[Pharo-project] random crashes... help needed (Pharo, Magma, Seaside... I dunno)
Hi, I'm having random crashes in a production image. This crashes occurs without leaving any information available... it just crash. No PharoDebug.log, no dump, nothing... finally, yesterday I get a PharoDebug.log (which I'm attaching) with some error information... but frankly I don't know if this is the reason the image is crashing or some other random error. Some other elements I'm seeing is: - Seaside sessions are in fact never removed unless I manually execute garbage collect. - WeakArray FinalizationDependents is bigger than usual (90 instances), elements are mostly WeakIdentityDictionary and MaWeakIdentityDictionary (most of them empty but some other with sizes from 700 to 4000). I'm using a Pharo 1.1.1 with Seaside 3.0 and Magma 1.1r2 The VM is any VM (exuperi 0.15f or unix cog) . OS is "some kind of linux"... looks like a "custom suse". Well... and I'm completely lost. Any help will be very appreciated... even on ways on how to get more info of my crashing system (redirecting standard and error output does not works either)... And the access I have to the O.S. is also limited. Cheers, Esteban
On 27 December 2010 23:49, Esteban Lorenzano <estebanlm@gmail.com> wrote:
Hi, I'm having random crashes in a production image. This crashes occurs without leaving any information available... it just crash. No PharoDebug.log, no dump, nothing... finally, yesterday I get a PharoDebug.log (which I'm attaching) with some error information... but frankly I don't know if this is the reason the image is crashing or some other random error. Some other elements I'm seeing is:
- Seaside sessions are in fact never removed unless I manually execute garbage collect. - WeakArray FinalizationDependents is bigger than usual (90 instances), Â elements are mostly WeakIdentityDictionary and MaWeakIdentityDictionary (most of them empty but some other with sizes from 700 to 4000).
I'm using a Pharo 1.1.1 with Seaside 3.0 and Magma 1.1r2 The VM is any VM (exuperi 0.15f or unix cog) . OS is "some kind of linux"... looks like a "custom suse".
Okay, you're using oscog.35 which predates new finalization.. at least it is not in a play.
Well... and I'm completely lost. Any help will be very appreciated... even on ways on how to get more info of my crashing system (redirecting standard and error output does not works either)... And the access I have to the O.S. is also limited.
well, all i can say that you should start digging from here: 'Error: highBit is not defined for negative integers' what if some error message prints numbers and using highbit for on negative one? it will then loop forever, recursively until there is a no free space left. but that of course could be anything else.
Cheers, Esteban
-- Best regards, Igor Stasenko AKA sig.
On Mon, Dec 27, 2010 at 8:20 PM, Igor Stasenko <siguctua@gmail.com> wrote:
On 27 December 2010 23:49, Esteban Lorenzano <estebanlm@gmail.com> wrote:
Hi, I'm having random crashes in a production image. This crashes occurs without leaving any information available... it just crash. No PharoDebug.log, no dump, nothing... finally, yesterday I get a PharoDebug.log (which I'm attaching) with some error information... but frankly I don't know if this is the reason the image is crashing or some other random error. Some other elements I'm seeing is:
- Seaside sessions are in fact never removed unless I manually execute garbage collect. - WeakArray FinalizationDependents is bigger than usual (90 instances), elements are mostly WeakIdentityDictionary and MaWeakIdentityDictionary (most of them empty but some other with sizes from 700 to 4000).
I'm using a Pharo 1.1.1 with Seaside 3.0 and Magma 1.1r2 The VM is any VM (exuperi 0.15f or unix cog) . OS is "some kind of linux"... looks like a "custom suse".
Okay, you're using oscog.35 which predates new finalization.. at least it is not in a play.
Well... and I'm completely lost. Any help will be very appreciated... even on ways on how to get more info of my crashing system (redirecting standard and error output does not works either)... And the access I have to the O.S. is also limited.
well, all i can say that you should start digging from here:
'Error: highBit is not defined for negative integers'
what if some error message prints numbers and using highbit for on negative one? it will then loop forever, recursively until there is a no free space left. but that of course could be anything else.
That should not be fine :(. I don't like infinite recursion to be uncontrolled... Can we control it?
