Hi All,
as you may know, squeaksource.com is down. I had a look and found that - the image had fallen into a GC cycle consuming 100% cpu and about 500 MB o memory - restarting the image didn't help with that - I added the -mmap 1024m parameter to its arguments, and it stopped the CPU grinding, and the memory usage has fallen below 500 MB since then. This also means that the VM's default memory limit is actually 500MB contrary to the 1GB stated by the VM. - but, the image is still not able to start up properly. it's probably stuck on something simple, but due to the lack of VNC, I just couldn't check what was wrong
Levente
Okay, I downloaded the image and started up on my machine. It greeted me with 2 errors. Both were related to not being able to start up the web server. Does this image save itself periodially? If so, it was saved when it had these debuggers open, basically killing itself.
Levente
On Fri, 24 Mar 2017, Levente Uzonyi wrote:
Hi All,
as you may know, squeaksource.com is down. I had a look and found that
- the image had fallen into a GC cycle consuming 100% cpu and about 500 MB o
memory
- restarting the image didn't help with that
- I added the -mmap 1024m parameter to its arguments, and it stopped the CPU
grinding, and the memory usage has fallen below 500 MB since then. This also means that the VM's default memory limit is actually 500MB contrary to the 1GB stated by the VM.
- but, the image is still not able to start up properly. it's probably stuck
on something simple, but due to the lack of VNC, I just couldn't check what was wrong
Levente
A few months ago, we lost weeks of user updates due to image persistence saving itself as an unlaunchable image (> 1GB interpreter). All the latest code for SqueakSource is ready and available at http://source.squeak.org/ss. This is the codebase that has been supporting source.squeak.org for several months now. It's stable.
Squeaksource is important. Is there any reason not to update the image and code ASAP?
On Fri, Mar 24, 2017 at 1:44 PM, Levente Uzonyi leves@caesar.elte.hu wrote:
Okay, I downloaded the image and started up on my machine. It greeted me with 2 errors. Both were related to not being able to start up the web server. Does this image save itself periodially? If so, it was saved when it had these debuggers open, basically killing itself.
Levente
On Fri, 24 Mar 2017, Levente Uzonyi wrote:
Hi All,
as you may know, squeaksource.com is down. I had a look and found that
- the image had fallen into a GC cycle consuming 100% cpu and about 500 MB
o memory
- restarting the image didn't help with that
- I added the -mmap 1024m parameter to its arguments, and it stopped the
CPU grinding, and the memory usage has fallen below 500 MB since then. This also means that the VM's default memory limit is actually 500MB contrary to the 1GB stated by the VM.
- but, the image is still not able to start up properly. it's probably
stuck on something simple, but due to the lack of VNC, I just couldn't check what was wrong
Levente
On Fri, 24 Mar 2017, Chris Muller wrote:
A few months ago, we lost weeks of user updates due to image persistence saving itself as an unlaunchable image (> 1GB interpreter). All the latest code for SqueakSource is ready and available at http://source.squeak.org/ss. This is the codebase that has been supporting source.squeak.org for several months now. It's stable.
Squeaksource is important. Is there any reason not to update the image and code ASAP?
Just one: someone has to do it. Someone who knows how to do it.
Levente
On Fri, Mar 24, 2017 at 1:44 PM, Levente Uzonyi leves@caesar.elte.hu wrote:
Okay, I downloaded the image and started up on my machine. It greeted me with 2 errors. Both were related to not being able to start up the web server. Does this image save itself periodially? If so, it was saved when it had these debuggers open, basically killing itself.
Levente
On Fri, 24 Mar 2017, Levente Uzonyi wrote:
Hi All,
as you may know, squeaksource.com is down. I had a look and found that
- the image had fallen into a GC cycle consuming 100% cpu and about 500 MB
o memory
- restarting the image didn't help with that
- I added the -mmap 1024m parameter to its arguments, and it stopped the
CPU grinding, and the memory usage has fallen below 500 MB since then. This also means that the VM's default memory limit is actually 500MB contrary to the 1GB stated by the VM.
- but, the image is still not able to start up properly. it's probably
stuck on something simple, but due to the lack of VNC, I just couldn't check what was wrong
Levente
I know how to do it. I can take it on in late April or early May, or assist someone else before then.
On Fri, Mar 24, 2017 at 2:04 PM, Levente Uzonyi leves@caesar.elte.hu wrote:
On Fri, 24 Mar 2017, Chris Muller wrote:
A few months ago, we lost weeks of user updates due to image persistence saving itself as an unlaunchable image (> 1GB interpreter). All the latest code for SqueakSource is ready and available at http://source.squeak.org/ss. This is the codebase that has been supporting source.squeak.org for several months now. It's stable.
