Valgrind errors while scanning for memory leaks

[Lagrang3]

Running pyln testing with LIGHTNINGD_DEV_MEMLEAK on any test file is triggering
a valgrind error. For example test_pay.py:

LIGHTNINGD_DEV_MEMLEAK=1 pytest tests/test_pay.py -x

Fails with:

==598650== Use of uninitialised value of size 8
==598650==    at 0x24BD6E: htable_val (htable.c:152)
==598650==    by 0x24C3EA: htable_firstval_ (htable.c:167)
==598650==    by 0x24C6A0: htable_del_ (htable.c:406)
==598650==    by 0x12B93A: memleak_ptr (memleak.c:91)
==598650==    by 0x12B975: scan_for_pointers (memleak.c:146)
==598650==    by 0x12B98F: scan_for_pointers (memleak.c:147)
==598650==    by 0x12BAD1: memleak_scan_obj (memleak.c:159)
==598650==    by 0x11162B: memleak_check (libplugin.c:1745)
==598650==    by 0x114BE3: ld_command_handle (libplugin.c:1829)
==598650==    by 0x11509E: ld_read_json_one (libplugin.c:1999)
==598650==    by 0x11512E: ld_read_json (libplugin.c:2019)
==598650==    by 0x24C89D: next_plan (io.c:60)
==598650== 
{
   <insert_a_suppression_name_here>
   Memcheck:Value8
   fun:htable_val
   fun:htable_firstval_
   fun:htable_del_
   fun:memleak_ptr
   fun:scan_for_pointers
   fun:scan_for_pointers
   fun:memleak_scan_obj
   fun:memleak_check
   fun:ld_command_handle
   fun:ld_read_json_one
   fun:ld_read_json
   fun:next_plan
}

at 9afc10b

[jgriffiths]

@Lagrang3 I've found python needs the envvars PYTHONMALLOC=malloc and PYTHONDEVMODE=1 to be set in in order for valgrind output to be useful.

[rustyrussell]

Yes, this is why the tests suppress memleak with valgrind on. We troll through memory looking for pointers, so of course we test uninitialized memory.