Cheers, Esteban
-- Best regards, Igor Stasenko AKA sig.
Oh, I just attached the error I could get... but this is also happening on exuperi vm 0.15f. There are a new cog version I can try? El 27/12/2010, a las 8:20p.m., Igor Stasenko escribió:
On 27 December 2010 23:49, Esteban Lorenzano <estebanlm@gmail.com> wrote:
Hi, I'm having random crashes in a production image. This crashes occurs without leaving any information available... it just crash. No PharoDebug.log, no dump, nothing... finally, yesterday I get a PharoDebug.log (which I'm attaching) with some error information... but frankly I don't know if this is the reason the image is crashing or some other random error. Some other elements I'm seeing is:
- Seaside sessions are in fact never removed unless I manually execute garbage collect. - WeakArray FinalizationDependents is bigger than usual (90 instances), elements are mostly WeakIdentityDictionary and MaWeakIdentityDictionary (most of them empty but some other with sizes from 700 to 4000).
I'm using a Pharo 1.1.1 with Seaside 3.0 and Magma 1.1r2 The VM is any VM (exuperi 0.15f or unix cog) . OS is "some kind of linux"... looks like a "custom suse".
Okay, you're using oscog.35 which predates new finalization.. at least it is not in a play.
Well... and I'm completely lost. Any help will be very appreciated... even on ways on how to get more info of my crashing system (redirecting standard and error output does not works either)... And the access I have to the O.S. is also limited.
well, all i can say that you should start digging from here:
'Error: highBit is not defined for negative integers'
what if some error message prints numbers and using highbit for on negative one? it will then loop forever, recursively until there is a no free space left. but that of course could be anything else.
Cheers, Esteban
-- Best regards, Igor Stasenko AKA sig.
On 27.12.2010 23:49, Esteban Lorenzano wrote:
Hi, I'm having random crashes in a production image. This crashes occurs without leaving any information available... it just crash. No PharoDebug.log, no dump, nothing... finally, yesterday I get a PharoDebug.log (which I'm attaching) with some error information... but frankly I don't know if this is the reason the image is crashing or some other random error. Some other elements I'm seeing is:
- Seaside sessions are in fact never removed unless I manually execute garbage collect.
No objects are removed in Smalltalk unless a garbage collection happens. There was an issue [1] where you had to access sessions and not only create them so that they get expired. [1] http://code.google.com/p/seaside/issues/detail?id=611 Cheers Philippe
Hi Esteban, there are a couple of Magma-related items you might want to check: - Is it possible a proxy is being passed as an argument to a primitive? I've only ever seen this problem manifest as a primitive failure, but maybe with the right primitive it could cause what you're seeing..? If it is, send #yourself to the argument in the primitive sender. - Do you have BlockContexts or BlockClosures in your persistent model? Check your repository definition's 'classDefinitions' dictionary, if they're in there, remember that only simple blocks are supported, e.g., only reference the block args, like in a sortBlock. On Mon, Dec 27, 2010 at 4:49 PM, Esteban Lorenzano <estebanlm@gmail.com> wrote:
Hi, I'm having random crashes in a production image. This crashes occurs without leaving any information available... it just crash. No PharoDebug.log, no dump, nothing... finally, yesterday I get a PharoDebug.log (which I'm attaching) with some error information... but frankly I don't know if this is the reason the image is crashing or some other random error. Some other elements I'm seeing is:
- Seaside sessions are in fact never removed unless I manually execute garbage collect. - WeakArray FinalizationDependents is bigger than usual (90 instances), Â elements are mostly WeakIdentityDictionary and MaWeakIdentityDictionary (most of them empty but some other with sizes from 700 to 4000).
I'm using a Pharo 1.1.1 with Seaside 3.0 and Magma 1.1r2 The VM is any VM (exuperi 0.15f or unix cog) . OS is "some kind of linux"... looks like a "custom suse".
Well... and I'm completely lost. Any help will be very appreciated... even on ways on how to get more info of my crashing system (redirecting standard and error output does not works either)... And the access I have to the O.S. is also limited.
