Gathering data on the CircuitPython memory weirdness. #6992
Description
CircuitPython version
Adafruit CircuitPython 8.0.0-beta.1 on 2022-10-01; Raspberry Pi Pico W with rp2040Code/REPL
from gc import mem_free, collect
from sys import exit
targ_mem = None
mem_offs = 69420 # offset to add, value set to allocate the needed ram
# tests
# for loop test
for_t = False
# int test
int_t = False
mem_offs = 0 # revert the value to 0, so we can actually use it
# script begin
def memstep(fail=True):
global targ_mem, mem_offs
collect()
f = mem_free()
diff = targ_mem - f + mem_offs
if diff:
print("-> Diff: " + str(diff), end="")
if fail:
print(" | This is not fine.\n\n--- FAIL ---", end="")
exit(1)
else:
print(" | This is fine.",end="")
print()
del f, fail, diff
collect()
collect()
targ_mem = mem_free()
print(f"\nTarget mem is set to {targ_mem}.")
# fstrings with vars leave collectable garbage
print("\nInitial test")
memstep()
if for_t:
print("Before \"for\" test")
memstep()
print("In \"for\" test")
mem_offs -= 16
for i in range(30):
memstep()
mem_offs += 16
collect()
print("After \"for\" test")
memstep()
if int_t:
print("In \"int\" test")
i = 69420531221
#collect()
mem_offs -= 16
memstep()
print("Del \"int\" test")
del i
mem_offs += 16
#collect()
memstep()
print("End test")
collect()
memstep()
print("--- PASS ---")
exit(0)Behavior
As is, it works. This means the memstep function and the prints aren't leaving anything behind.
If we now for_t = True, we instead fail:
Target mem is set to 158560.
Initial test
Before "for" test
In "for" test
After "for" test
-> Diff: 16 | This is not fine.
--- FAIL ---
Code done running.
Press any key to enter the REPL. Use CTRL-D to reload.
if the offset wasn't there we would fail within the "for" test.
The outcome doesn't change if I replace the memstep with a pass inside the for loop.
If we then for_t = False and int_t = True, we yet again fail:
Target mem is set to 158560.
Initial test
In "int" test
Del "int" test
-> Diff: 16 | This is not fine.
--- FAIL ---
Code done running.
Press any key to enter the REPL. Use CTRL-D to reload.
What even holds those 16 bytes???????????
Description
No response
Additional information
For anyone who wonders as to why this is important, ljinux runs stuff on a loop pretty much forever, these leaks accumulate into many kilobytes.