Page 3 of 7

Re: Serious memory problem in 3.08.1 on all VST - even stock

PostPosted: Mon Jul 17, 2017 1:08 pm
by adamszabo
Im on Win 7 32 bit FL Studio.

But on closer look im getting the same as spogg. It kinda increases but very slowly, not as fast as the others. Then it seems to stop doing it.

Im getting the dissapearing mouse cursos as well if I constantly switch between the task manager and fl studio.

Btw can you guys check this dll too?

Re: Serious memory problem in 3.08.1 on all VST - even stock

PostPosted: Mon Jul 17, 2017 3:27 pm
by Spogg
adamszabo wrote:
Btw can you guys check this dll too?


Hi Adam

Your knob VST continues to increase the memory as I twiddle. Plus it still shows the mouse cursor issue.

Cheers

Spogg

Re: Serious memory problem in 3.08.1 on all VST - even stock

PostPosted: Mon Jul 17, 2017 4:32 pm
by TheOm
Usually this mouse cursor issue happens if you call showCursor in mouseLUp instead of mouseLUpCaptured.
mouseLUp only gets called if the cursor is over the view (which it might not be if you captured the mouse).

Re: Serious memory problem in 3.08.1 on all VST - even stock

PostPosted: Mon Jul 17, 2017 5:09 pm
by Spogg
Well this is getting progressively weirder!

I have a Linkx 1010B tablet running Windows 10 32 bit with 2GB RAM

I thought I’d test the memory issue on this.

My Ruby knob did the same memory thing as on Windows 7, the SM knob was still ok.
The mouse issue on the vector knob was the same.

However, and this is something I just don’t understand, when I tested my 4EA+ and also the Harvester I got NO memory increase no matter what I did, including rapid twiddling of any knob or control. This was tried with Record set and not set. I should add that both plugins worked fine, much to my surprise.

Go figure!

Cheers

Spogg

Re: Serious memory problem in 3.08.1 on all VST - even stock

PostPosted: Mon Jul 17, 2017 10:23 pm
by tulamide
Let me start with the cursor issue, because it is not related to the memory leak:

TheOm wrote:Usually this mouse cursor issue happens if you call showCursor in mouseLUp instead of mouseLUpCaptured.
mouseLUp only gets called if the cursor is over the view (which it might not be if you captured the mouse).
Indeed, TheOm spotted the issue. I created the knob long before we were aware of the mouseLUp issue. mouseLUpCaptured was never documented, so we all used the prior. I used the "Show Cursor" prim to keep the mouse in place, and this prim needs a boolean, which is switched by mouse down/up events. I will correct the issue and post on Flowstone GURU.

So, I used a vector knob for a specific reason; to exclude any issues that might arise from reserving memory for pictures converted to bitmaps. This knob only uses the view bitmap (draws directly to the view). Also, the test is reduced to pure Ruby, to concentrate on this issue alone.

It is totally normal for any process to increase memory usage during execution. It can't foresee the future. If things are defined later, bitmaps get loaded in later, audio is recorded to mem, etc., it will reserve that amount of memory during execution, not in the beginning. And of course it can't release memory that's still being used.

But there's also a clear difference between normal memory increase and a memory leak. In the first dll (the one I posted), there obviously is no memory leak. It acts as expected (yes, we ignore the cursor issue).

The second one (posted by Adam), does have a typical memory leak.

Both are exactly the same schematic. Both are exported exactly the same way. But, number one was created with 3.0.6 and number two with 3.0.8.1. The comparison was important to exclude the rumours that there would still be an old synthmaker leak in there.

The issue therefore is reduced to one spot for MyCo to look at: The export mechanism. Something has changed between 3.0.6 and 3.0.8.1 in the way Ruby code is embedded during export. It might be another Ruby version or a mistake in the programming of the export routines (I wonder what would happen if Ruby got loaded with every knob move :shock: ). There's still a lot to investigate. But MyCo can concentrate on the export routines and dependencies and I hope this saves him a lot of time.

But let me also be clear: It is not an issue from Ruby.

Re: Serious memory problem in 3.08.1 on all VST - even stock

PostPosted: Mon Jul 17, 2017 11:11 pm
by Spogg
Totally excellent investigative work tulamide (as expected!) :ugeek:

But I still don't get my Windows 10 results :?

Cheers

Spogg

Re: Serious memory problem in 3.08.1 on all VST - even stock

PostPosted: Mon Jul 17, 2017 11:26 pm
by lalalandsynth
@ Tulamide. The knob you posted does leak on win 7 . But not on win 8 .

Re: Serious memory problem in 3.08.1 on all VST - even stock

PostPosted: Mon Jul 17, 2017 11:49 pm
by tulamide
lalalandsynth wrote:@ Tulamide. The knob you posted does leak on win 7 . But not on win 8 .

It does not. Read my explanations and the test results of the others.

Re: Serious memory problem in 3.08.1 on all VST - even stock

PostPosted: Tue Jul 18, 2017 1:17 am
by lalalandsynth
I will post screenshots.

Re: Serious memory problem in 3.08.1 on all VST - even stock

PostPosted: Tue Jul 18, 2017 11:42 am
by lalalandsynth
Volume knob test opened in Reaper 5.4 , Win 7 . Tested on 2 computers.
Took quite some wiggling and patience but it leaks.

Initial Memory before moving.
Volume knob test 1.jpg
Volume knob test 1.jpg (92.17 KiB) Viewed 23501 times


After about 6 minutes of wiggling the knob....it seems much slower then other tests but still leaks.
Therefore I will suggest you guys on Win 7 test this again , taking into account that its slower then other leaks.
After I stop moving the knob it does not let go of the memory.
Does not do this on my Win 8 machine but leaks on both Win 7 machines.
I have no clue why , I can only report what I am seeing here.

Volume knob test 2.jpg
Volume knob test 2.jpg (130.01 KiB) Viewed 23501 times