Squeaksource is important. Is there any reason not to update the image and code ASAP?
Just one: someone has to do it. Someone who knows how to do it.
Levente
On Fri, Mar 24, 2017 at 1:44 PM, Levente Uzonyi leves@caesar.elte.hu wrote:
Okay, I downloaded the image and started up on my machine. It greeted me with 2 errors. Both were related to not being able to start up the web server. Does this image save itself periodially? If so, it was saved when it had these debuggers open, basically killing itself.
Levente
On Fri, 24 Mar 2017, Levente Uzonyi wrote:
Hi All,
as you may know, squeaksource.com is down. I had a look and found that
- the image had fallen into a GC cycle consuming 100% cpu and about 500
MB o memory
- restarting the image didn't help with that
- I added the -mmap 1024m parameter to its arguments, and it stopped the
CPU grinding, and the memory usage has fallen below 500 MB since then. This also means that the VM's default memory limit is actually 500MB contrary to the 1GB stated by the VM.
- but, the image is still not able to start up properly. it's probably
stuck on something simple, but due to the lack of VNC, I just couldn't check what was wrong
Levente
On Fri, Mar 24, 2017 at 07:24:08PM +0100, Levente Uzonyi wrote:
Hi All,
as you may know, squeaksource.com is down. I had a look and found that
- the image had fallen into a GC cycle consuming 100% cpu and about 500 MB
o memory
- restarting the image didn't help with that
- I added the -mmap 1024m parameter to its arguments, and it stopped the
CPU grinding, and the memory usage has fallen below 500 MB since then. This also means that the VM's default memory limit is actually 500MB contrary to the 1GB stated by the VM.
- but, the image is still not able to start up properly. it's probably
stuck on something simple, but due to the lack of VNC, I just couldn't check what was wrong
Levente
Sorry I was away, I just restarted it.
What I did:
- download the image, run it locally - close the debuggers - open process browser, find a process stuck in trying to send mail, terminate it - check memory size, verify that the web page opens on my localhost - copy it back to dan.box.squeak.org - kill the running image and let it restart with fixed image
VNC and mail from the squeaksource image do not work any more. The VNC used to work on dan.box.squeak.org but not now, possibly something changed in network configuration that the ancient image does not handle right. The stuck process that caused the failure seems to fail on an error that tries to send mail to inform someone of the problem, which fails because smtp does not work because of network address resolution, which fails and throws an error that tries to send more mail, etc.
As Chris said, the image is overdue for an update.
The service was off line but no data was loss, and as in previous incidents the data is safe on disk.
Dave
Thanks Dave!
Levente
On Fri, 24 Mar 2017, David T. Lewis wrote:
On Fri, Mar 24, 2017 at 07:24:08PM +0100, Levente Uzonyi wrote:
Hi All,
as you may know, squeaksource.com is down. I had a look and found that
- the image had fallen into a GC cycle consuming 100% cpu and about 500 MB
o memory
- restarting the image didn't help with that
- I added the -mmap 1024m parameter to its arguments, and it stopped the
CPU grinding, and the memory usage has fallen below 500 MB since then. This also means that the VM's default memory limit is actually 500MB contrary to the 1GB stated by the VM.
- but, the image is still not able to start up properly. it's probably
stuck on something simple, but due to the lack of VNC, I just couldn't check what was wrong
Levente
Sorry I was away, I just restarted it.
What I did:
- download the image, run it locally
- close the debuggers
- open process browser, find a process stuck in trying to send mail, terminate it
- check memory size, verify that the web page opens on my localhost
- copy it back to dan.box.squeak.org
- kill the running image and let it restart with fixed image
VNC and mail from the squeaksource image do not work any more. The VNC used to work on dan.box.squeak.org but not now, possibly something changed in network configuration that the ancient image does not handle right. The stuck process that caused the failure seems to fail on an error that tries to send mail to inform someone of the problem, which fails because smtp does not work because of network address resolution, which fails and throws an error that tries to send more mail, etc.
As Chris said, the image is overdue for an update.
The service was off line but no data was loss, and as in previous incidents the data is safe on disk.
Dave
On Fri, Mar 24, 2017 at 01:57:25PM -0500, Chris Muller wrote:
A few months ago, we lost weeks of user updates due to image persistence saving itself as an unlaunchable image (> 1GB interpreter).
