Help debugging memory leak

[Eloitor]

I'm using marimo to run sagemath. The notebooks work fine, but when I edit them sometimes I get mysterious memory leaks.

I'd like to get a minimal example of this memory leak without using marimo, so I can report this problem to sagemath.

Using marimo I found this minimal example:

from sage.all import QuadraticField
F = QuadraticField(5)
F.prime_above(2) 

pari_leak.py.txt

After editing the cell and running again, it crashes the python kernel:

log.txt
log2.txt

Sometimes I get this

[akshayka]

Sorry, to clarify, do you think this is a leak in marimo or in SageMath?

[Eloitor]

I think it is a leak in Sagemath, but I'm unable to reproduce it without marimo...

[Eloitor]

I got a segmentation fault.

I wrote this in a single cell and run it:

from sage.all import QuadraticField
F = QuadraticField(5)
F.prime_above(2) 

Then I deleted the last line and run it again, getting a segmentation fault.

I'm using

[experimental]
execution_type = "strict"

[akshayka]

cc @dmadisetti

[dmadisetti]

That's exciting. I'll try this out, strict mode cannot manage FFI memory, so something might be happening under the hood.

Does this happen only in strict mode?

[dmadisetti]

I do get the crash, but not consistently / reliably. I don't know if this is strict mode specific either since I got the crash after I turned strict mode off.

Seemed a little hit or miss, sometimes restarting the kernel was fine and I couldn't trigger the response at all. Other times crashed first go. Seems like we need someone with an idea of sage internals to give some insight

[dmadisetti]

Notebook I was using for reference

Sage.md