Author Topic: Sequence crashes in 4.00.35  (Read 2800 times)

Offline gerry

  • Hero Member
  • *****
  • Posts: 695
    • View Profile
Sequence crashes in 4.00.35
« on: May 26, 2015, 11:38:57 PM »

Attached is the debug report.

Open and saved a few times without doing anything. Okay.
Added an effect . Saved . Okay
Added an effect . Saved . Okay
Moved two models up the sequencer view . Saved . okay
Added a new effect (on the vertical group) . Saved . okay
Added a new effect (on the vertical group) . Changed colours of the effect. Saved . Crash

Added to github as well and cross refereced.
Gerry

Offline gerry

  • Hero Member
  • *****
  • Posts: 695
    • View Profile
Re: Sequence crashes in 4.00.35
« Reply #1 on: May 27, 2015, 03:17:28 AM »
Attached is another debug report.

I opened an existing sequence , and pressed Save 3 or 4 times without doing anything at all, other than dhift the same effect a bit.

Stopped when it crashed. Watched the messages . the last one is that the fseq file has been created successfully in nn.nn seconds. Then  a couple of secs later it gracefully crashed.

Most of the time that it crashes , is after a save . Having said that , I am saving it every 5 minutes ie after dropping every 2 -3 effects that I wish to keep.

Not sure if it has any bearing , but I normally close the Perspectives and Effects Windows and move the Model window partially over the sequencer window.

Regards


« Last Edit: May 27, 2015, 04:36:14 AM by gerry »
Gerry

Offline gerry

  • Hero Member
  • *****
  • Posts: 695
    • View Profile
Re: Sequence crashes in 4.00.35
« Reply #2 on: May 27, 2015, 05:25:51 PM »
Please find attached another two debug report.

In both cases the crash is immediately after a save and after the message is displayed that the fseq file has been created in nn.nn secs.

So IMHO , either the creation of the fseq is failing at the end (but the message indicating success displays first), or something is failing in a refresh / repaint after the save. 

I may add that in this seq , the singing faces have been imported from a LOR LMS purchased file using the new Import effects function  (and looks great in preview) , but the rest if the effects (not many) have been added in xlights.
Gerry

Offline mikejk67

  • Newbie
  • *
  • Posts: 12
    • View Profile
Re: Sequence crashes in 4.00.35
« Reply #3 on: May 27, 2015, 06:50:13 PM »
In my case, sequences that worked fine (adding effects, saving, loading) in 4.0.33 are no longer working in 4.0.34 or 4.0.35.

Offline gerry

  • Hero Member
  • *****
  • Posts: 695
    • View Profile
Re: Sequence crashes in 4.00.35
« Reply #4 on: May 27, 2015, 08:10:39 PM »
In my case, sequences that worked fine (adding effects, saving, loading) in 4.0.33 are no longer working in 4.0.34 or 4.0.35.

Hello Mike,

Skip 4.0034 as there was an issue that meant it was replaced by 4.00.35 the same day.

For 4.00.35 - what is an example of a component that is not working  and do you get a crash or it is not working as expected ?
 
Gerry

Offline gerry

  • Hero Member
  • *****
  • Posts: 695
    • View Profile
Re: Sequence crashes in 4.00.35
« Reply #5 on: May 28, 2015, 12:11:50 AM »
Another couple of debug reports.

Both are just after the seq has been saved and the successful message has been displayed after fseq has been created.
One had no effects in it - had only added all the models  and then saved.

Is there any environmental variable/setup that could be causing this , that I could check?
 
Am using Windows 7.1 - Enterprise version.