Based on the ss.log, there were only a few updates lost, and they appeared to be junk commits from some rather questionable origins. I asked on the Squeak and Pharo lists and heard of no issues. I am fairly confident that there was no loss.
All the latest code for SqueakSource is ready and available at http://source.squeak.org/ss. This is the codebase that has been supporting source.squeak.org for several months now. It's stable.
Squeaksource is important. Is there any reason not to update the image and code ASAP?
I cannot put much time into it now, and I don't mind if you want to take it over as your schedule permits.
If I do the updates, I want to get it to a 64 bit image/vm.
Dave
On Fri, Mar 24, 2017 at 07:44:35PM -0400, David T. Lewis wrote:
VNC and mail from the squeaksource image do not work any more. The VNC used to work on dan.box.squeak.org but not now, possibly something changed in network configuration that the ancient image does not handle right. The stuck process that caused the failure seems to fail on an error that tries to send mail to inform someone of the problem, which fails because smtp does not work because of network address resolution, which fails and throws an error that tries to send more mail, etc.
I am somewhat embarassed to admit that I just fixed the "broken VNC" problem by clicking the "enable remote connections" checkbox in the RFBServer. Apparently I accidentally left it disabled the last time I restarted the image.
VNC access to the squeaksource.com image works again, as explained in dan.box.squeak.org:/srv/squeaksourcecom/README. Log in to dan.box.squeak.org, send SIGUSR2 to the VM process, then connect from your VNC localhost:1 connection within 60 seconds of the SIGUSR2 (after setting up the port forwarding to dan). Use the VNC password that is mentioned in the README.
This is probably way too many layers of security now that we have the port forwarding set up on Rackspace, so maybe I should get rid of that requirement to send a SIGUSR2 to enable RFBServer. Leaving VNC running all the time would probably be fine. But maybe later, I will not change it now.
Dave
Hi Dave,
it should work without enabling remote connections if you create an ssh tunnel to dan instead of ian.
Levente
On Sat, 25 Mar 2017, David T. Lewis wrote:
On Fri, Mar 24, 2017 at 07:44:35PM -0400, David T. Lewis wrote:
VNC and mail from the squeaksource image do not work any more. The VNC used to work on dan.box.squeak.org but not now, possibly something changed in network configuration that the ancient image does not handle right. The stuck process that caused the failure seems to fail on an error that tries to send mail to inform someone of the problem, which fails because smtp does not work because of network address resolution, which fails and throws an error that tries to send more mail, etc.
I am somewhat embarassed to admit that I just fixed the "broken VNC" problem by clicking the "enable remote connections" checkbox in the RFBServer. Apparently I accidentally left it disabled the last time I restarted the image.
VNC access to the squeaksource.com image works again, as explained in dan.box.squeak.org:/srv/squeaksourcecom/README. Log in to dan.box.squeak.org, send SIGUSR2 to the VM process, then connect from your VNC localhost:1 connection within 60 seconds of the SIGUSR2 (after setting up the port forwarding to dan). Use the VNC password that is mentioned in the README.
This is probably way too many layers of security now that we have the port forwarding set up on Rackspace, so maybe I should get rid of that requirement to send a SIGUSR2 to enable RFBServer. Leaving VNC running all the time would probably be fine. But maybe later, I will not change it now.
Dave
On Sun, Mar 26, 2017 at 02:09:45PM +0200, Levente Uzonyi wrote:
Hi Dave,
it should work without enabling remote connections if you create an ssh tunnel to dan instead of ian.
Levente
I think that is what I am doing. Just to be clear, I do this:
lewis@lewis-Gazelle-Pro:~$ ssh -L5901:localhost:5901 dan.box.squeak.org
The programs included with the Debian GNU/Linux system are free software; the exact distribution terms for each program are described in the individual files in /usr/share/doc/*/copyright.
Debian GNU/Linux comes with ABSOLUTELY NO WARRANTY, to the extent permitted by applicable law. Last login: Sun Mar 26 14:19:37 2017 from 10.208.225.29 davidlewis@dan:~$ sudo su - squeaksourcecom [sudo] password for davidlewis: squeaksourcecom@dan:~$ pgrep -l squeakvm 9388 squeakvm squeaksourcecom@dan:~$ kill -SIGUSR2 9388 squeaksourcecom@dan:~$ # keep this connection open until the VMC connection is made squeaksourcecom@dan:~$ exit
While the ssh connection is active, I use a VNC client to connect to localhost:1 which is port 5901 forwarded to dan.box.squeak.org, and then I enter the password for the VNC connection.
