Component Corruption Problem - instantly crashes aws

[TheSmitchell]

I have been battling corrupt components all day. I have a tabbed ui that - suddenly - started taking out the aws hard. I think it got corrupted. So I rolled back to an older version. But then some of my other grids starting producing the same error. ???

I was wondering if anyone else has encountered a similar problem? The aws quits immediately and shuts down. Doesn't even try to restart. It issues a message that it "stopped working" and the process is completely gone. No dump files. No errors in the error log - save for a report in the Windows Event Viewer:

Event ID: 1000
Level: Error
Task Category: (100)
Exception code: 0xc0000005
Fault offset: 0x0007d0be
Faulting Application: A5ApplicationServer.exe
Faulting module: A5ENGINE.dll

I'm running on windows 8. Current A5 version below. Maybe that has something to do with it? How does 12 stack up? I'll decidedly upgrade if it's more stable.

---
Sarah

[christappan]

Re: Component Corruption Problem - instantly crashes aws

No idea if it's related, but from the 6/6/13 v11 prerelease (here):

Application Server - SSL - Some users who are using SSL have reported server stability issues. We have found and fixed an issue with OpenSSL - the third party library used by Alpha Five to implement SSL - that we believe will address this issue.

How does 12 stack up? I'll decidedly upgrade if it's more stable.

Working with Alpha Five since version 9, I'll say that highly unstable builds get replaced pretty quickly, but almost every build I've used has seemed to have some kind of minor bugs in it. So I wouldn't put too much additional confidence in v12 over v11 since it's essentially just newer builds on top of the old.

[TheSmitchell]

Re: Component Corruption Problem - instantly crashes aws

Probably. Probably not. I'm not running with SSL on, so I don't see how that would potentially fix the issue - especially since I can just rollback the components and re-build them to the same "state" and they're fine. I think it's a problem with component compilation/generation for publishing. But, like I said, the logs are useless. They show activity up to the point of accessing the problem component and then nothing. No errors or anything-I have all the logs. I think I'll send a report into Alpha today - after generating a minimal set of logs - and see what they say. This crash is pretty epic.

[TheSmitchell]

Re: Component Corruption Problem - instantly crashes aws

Hmmm. I fixed my grid? I was pre-pending {grid.componentName} to my js functions defined in the javascript section. I decided that was no longer a good idea and that passing the {grid.object} or whatever I needed into the function was better --- no sense in generating tens of hundreds of the same function when I can just pass the data I need. So I refactored and now my grid is behaving properly. I don't know what caused the AWS to choke and die when this grid was requested, but at any rate...

This:

var {grid.componentName}_foo = function(stuff)
{
  {grid.object}.etc...
}

Now looks like this:

var foo = function(grid,stuff)
{
  grid.etc...
}

On a completely unrelated note (and because this grid has code to address this particular): $ is present in {grid.componentName} when grid names get long enough. $ make the jQuery go crazy! You try something like this:

var allATagsInNotes = jQuery('#{gridComponentName}\\.V\\.R'+rowNum+'\\.NOTES a');
allATagsInNotes.attr("target","_blank");

And it utterly crashes and fails seemingly at random. So you do a lot of debugging and find all these crazy $ in your {grid.componentName} strings! So, then you write code to escape all the crazy:

var validSelector = function(str)
{
	return str.replace(/([;&,\.\+\*\~':"\!\^#$%@\[\]\(\)=>\|])/g, '\\$1');

}

var allATagsInNotes = jQuery('#' + validSelector('{grid.ComponentName}.V.R' + rowNum + '.NOTES') + ' a'); 
allATagsInNotes.attr("target","_blank");

Then you pull your hair out some more later because this grid that gave you so much grief for the $ in the jQuery is now randomly corrupt for whatever reason. SO then you refactor your code - trying _anything_ to make it more better without rolling back 8 hours of work and magic sauce everything is fixed.

I don't know what you call this. Heisenbug?

Either way, I've mucked around enough to get whatever was broken to fix itself.

