On Sun, Jun 29, 2014 at 7:14 PM, Chris Muller asqueaker@gmail.com wrote:
I was operating under the assumption that CC was referring to a Spur VM. I mistakenly assumed the higher number automatically meant Spur, but you've made me realize it's just a new Cog VM, not Spur. Yes, Cog has been quite stable for some time and you know the details and the risks. We know your ethic and we trust you.
except that you're right :-(. I've found that my recent changes to eliminate implementation of division with shift when the divisor is a power of two breaks the LargeIntegers plugin because division is the only way that Slang supports signed bitShift: (which is a bug). Phhh...
So I'll be taking 3021 down PDQ. 3018 should be OK.
Having said that, I would still understand and support any box-admin who wanted some testing first. If something is already _proven_ working in production, not crashing, and we're not really having any immediate production impetus to upgrade it, then this almost amounts to "testing in production". No benefit, but a small chance for a problem. There's plenty of time and opportunity to shake them out in test environments first.
On Sun, Jun 29, 2014 at 8:11 PM, Eliot Miranda eliot.miranda@gmail.com wrote:
On Sun, Jun 29, 2014 at 9:20 AM, Chris Muller asqueaker@gmail.com wrote:
Didn't Levente ask us to use stable VM's for production services?
Given that 3018 has nothing radically new but contains bug fixes IMO it is stable. Even the SPur VM is more stable than the one that preceeds it. What extra testing has 2776 had to make it more or less stable than 2987 or 3021? 2987 and subsequent all run the standard test suite with equivalent results. I don't make arbitrary changes to the Cog VMs (Spur is obviously evolving much faster, but again is getting more, not less stable). You need to specify a criterion by which a VM is judged stable, not merely go by age, which is, at least on average, inversely correlated with stability, older VMs containing more bugs, if you want to argue against using newer VMs. I don't put newer VMs out for amusement. I put them out because they're better (faster, fix bugs, etc).
On Fri, Jun 27, 2014 at 4:01 PM, Eliot Miranda eliot.miranda@gmail.com wrote:
Hi Chris,
On Fri, Jun 27, 2014 at 1:14 PM, Chris Cunnington brasspen@gmail.com wrote:
I intend to upgrade the VM on box4 VM.r2776 (Aug. 2013) to VM.r2987 (June 2014). The intention here is to refine the install/removal process. I just did this on my Debian Wheezy box, so it seems pretty straightforward.
3018 is the latest and greatest. Please use it.
Installing:
- get a Cog binary from mirandabanda.org
- use Ken's cogdeb.zip to make a deb
- install with dpkg (i.e. dpkg -i cogvm_2776-1_i386.deb)
- execute "man squeak" to see if it's loaded
Removing:
- check the package name with dpkg-query -l
- see it's actually called "coglinux" as opposed to cogvm or squeak
- dpkg -r coglinux
So, as I said, I just did this on my server and is seemed to work OK. I plan to upgrade the VM from last year's model (2776) to this year's (2987). If there's anything people with more LInux experience could add to this process, I'd be happy to hear it.
Chris
-- best, Eliot
-- best, Eliot
2014-06-30 19:09 GMT+02:00 Eliot Miranda eliot.miranda@gmail.com:
On Sun, Jun 29, 2014 at 7:14 PM, Chris Muller asqueaker@gmail.com wrote:
I was operating under the assumption that CC was referring to a Spur VM. I mistakenly assumed the higher number automatically meant Spur, but you've made me realize it's just a new Cog VM, not Spur. Yes, Cog has been quite stable for some time and you know the details and the risks. We know your ethic and we trust you.
except that you're right :-(. I've found that my recent changes to eliminate implementation of division with shift when the divisor is a power of two breaks the LargeIntegers plugin because division is the only way that Slang supports signed bitShift: (which is a bug). Phhh...
Which sounds completely weird because LargeIntegers are treated as unsigned!!! (Or sign-magnitude if you prefer).
Nicolas
So I'll be taking 3021 down PDQ. 3018 should be OK.
Having said that, I would still understand and support any box-admin who wanted some testing first. If something is already _proven_ working in production, not crashing, and we're not really having any immediate production impetus to upgrade it, then this almost amounts to "testing in production". No benefit, but a small chance for a problem. There's plenty of time and opportunity to shake them out in test environments first.
On Sun, Jun 29, 2014 at 8:11 PM, Eliot Miranda eliot.miranda@gmail.com wrote:
On Sun, Jun 29, 2014 at 9:20 AM, Chris Muller asqueaker@gmail.com wrote:
Didn't Levente ask us to use stable VM's for production services?
Given that 3018 has nothing radically new but contains bug fixes IMO it is stable. Even the SPur VM is more stable than the one that preceeds it. What extra testing has 2776 had to make it more or less stable than 2987 or 3021? 2987 and subsequent all run the standard test suite with equivalent results. I don't make arbitrary changes to the Cog VMs (Spur is obviously evolving much faster, but again is getting more, not less stable). You need to specify a criterion by which a VM is judged stable, not merely go by age, which is, at least on average, inversely correlated with stability, older VMs containing more bugs, if you want to argue against using newer VMs. I don't put newer VMs out for amusement. I put them out because they're better (faster, fix bugs, etc).
On Fri, Jun 27, 2014 at 4:01 PM, Eliot Miranda <eliot.miranda@gmail.com
wrote:
Hi Chris,
On Fri, Jun 27, 2014 at 1:14 PM, Chris Cunnington brasspen@gmail.com wrote:
I intend to upgrade the VM on box4 VM.r2776 (Aug. 2013) to VM.r2987 (June 2014). The intention here is to refine the install/removal process. I just did this on my Debian Wheezy box, so it seems pretty straightforward.
3018 is the latest and greatest. Please use it.
Installing:
- get a Cog binary from mirandabanda.org
- use Ken's cogdeb.zip to make a deb
- install with dpkg (i.e. dpkg -i cogvm_2776-1_i386.deb)
- execute "man squeak" to see if it's loaded
Removing:
- check the package name with dpkg-query -l
- see it's actually called "coglinux" as opposed to cogvm or squeak
- dpkg -r coglinux
So, as I said, I just did this on my server and is seemed to work OK. I plan to upgrade the VM from last year's model (2776) to this year's (2987). If there's anything people with more LInux experience could add to this process, I'd be happy to hear it.
Chris
-- best, Eliot
-- best, Eliot
-- best, Eliot
vm-dev@lists.squeakfoundation.org
-
Eliot Miranda -
Nicolas Cellier