Dave
On Sat, 25 Mar 2017, David T. Lewis wrote:
On Fri, Mar 24, 2017 at 07:44:35PM -0400, David T. Lewis wrote:
VNC and mail from the squeaksource image do not work any more. The VNC used to work on dan.box.squeak.org but not now, possibly something changed in network configuration that the ancient image does not handle right. The stuck process that caused the failure seems to fail on an error that tries to send mail to inform someone of the problem, which fails because smtp does not work because of network address resolution, which fails and throws an error that tries to send more mail, etc.
I am somewhat embarassed to admit that I just fixed the "broken VNC" problem by clicking the "enable remote connections" checkbox in the RFBServer. Apparently I accidentally left it disabled the last time I restarted the image.
VNC access to the squeaksource.com image works again, as explained in dan.box.squeak.org:/srv/squeaksourcecom/README. Log in to dan.box.squeak.org, send SIGUSR2 to the VM process, then connect from your VNC localhost:1 connection within 60 seconds of the SIGUSR2 (after setting up the port forwarding to dan). Use the VNC password that is mentioned in the README.
This is probably way too many layers of security now that we have the port forwarding set up on Rackspace, so maybe I should get rid of that requirement to send a SIGUSR2 to enable RFBServer. Leaving VNC running all the time would probably be fine. But maybe later, I will not change it now.
Dave
Unless localhost is mapped to anything other than 127.0.0.1, that should work without allowing remote connetions. I always use 127.0.0.1 instead of localhost, because the image will always accept that.
Levente
On Sun, 26 Mar 2017, David T. Lewis wrote:
On Sun, Mar 26, 2017 at 02:09:45PM +0200, Levente Uzonyi wrote:
Hi Dave,
it should work without enabling remote connections if you create an ssh tunnel to dan instead of ian.
Levente
I think that is what I am doing. Just to be clear, I do this:
lewis@lewis-Gazelle-Pro:~$ ssh -L5901:localhost:5901 dan.box.squeak.org
The programs included with the Debian GNU/Linux system are free software; the exact distribution terms for each program are described in the individual files in /usr/share/doc/*/copyright.
Debian GNU/Linux comes with ABSOLUTELY NO WARRANTY, to the extent permitted by applicable law. Last login: Sun Mar 26 14:19:37 2017 from 10.208.225.29 davidlewis@dan:~$ sudo su - squeaksourcecom [sudo] password for davidlewis: squeaksourcecom@dan:~$ pgrep -l squeakvm 9388 squeakvm squeaksourcecom@dan:~$ kill -SIGUSR2 9388 squeaksourcecom@dan:~$ # keep this connection open until the VMC connection is made squeaksourcecom@dan:~$ exit
While the ssh connection is active, I use a VNC client to connect to localhost:1 which is port 5901 forwarded to dan.box.squeak.org, and then I enter the password for the VNC connection.
Dave
On Sat, 25 Mar 2017, David T. Lewis wrote:
On Fri, Mar 24, 2017 at 07:44:35PM -0400, David T. Lewis wrote:
VNC and mail from the squeaksource image do not work any more. The VNC used to work on dan.box.squeak.org but not now, possibly something changed in network configuration that the ancient image does not handle right. The stuck process that caused the failure seems to fail on an error that tries to send mail to inform someone of the problem, which fails because smtp does not work because of network address resolution, which fails and throws an error that tries to send more mail, etc.
I am somewhat embarassed to admit that I just fixed the "broken VNC" problem by clicking the "enable remote connections" checkbox in the RFBServer. Apparently I accidentally left it disabled the last time I restarted the image.
VNC access to the squeaksource.com image works again, as explained in dan.box.squeak.org:/srv/squeaksourcecom/README. Log in to dan.box.squeak.org, send SIGUSR2 to the VM process, then connect from your VNC localhost:1 connection within 60 seconds of the SIGUSR2 (after setting up the port forwarding to dan). Use the VNC password that is mentioned in the README.
This is probably way too many layers of security now that we have the port forwarding set up on Rackspace, so maybe I should get rid of that requirement to send a SIGUSR2 to enable RFBServer. Leaving VNC running all the time would probably be fine. But maybe later, I will not change it now.
Dave
Now that VNC is working again, I connected to the image to check on it.