EDIT: I have reworded this because correlation does not imply causation. Original post indicated that using {grid.componentName} to make unique javascript function names caused the crash. I can't prove that. Claiming that was effectively a load of ...

[christappan]

Re: Component Corruption Problem - instantly crashes aws

I'm quite far behind you in terms of javascript and jQuery knowledge, but I'm curious about the function names with grid component names included. Do I understand correctly that you're suggesting it's a problem for "$$$$" to show up in the function name itself but if "$$$$" was in the "grid" variable of the function it would work correctly?

And is there a component name length at which the "$$$$" show up?

[TheSmitchell]

Re: Component Corruption Problem - instantly crashes aws

Oh. Oh yes. $ shows up in component names. Nested. Nested deep. And then your jQueries explode.

It's not necessarily that $ is not permitted in a javascript function name. Quite the opposite. You can totally put them in function names. But the generated function names were rather long... And my corrupted grid got fixed after I modified them to take a grid variable instead of generating a unique name for each grid instance using the {grid.componentName}. And {grid.componentName} makes me nervous. And I think that approach was generally a bad idea to begin with. Passing the grid makes it mega easier - I could even just put my code in a separate js file that way and there's less code the server has to generate before the data is served up to the client.

[christappan]

Re: Component Corruption Problem - instantly crashes aws

I guess my question is this: In the code

var foo = function(grid,stuff)
{
  grid.etc...
}

are you setting

var grid = {grid.ComponentName}

? If so, wouldn't you just be passing a variable with the same $$$$ problems? And if not, where is "grid" getting its value?

[TheSmitchell]

Re: Component Corruption Problem - instantly crashes aws

I guess my question is this: In the code

var foo = function(grid,stuff)
{
  grid.etc...
}

are you setting

var grid = {grid.ComponentName}

? If so, wouldn't you just be passing a variable with the same $$$$ problems? And if not, where is "grid" getting its value?

No. I'm passing the grid object. Figured that would be noticed in my first code snippet?

var {grid.componentName}_foo = function(stuff)
{
  {grid.object}.etc...
}

[...] So I've updated my js to look more like this:

var foo = function(grid,stuff)
{
  grid.etc...
}

[christappan]

Re: Component Corruption Problem - instantly crashes aws

I did indeed miss that. Sorry. (and thanks btw for all the explaining). So is {grid.object} safe(r) from the added "$$$$"?

[TheSmitchell]

Re: Component Corruption Problem - instantly crashes aws

If Alpha was generating a {grid.object} variable with a malformed variable name, you'd have other problems well before any client-side code you wrote was executed.

It's not that {grid.object} is safer from the added $, it's that my function names aren't being generated by the AWS before being sent to the client. And that using the parameter/variable "grid" as an alias for {grid.object} achieves the same result - since I'm requiring that the {grid.object} be passed to the function on call rather than having to figure out how to force Alpha to create a unique function object for each instance of my grid.

I don't know why effectively "cleaning up" my javascript fixed my crash. It's a mystery. But it did. Most likely what happened was I made enough changes in a number of locations that whatever thing - some control? a linked grid? even maybe the javascript code section itself - that was causing my grid to insta-crash the AWS went away.

Here's an image for you:
alphaNames.png

It doesn't take long for me to find these. The grid object name contains just as many $ as anything else. Something to keep in mind if you ever get into using the jQuery, though. Because $ is a special character in jQuery selector strings. So you have to escape them.

Looking back, my post with the jQuery bits was more me rambling on about something completely unrelated and also my grid is working again and I don't know why. Also that passing your {grid.object} or {grid.componentName} to your functions is probably better/easier to think about than trying to figure out how to make unique functions per grid instance. /end more rambling

[TheSmitchell]

Re: Component Corruption Problem - instantly crashes aws

I have solved my problem.

It's an FTP upload problem. If I try to publish a change to a component while the AWS is under load, sometimes the FTP process doesn't copy the file up cleanly. Then the AWS crashes. I've never had this problem before. If I publish while the AWS isn't running or when nobody is using the system, then uploads are fine. No component is safe.