Ludovic Courtès writes:
> Hello,
>
> Roel Janssen <r...@gnu.org> skribis:
>
>> ==24971== 4,104 bytes in 1 blocks are possibly lost in loss record 351 of 365
>> ==24971== at 0x4C2AAD6: malloc (in
>> /gnu/store/18w3ykyqkcq5zp1qx17qhamkxlczzl0n-valgrind-3.12.0/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
>> ==24971== by 0x4E719E3: sqlite3MemMalloc (in
>> /gnu/store/6d4ihp7xbdh3a0ffbpm5n45q4v3w0l35-sqlite-3.19.3/lib/libsqlite3.so.0.8.6)
>> ==24971== by 0x4E4DB8B: sqlite3Malloc (in
>> /gnu/store/6d4ihp7xbdh3a0ffbpm5n45q4v3w0l35-sqlite-3.19.3/lib/libsqlite3.so.0.8.6)
>> ==24971== by 0x4E51316: pcache1Alloc (in
>> /gnu/store/6d4ihp7xbdh3a0ffbpm5n45q4v3w0l35-sqlite-3.19.3/lib/libsqlite3.so.0.8.6)
>> ==24971== by 0x4E6E71A: sqlite3BtreeCursor (in
>> /gnu/store/6d4ihp7xbdh3a0ffbpm5n45q4v3w0l35-sqlite-3.19.3/lib/libsqlite3.so.0.8.6)
>> ==24971== by 0x4EA9053: sqlite3VdbeExec (in
>> /gnu/store/6d4ihp7xbdh3a0ffbpm5n45q4v3w0l35-sqlite-3.19.3/lib/libsqlite3.so.0.8.6)
>> ==24971== by 0x4EB271E: sqlite3_step (in
>> /gnu/store/6d4ihp7xbdh3a0ffbpm5n45q4v3w0l35-sqlite-3.19.3/lib/libsqlite3.so.0.8.6)
>> ==24971== by 0x4EB34D1: sqlite3_exec (in
>> /gnu/store/6d4ihp7xbdh3a0ffbpm5n45q4v3w0l35-sqlite-3.19.3/lib/libsqlite3.so.0.8.6)
>> ==24971== by 0x426886: nix::LocalStore::openDB(bool) (local-store.cc:293)
>> ==24971== by 0x42BEF4: nix::LocalStore::LocalStore(bool)
>> (local-store.cc:169)
>> ==24971== by 0x40A356: acceptConnection(int)::{lambda()#1}::operator()()
>> const (nix-daemon.cc:755)
>> ==24971== by 0x40E16B: std::_Function_handler<void (),
>> acceptConnection(int)::{lambda()#1}>::_M_invoke(std::_Any_data const&)
>> (functional:1871)
>
> I suspect these “possibly lost” reports are false alarms.
>
> Anyway, there’s no “invalid read” or “invalid write” report, which is
> what we were looking for. :-/
Indeed. There are some 'definitely lost' reports as well, but these are
very small.
I can't seem to reproduce this on my laptop, which is strange because
the machine I do get the test failure on has ECC memory, and my laptop
doesn't have that.
Kind regards,
Roel Janssen