From the image, we have this:
NetNameResolver localHostAddress ==> #[104 130 170 38]
I suspect that this will be somehow related to the smtp problem. I left a workspace open in the image with some notes about this, so hopefully when it happens again we will be able to connect to the image and see what is wrong.
Dave
On Sun, Mar 26, 2017 at 06:08:25PM +0200, Levente Uzonyi wrote:
Unless localhost is mapped to anything other than 127.0.0.1, that should work without allowing remote connetions. I always use 127.0.0.1 instead of localhost, because the image will always accept that.
Levente
On Sun, 26 Mar 2017, David T. Lewis wrote:
On Sun, Mar 26, 2017 at 02:09:45PM +0200, Levente Uzonyi wrote:
Hi Dave,
it should work without enabling remote connections if you create an ssh tunnel to dan instead of ian.
Levente
I think that is what I am doing. Just to be clear, I do this:
lewis@lewis-Gazelle-Pro:~$ ssh -L5901:localhost:5901 dan.box.squeak.org
The programs included with the Debian GNU/Linux system are free software; the exact distribution terms for each program are described in the individual files in /usr/share/doc/*/copyright.
Debian GNU/Linux comes with ABSOLUTELY NO WARRANTY, to the extent permitted by applicable law. Last login: Sun Mar 26 14:19:37 2017 from 10.208.225.29 davidlewis@dan:~$ sudo su - squeaksourcecom [sudo] password for davidlewis: squeaksourcecom@dan:~$ pgrep -l squeakvm 9388 squeakvm squeaksourcecom@dan:~$ kill -SIGUSR2 9388 squeaksourcecom@dan:~$ # keep this connection open until the VMC connection is made squeaksourcecom@dan:~$ exit
While the ssh connection is active, I use a VNC client to connect to localhost:1 which is port 5901 forwarded to dan.box.squeak.org, and then I enter the password for the VNC connection.
Dave
On Sat, 25 Mar 2017, David T. Lewis wrote:
On Fri, Mar 24, 2017 at 07:44:35PM -0400, David T. Lewis wrote:
VNC and mail from the squeaksource image do not work any more. The VNC used to work on dan.box.squeak.org but not now, possibly something changed in network configuration that the ancient image does not handle right. The stuck process that caused the failure seems to fail on an error that tries to send mail to inform someone of the problem, which fails because smtp does not work because of network address resolution, which fails and throws an error that tries to send more mail, etc.
I am somewhat embarassed to admit that I just fixed the "broken VNC" problem by clicking the "enable remote connections" checkbox in the RFBServer. Apparently I accidentally left it disabled the last time I restarted the image.
VNC access to the squeaksource.com image works again, as explained in dan.box.squeak.org:/srv/squeaksourcecom/README. Log in to dan.box.squeak.org, send SIGUSR2 to the VM process, then connect from your VNC localhost:1 connection within 60 seconds of the SIGUSR2 (after setting up the port forwarding to dan). Use the VNC password that is mentioned in the README.
This is probably way too many layers of security now that we have the port forwarding set up on Rackspace, so maybe I should get rid of that requirement to send a SIGUSR2 to enable RFBServer. Leaving VNC running all the time would probably be fine. But maybe later, I will not change it now.
Dave
On Fri, Mar 24, 2017 at 01:57:25PM -0500, Chris Muller wrote:
Squeaksource is important. Is there any reason not to update the image and code ASAP?
Hi Chris,
I have not had much time to look at this, but can you say anything about what is required to read the serialized repository from the old image into an up to date SqueakSource image? I think I remember you saying that you had used Magma serializer to do this, but I don't know what issues may have been involved.
I have tried generating a new data.obj from the current squeaksource.com image, then loading it into Squeak 4.6 or a 32-bit Squeak 5.0, but ReferenceStream deserializing is failing fairly early in the load. The data.obj is fine, because I can reload it into the original image. But I am not able to load it into either the V3 or Spur 32 bit images with newer SqueakSource package, so I am suspecting that this may be something to do with class changes in the SqueakSource package. I can probably figure it out, but I thought I should ask you first before I spend much time debugging. Does any of this ring a bell for you?
Thanks! Dave
Hi Dave,
Yes, because of the format change I used Ma object serializer, because it is mostly agnostic to the squeak version. I've attached version of Ma Object Serializer of that time which can run correctly in that image. After loading that SAR, you can export the object with something like this:
MaObjectSerializer new fileOut: SSRepository current storage toFileNamed: 'data.obj.ma' in: FileDirectory default.