Cheers, Esteban
Hi, No, I shouldn't have any of those problems... but I get another PharoDebug.log with some more information... something sorting things on WeakArray. Well, I'm attaching the log, maybe this is more helpful than the other one. Esteban El 28/12/2010, a las 4:40p.m., Chris Muller escribió:
Hi Esteban, there are a couple of Magma-related items you might want to check:
- Is it possible a proxy is being passed as an argument to a primitive? I've only ever seen this problem manifest as a primitive failure, but maybe with the right primitive it could cause what you're seeing..? If it is, send #yourself to the argument in the primitive sender.
- Do you have BlockContexts or BlockClosures in your persistent model? Check your repository definition's 'classDefinitions' dictionary, if they're in there, remember that only simple blocks are supported, e.g., only reference the block args, like in a sortBlock.
On Mon, Dec 27, 2010 at 4:49 PM, Esteban Lorenzano <estebanlm@gmail.com> wrote:
Hi, I'm having random crashes in a production image. This crashes occurs without leaving any information available... it just crash. No PharoDebug.log, no dump, nothing... finally, yesterday I get a PharoDebug.log (which I'm attaching) with some error information... but frankly I don't know if this is the reason the image is crashing or some other random error. Some other elements I'm seeing is:
- Seaside sessions are in fact never removed unless I manually execute garbage collect. - WeakArray FinalizationDependents is bigger than usual (90 instances), elements are mostly WeakIdentityDictionary and MaWeakIdentityDictionary (most of them empty but some other with sizes from 700 to 4000).
I'm using a Pharo 1.1.1 with Seaside 3.0 and Magma 1.1r2 The VM is any VM (exuperi 0.15f or unix cog) . OS is "some kind of linux"... looks like a "custom suse".
Well... and I'm completely lost. Any help will be very appreciated... even on ways on how to get more info of my crashing system (redirecting standard and error output does not works either)... And the access I have to the O.S. is also limited.
Cheers, Esteban
for your info this is on the todo of igor for 1.3 to dive and fix weak structure and finalization well. Stef On Dec 29, 2010, at 1:59 AM, Esteban Lorenzano wrote:
Hi, No, I shouldn't have any of those problems... but I get another PharoDebug.log with some more information... something sorting things on WeakArray. Well, I'm attaching the log, maybe this is more helpful than the other one.
Esteban
<PharoDebug.log>
El 28/12/2010, a las 4:40p.m., Chris Muller escribió:
Hi Esteban, there are a couple of Magma-related items you might want to check:
- Is it possible a proxy is being passed as an argument to a primitive? I've only ever seen this problem manifest as a primitive failure, but maybe with the right primitive it could cause what you're seeing..? If it is, send #yourself to the argument in the primitive sender.
- Do you have BlockContexts or BlockClosures in your persistent model? Check your repository definition's 'classDefinitions' dictionary, if they're in there, remember that only simple blocks are supported, e.g., only reference the block args, like in a sortBlock.
On Mon, Dec 27, 2010 at 4:49 PM, Esteban Lorenzano <estebanlm@gmail.com> wrote:
Hi, I'm having random crashes in a production image. This crashes occurs without leaving any information available... it just crash. No PharoDebug.log, no dump, nothing... finally, yesterday I get a PharoDebug.log (which I'm attaching) with some error information... but frankly I don't know if this is the reason the image is crashing or some other random error. Some other elements I'm seeing is:
- Seaside sessions are in fact never removed unless I manually execute garbage collect. - WeakArray FinalizationDependents is bigger than usual (90 instances), elements are mostly WeakIdentityDictionary and MaWeakIdentityDictionary (most of them empty but some other with sizes from 700 to 4000).
I'm using a Pharo 1.1.1 with Seaside 3.0 and Magma 1.1r2 The VM is any VM (exuperi 0.15f or unix cog) . OS is "some kind of linux"... looks like a "custom suse".
Well... and I'm completely lost. Any help will be very appreciated... even on ways on how to get more info of my crashing system (redirecting standard and error output does not works either)... And the access I have to the O.S. is also limited.
Cheers, Esteban
participants (6)
-
Chris Muller -
Esteban Lorenzano -
Guillermo Polito -
Igor Stasenko -
Philippe Marschall -
Stéphane Ducasse