Memory leak when saving pyramid files

When I open a TIFF image using access='sequential' and then save it using compression='jpeg' memory is not released.

Minimal example to reproduce the issue (run passing a large image file path as first parameter:

#!/usr/bin/env python3

import sys

import pyvips

fpath = sys.argv[1]

i = 1
while True:
    print('Iteration {}'.format(i))
    with open(fpath, 'rb') as fh:
        img = pyvips.Image.new_from_buffer(fh.read(), '', access='sequential')
        img.tiffsave('/tmp/flush.tif', compression='jpeg', Q=90)
    i += 1

You can observe memory constantly increasing until eventually the program crashes.

Note that if I don’t use sequential access, or JPEG compression, the memory is flushed as normal.

[scossu]

@workergnome track this

[sixtyfive]

For anyone googling for how to deal with extracting the pages from large PDF files as images, this Issue got me on the right path. The cache doesn’t have to be disabled completely, just limited. Example script here.