WinXP - Win 8 --- something I need to know ?!?

[RJHollins]

BREAKING NEWS !!!!!

I did a mod to NuBeats test module. Made a new VST.

For the first time ... we have had the VST actually LOAD the file back in and populate the menus !


as a VST ... using the same Reaper [i think] on Win 8.1

I've ask for verifications !!!

[Nubeat7]

cool what you changed?

[RJHollins]

Getting more test feedback from BT'r.
In his last email he say:

Here ya go. I did screenshots for each stage. The menus did populate this time. I did this on the installed (not portable) version of the newest 32bit pre-release too. But it took 2 tries.... The first load did nothing, it did not even generate a file in the appdata folder. Did this test "work" this time? I am going to try it on the older portable version (4.32, the same as yours)

Worked the same way on the older portable install. By the way, AFTER is started working, it would repopulate after hitting save again, then blank when hitting delete. Worked back and forth, but like I said, only after I had to load and remove a few times before it would work at all.

after further testing he wrote:

If I close Reaper, it takes 2-4 load/remove cycles of the plugin to get it to work. And I know that this makes no sense, but like the first bach of testing I did for you months ago, it will not load at all into my master buss until I get it to work on a "normal" track. I have tried to just load/remove/load into a master buss up to 11 times, and it will not work. But as soon as I get it to work on another track, it works on master buss (until I close and reopen Reaper, then the process starts over again). Weird, I know, But that is what i am seeing.

The change I made to the code was influenced by the small piece of code I posted earlier. I saw this 'nil' value.

Code: Select all

def event trg
   if trg=='load' && @path!=''
      obj=nil
      File.open(@path, 'rb') {|f| obj=Marshal.load(f)}
      output 0, obj
      output 1, nil
   end
end


So what I did was precede the loading of the arrays with 17 'nils' sent to the respective outputs. I also did the same when the DELETE command executed, thereby clearing all menus.
However, I did NOT 'nil' the array BEFORE the OPEN ... at least not yet

I'm posting the mod'd file [4a], for expert review.

Save-Load Marshal_testschem4a.fsm

mod'd with 'nil'

(21.06 KiB) Downloaded 856 times

From initial reports, we have intermittent success ... as compared to none ... maybe this hints at the cause of the problem. Hopefully this can expose it fully, and provide THE solution.

[Nubeat7]

so your tester says it will work on the 4th time trying it - what does the infopanel says when its not working because if it is saved, the file should exist, file should be a file , size, ext, and loaded or error ...

btw your infolabel (under the selectors)dont update properly because when selector items are changing there is no trigger generated at the droplist output! this means that also the selected item is not updated, after loading the file you have to reselect it to be updated

[Nubeat7]

i don`t think that this nilling the outputs befor outputting the arrays is the reason because it dont affects the file open or the marshalling methode, it just clears the outputs before sending the array data but when it is working like this it also should work without nilling the outputs.

in the snippet from where you got the idea the variable for the loaded file is initialised with nil, would be the same like in your example

Code: Select all

container = nil
File.open(@filename,"rb"){|file| container =Marshal.load(file)}

which also shouldn`t make any difference!

[RJHollins]

so your tester says it will work on the 4th time trying it - what does the infopanel says when its not working because if it is saved, the file should exist file should be a file , size, ext, and loaded or error ...

btw your infolabel (under the selectors)dont update properly because when selector items are changing there is no trigger generated at the droplist output! this means that also the selected item is not updated, after loading the file you have to reselect it to be updated

mmmmm .... NuBeat, I need to read/understand what you are seeing here. I'm playing a Jazz gig tonite ...but when I get back will definately reread.

BTW ... the Selection readouts all work for me ... at least how I've always expected [WinXP] ... maybe we are seeing an issue difference between OS's ... this may be the problem ??????

Thanks for looking at this !!!!! Very important.
thanks

[RJHollins]

OK ... back from gig ... man was that nice Great players and singer !

anyway I'm looking at the schematic and reading NuBeats' post:

btw your infolabel (under the selectors)dont update properly because when selector items are changing there is no trigger generated at the droplist output! this means that also the selected item is not updated, after loading the file you have to reselect it to be updated

I think I follow what you're describing ... I not sure if this has been an issue with using my program [which no one has actually been able to try]. The selector menus don't really need to be changed ... basically, once they get loaded from the config file the first time, that's all we need. The problem has been GETTING that data to load in !

I've sent an email to my tester to get us a screenshot of the status section when first failed. Will post that as soon as I get it.

As to the experiment with NIL ...

The difficulty troubleshooting this has been ME not having the problem. So I'm really guessing what or where the problem is. I find it tough to believe that file access would be it ... it's NOT like we haven't been able to read a file. We can read the XML file and parse without a problem. We have had some intermittent issue writing the config file, but the consistent issue has been trying to then READ the file.

Just to highlight ... prior to the Marshal routine, I had a standard text file being saved using nothing but Primitives [just like we had in SM]. These have never given me issue [on my system, but has on BT'r system]. I then went to TROG's suggestion of Marshal and to the Raw Binary version ... but each time to find it not working on the testers' system.

