Presumably this is from address space being consumed for each memory.
Eric, can you repro this?
Also, do we have an issue filed on changing navigation to clear all the live memory objects?

Hi, do you have a detailed repro for this issue so that we can verify that the application is not holding onto something that transitively holds onto the WASM instance (and therefore memory)?

Yes, I attached a small zipped web site with this issue.

Inside you will find 2 pages:
 -The main page: memleak.htm
 -and the iframe content: wasmiframe.htm

To reproduce:
 1_Open Chrome and start the developer tools, memory tab
 2_Open the memleak.htm page
 3_Take a heap snapshot
 4_Click on the "reload iframe" button, wait for the tank game to start
 5_loop on #3

As you can see in the attached screenshot, the heap is growing constantly and if you look at the heap you will see the previous ArrayBuffer & code not released.

Hope this helps,
Let me know if you need anything else.

Thanks.

-Thomas

Thanks for the repro. If you don't open dev tools, do you still see the memory leak? If you close and re-open dev tools, does the memory get reclaimed?

No, if I don't open the devtools at all, I don't see the memory leak (According to the Chrome task manager).

If I close and re-open the devtools, then the leak is still there. GC does not get rid of the unused objects.

This looks like  https://crbug.com/750256  to me.