[attachment deleted by admin]
« Last Edit: May 28, 2015, 04:59:48 AM by gerry »
Gerry

Offline Gilrock

  • Supporting Member
  • Hero Member
  • *
  • Posts: 6946
    • View Profile
Re: Sequence crashes in 4.00.35
« Reply #6 on: May 28, 2015, 08:12:26 AM »
Gerry,

A lot of people are dealing with crashes.  I've only been able to fix them if I can cause them to happen on my machine.  Most of your crashes all appear to be happening when multiple render threads are spawned to render the effects for saving the sequence.  Unfortunately nothing in your backtrace.txt files gives me a clue of whats wrong.  Dan tried to add more info to the backtrace files and I'm seeing the thread info showing up in your reports but it doesn't really tell me anything.  I'm not sure if Dan is able to figure anything out from that info.  But now the reports seem to be missing the actual functions where the crash started.  In your reports we were usually seeing RenderJob.o or Render.o and those aren't showing up at all now.

But do realize that the Save operation has 2 steps.  We save the XML file first and then we spawn the threads to render effects to create the fseq file.  So I don't think I've ever had a crash that didn't save my XML file.  What that means is that if you add a bunch of effects and then get a crash on Save it doesn't mean you lost that work.  Those effects should still have been saved in the XML file.  They just failed to render into the fseq file.  That allows you to restart and try the save again.

I'm sorry your getting crashes.  I get them also and it sucks.  I hate programs that do this.  Some of the frustations for me is a lot of issues come from code that we didn't even write.  We use so much wxWidgets code and sometimes we find issues where we have to try to code workarounds.  The other thing is we are using the multi-threading to speed up the rendering and that increases the risk of crashes especially on the Windows platform.  Dan runs on Mac and doesn't experience all these crashing problems.  I think some of these problems may be the wxWidgets threading implementation on Windows.

We could probably eliminate a lot of these crash on save issues if we made it single threaded but it could take 4 to 10 times longer to do the saves.  Our speed it what makes us better than the rest but if we can't fix the crashes it erases all that.

Offline gerry

  • Hero Member
  • *****
  • Posts: 695
    • View Profile
Re: Sequence crashes in 4.00.35
« Reply #7 on: May 28, 2015, 06:10:49 PM »
Hi Gil,

Thanks for your reply. The reason I am posting as many crashes as I can (or making suggestions) is that I am hoping that one of them will assist you folks in identifying the causes.

Did you mean to imply anything else. This is a Beta version , so would expect there to be crashes, but I also know that if the fseq (post)/creation crash is not fixed , then it cant proceed .

Yesterday I uninstalled everything and re-installed xlights just in case it was something left over from a previous install.

Am happy to assist in any way I can . If you or Dan wish to have versions with more debug statements etc that you wish to provide me outside the regular release , I can run with those and see if it helps identify the cause.
Similarly if there was a way (eg by a config variable) to specify how many threads should be spawned , I can experiment with that and see if it makes a difference to the crash occurrence.

For example , you/I could set it to 1 for a couple of days and see what happens

There is noting special about my Windows setup - on the contrary = it is a 'boring fairly new standard issue'  machine with basic software eg Office etc installed .

I also have Vixen and LSP installed on it, but LSP  is still single threaded (sigh) .

Regards
 
Gerry

Offline gerry

  • Hero Member
  • *****
  • Posts: 695
    • View Profile
Re: Sequence crashes in 4.00.35
« Reply #8 on: May 29, 2015, 05:09:33 AM »
Hello,

I have attached two more reports , only because the last one seems to be different.

This evening , I defined a custom model for the spinner , but I redefined the same channels eg called it Spinner Custom , defined it as a custom model with 8 spokes and tried the fan effect on it. It looked good on the preview , but the save did not like it ie xlights stopped responding and shut down without even a debug report . First time that has happened.

So , I experimented a bit , changing the spinner channels so that they did not overlap. Still crashed but 'normally'.
Now it was not even getting to the 'fseq created message'  but crashing before that.
So I decided to remove the models which had 'custom' ie the faces and the last spinner.
It still crashed before the 'fseq created'
 
Then I decided to remove the Groups from the sequencer (one by one) , just in case the group and items within the group were going to different threads and causing an issue.
looked at the trace each time and it looked (to my naked eye) kind off the same.
But when I removed the Stars group , the trace looked a bit diff , so am posting it.
Have all the others in case u wish to have them.
Regards
Gerry