to make it work >> in 5.0 Spur. We still need to get these latest updates into the standard >> update stream for squeak46, but as soon as that is complete it should >> work >> for those images. I am less certain of the timing for 5.0 images. > > I must've missed the conversation; what is the problem with 5.0 that > Colin's Environments fix won't work? > > |
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
In reply to this post by Chris Muller-3
On Thu, Dec 3, 2015 at 9:35 AM, Chris Muller <[hidden email]> wrote: I must've missed the conversation; what is the problem with 5.0 that It's basically that Spur doesn't support ImageSegments yet. It crashes on when saving a segment, and fails to load one properly (but doesn't crash). -Colin |
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
Hi Colin, AFAIA the Spur VM does provide image segment support ;-). If the system is crashing there's a bug. But the image segment support does differ somewhat from the previous version since Sour does not support stable object enumeration and so relies on the segment load prim answering an array of the loaded objects and the image level code using this to post process the loaded objs rather than nextObject. There were a raft of changes about a month ago that made the image segment tests run successfully. Can you say more about what's broken? _,,,^..^,,,_ (phone)
|
Re: Mac release broken? ClassBinding>>objectForDataStrea
|
|
On Sun, Dec 6, 2015 at 9:26 AM, Eliot Miranda <[hidden email]> wrote:
The crash occurs when saving a project. I don't know what the issue is, but there's a reproducible case here: (Thanks to David for rescuing my overly large attachment!) In an apparently separate issue, when loading a project, I get an error, 'Segment has not been becommed into the loaded objects', raised by ImageSegment>>checkAndReportLoadError. The segment inst var is still an instance of WordArrayForSegment, rather than an Array, so it appears the primitive failed. Colin |
«
Return to Squeak - Dev
|
1 view|%1 views
| Free forum by Nabble | Edit this page |