Now, you can go into a 5.1 / recent image and use Ma object serializer to load it. Unfortunately, there was a minor file-format change since that 2009 version, so you have to use that same version to load it initially. Load the same SAR into the 5.1 image to iniitally load the object.
MaObjectSerializer fileIn: '/the/fully/qualified/filename' "inspect it?"
Now you can finally load Ma Serializer 1.6 from SqueakMap, and export the object one last time, this time in the "new" Ma format.
MaObjectSerializer new fileOut: self toFileNamed: 'data.obj' in: FileDirectory default
It should now be ready to use with the latest code at source.squeak.org/ss. You can test with just the data.obj file you don't need any of the MCZ's.
When the time comes, we should run the new server with just good ol'e SSFilesystem repository. No image persistence and no Magma. The save time is slow, but it runs in the background and if new commits come in in the meantime, it'll just save again after the current save is done. It keeps the last 35 or so generations of backup obj files.
- Chris
On Thu, Mar 30, 2017 at 6:01 PM, David T. Lewis lewis@mail.msen.com wrote:
On Fri, Mar 24, 2017 at 01:57:25PM -0500, Chris Muller wrote:
Squeaksource is important. Is there any reason not to update the image and code ASAP?
Hi Chris,
I have not had much time to look at this, but can you say anything about what is required to read the serialized repository from the old image into an up to date SqueakSource image? I think I remember you saying that you had used Magma serializer to do this, but I don't know what issues may have been involved.
I have tried generating a new data.obj from the current squeaksource.com image, then loading it into Squeak 4.6 or a 32-bit Squeak 5.0, but ReferenceStream deserializing is failing fairly early in the load. The data.obj is fine, because I can reload it into the original image. But I am not able to load it into either the V3 or Spur 32 bit images with newer SqueakSource package, so I am suspecting that this may be something to do with class changes in the SqueakSource package. I can probably figure it out, but I thought I should ask you first before I spend much time debugging. Does any of this ring a bell for you?
Thanks! Dave
Thanks Chris, I'll give it a try.
Dave
On Thu, Mar 30, 2017 at 08:37:13PM -0500, Chris Muller wrote:
Hi Dave,
Yes, because of the format change I used Ma object serializer, because it is mostly agnostic to the squeak version. I've attached version of Ma Object Serializer of that time which can run correctly in that image. After loading that SAR, you can export the object with something like this:
MaObjectSerializer new fileOut: SSRepository current storage toFileNamed: 'data.obj.ma' in: FileDirectory default.
Now, you can go into a 5.1 / recent image and use Ma object serializer to load it. Unfortunately, there was a minor file-format change since that 2009 version, so you have to use that same version to load it initially. Load the same SAR into the 5.1 image to iniitally load the object.
MaObjectSerializer fileIn: '/the/fully/qualified/filename' "inspect it?"
Now you can finally load Ma Serializer 1.6 from SqueakMap, and export the object one last time, this time in the "new" Ma format.
MaObjectSerializer new fileOut: self toFileNamed: 'data.obj' in: FileDirectory default
It should now be ready to use with the latest code at source.squeak.org/ss. You can test with just the data.obj file you don't need any of the MCZ's.
When the time comes, we should run the new server with just good ol'e SSFilesystem repository. No image persistence and no Magma. The save time is slow, but it runs in the background and if new commits come in in the meantime, it'll just save again after the current save is done. It keeps the last 35 or so generations of backup obj files.
- Chris
On Thu, Mar 30, 2017 at 6:01 PM, David T. Lewis lewis@mail.msen.com wrote:
On Fri, Mar 24, 2017 at 01:57:25PM -0500, Chris Muller wrote:
Squeaksource is important. Is there any reason not to update the image and code ASAP?
Hi Chris,
I have not had much time to look at this, but can you say anything about what is required to read the serialized repository from the old image into an up to date SqueakSource image? I think I remember you saying that you had used Magma serializer to do this, but I don't know what issues may have been involved.
I have tried generating a new data.obj from the current squeaksource.com image, then loading it into Squeak 4.6 or a 32-bit Squeak 5.0, but ReferenceStream deserializing is failing fairly early in the load. The data.obj is fine, because I can reload it into the original image. But I am not able to load it into either the V3 or Spur 32 bit images with newer SqueakSource package, so I am suspecting that this may be something to do with class changes in the SqueakSource package. I can probably figure it out, but I thought I should ask you first before I spend much time debugging. Does any of this ring a bell for you?
Thanks! Dave
box-admins@lists.squeakfoundation.org