Describe the Issue
the mv_gen_tl_producer leaks approximately 10 MB of native memory every time an ImageAcquirer is created and destroyed. memory is never reclaimed.
To Reproduce
run the below code.
Sample Code
I can show a piece of code that demonstrates the reported phenomenon:
import resource
from harvesters.core import Harvester
def rss_mb():
return resource.getrusage(resource.RUSAGE_SELF).ru_maxrss / 1024
h = Harvester()
h.add_file("file.cti")
h.update()
for i in range(20):
before = rss_mb()
ia = h.create(0)
ia.start()
with ia.fetch(timeout=2) as buf:
pass
ia.stop()
ia.destroy()
after = rss_mb()
print(f"iter={i} rss=+{after - before:.1f} MB total={after:.1f} MB")
If applicable, please paste the actual output (its whole traceback, etc) here:
iter=0 rss=+27.3 MB total=80.8 MB
iter=1 rss=+13.2 MB total=93.9 MB
iter=2 rss=+8.3 MB total=102.2 MB
iter=3 rss=+8.5 MB total=110.7 MB
iter=4 rss=+8.2 MB total=118.9 MB
iter=5 rss=+8.2 MB total=127.0 MB
iter=6 rss=+6.2 MB total=133.2 MB
iter=7 rss=+8.1 MB total=141.3 MB
iter=8 rss=+8.2 MB total=149.4 MB
iter=9 rss=+8.2 MB total=157.6 MB
iter=10 rss=+8.2 MB total=165.8 MB
iter=11 rss=+8.2 MB total=173.9 MB
iter=12 rss=+8.2 MB total=182.1 MB
iter=13 rss=+4.7 MB total=186.8 MB
iter=14 rss=+8.5 MB total=195.3 MB
iter=15 rss=+8.2 MB total=203.5 MB
iter=16 rss=+8.2 MB total=211.6 MB
iter=17 rss=+8.2 MB total=219.8 MB
iter=18 rss=+8.2 MB total=228.0 MB
iter=19 rss=+5.2 MB total=233.1 MB
Expected Behavior
memory usage should stabilize and not increase.
Configuration
- OS: Linux 6.12.75+rpt-rpi-2712 [arm64]
- Python: 3.12.13
- Harvester: 1.4.3
- GenTL Producer: BAI BVA - Balluff Impact Acquire 3.6.3
- Camera: Hikrobot MV-CS050-10UM
Reproducibility
This phenomenon can be stably reproduced:
i don't know where things happen, perhaps this issue is in the gentl producer?
Describe the Issue
the mv_gen_tl_producer leaks approximately 10 MB of native memory every time an ImageAcquirer is created and destroyed. memory is never reclaimed.
To Reproduce
run the below code.
Sample Code
I can show a piece of code that demonstrates the reported phenomenon:
If applicable, please paste the actual output (its whole traceback, etc) here:
Expected Behavior
memory usage should stabilize and not increase.
Configuration
Reproducibility
This phenomenon can be stably reproduced:
i don't know where things happen, perhaps this issue is in the gentl producer?