I wondering .... maybe it is NOT the filesystem.

There is something you [NuBeat] saw with the Selectors ... an issue of updating. I think you are seeing something that I cannot see on my system ... and maybe relates to the core issue my tester has had from the beginning.

Maybe we know better when he sends the info you asked ... Still ... could it be that the file actually DID re-load ... but that the Selectors receiving the array list is not correctly being transmitted ???

By me adding the NIL command prior to sending out the data, I was tracing the extra trigger from the NIL ... even though, as you said, this should be unnecessary ... we DID get some sporadic success of the menus loading. Even sporadic is more than my tester has gotten before

I don't have much to support any of this ... but it seems to me that is Write and Read functions had this failure ... I would think this forum would light up with screaming !!!

2 ... the feeding of these Selectors is a modification that I did Therefore highly suspect ! BUT, in my defense ... it always worked. Not a single glitch in 4 separate project ... only, at least, on my computer. Because it worked, I had nothing to question about this.

If what I'm describing relates in any way to your observation .... I am ready to listen to anything you have to say about this, or what needs to be done to trigger this data from the loaded arrays and send this to the Selector menus.

In my project ... I use a WIRELESS connector off of each TEXT primitive [like we use for the test program], and transmit each array to sub-modules throughout the schematic. Even though this has work for me ... this may be the issue causing the problem. Again, this is speculation on my part ... but it would make more sense that I have something wrong in what I did, rather than a fundamental disk operation problem [that basically we have no control over anyways].

OK ... so I'm a bit wound up after the GIG ... maybe you are able to follow what I'm trying to express and tell me your thoughts or diagnosis. Otherwise I'm left trying to fix something that I can't see broken here.

soon as I get info ... will post. Hope we come to a solution.

Thanks !

[RJHollins]

OK ... we got pics !!!

From the 4a test module, my beta-tester sent these Screenshots.

When the test VST is first loaded, this is what's seen:

first load.png (46.42 KiB) Viewed 16009 times

When the SAVE button is clicked ... he sees this result.

type and save.png (42.09 KiB) Viewed 16009 times

The file is not happening.

After he has tried several times ... it will start to work, and the Saved file will load and populate the menus.

NOTE: To cut down on confusion [since my BT'r is not part of the thread here], I made a display change to the 4a test module of NIL'ing the Status reports BEFORE it sends current message.

Basically output 17 -> 21 NIL

Code: Select all

def event(in_id,v,t)
output 17,nil
output 18,nil
output 19,nil
output 20,nil
output 21,nil
  case in_id
     when 'load' then load_arrays
      #output 18, TRUE  # LED Status for COnfig file
    when 'save'
      #output 18, nil   # LED Status for COnfig file
      save_arrays
      input 'load', nil, time + 1.5
    when 'delete'
      delete_file
      output 0, nil
           output 1, nil
           output 2, nil
           output 3, nil
           output 4, nil
           output 5, nil
           output 6, nil
           output 7, nil
           output 8, nil
           output 9, nil
           output 10, nil
           output 11, nil
           output 12, nil
           output 13, nil
           output 14, nil
           output 15, nil
           output 16, nil
      input 'load', nil, time + 1.5
  end

It seem that this might help to follow the events.

He may be down for the right of the nite ... I'll post any further test results when I get them WITH the change in CODE .

Also ... seeing the Status report seems to indicate the file system has the problem ... I really don't know

[Nubeat7]

i had the same thingin my head, to clear the info messages, last night, just was too tired to finish, but here is a qick mod of the testmodule where i also delayd the infmessages for the loading to better see whats happening, i also toook the reload after the saving into the savemethode to get shure filesave is finished before reload it (i`m not really shure if the "input 'load', nil,time+3" really happens after the file save but i`m pretty shure because the commands in a methode should work serial or?

any way info panel is cleared everytime an action happens,so less confusing old messages...

i also try to load file 3 sec after file is deleted to check if it is really deleted

like this your tester could give concret information what happens when not working..

ps: about the label update - i didnt mean the labels of the selectors i meant the outputs of the selectors (the lebel right under the selectors i changed the second one to see what i mean - also if you dont need it just to clrearify what i mean - just thuoght that your tester maybe was looking at this and didnt sawa change there?

[RJHollins]

I've compiled the VST of test #5 and sent it off for testing.

I think the changes you made help very much mitigate display confusion ... Thanks!

I did add back the clearing of all arrays to begin the test from a 'blank' reference point, as having the data preloaded into the menus could confuse how it got there. I don't know ... just trying to be helpful in figuring this out.

I'll hold off further speculating until we get these results back on this perplexing issue.

Is anyone else willing to help out on this test Considering the possible impact this might have on a released project [with various OS's] ... it would be helpful to see the scope of this issue.

Thanks to all who help ! Please be sure to include the OS and the DAW host [VST version] that was used.

THANKS!