On Sun, Jan 18, 2015 at 06:00:43PM +0100, Lucas Nussbaum wrote: > In this specific case, it seems to be a random failure: > (jessie-amd64-sbuild)user@ip-172-31-5-2:/tmp/lua-rings-1.3.0$ while true; do > dh_auto_test -O--buildsystem=lua &>/dev/null && echo OK || echo FAIL; done > OK > OK > OK > FAIL > OK > OK > FAIL > FAIL > OK > FAIL > OK > FAIL > OK > FAIL > OK > FAIL > FAIL > FAIL > OK > OK
Hum, after a bunch of attempts I managed to reproduce it. To me the test is too strict on Lua 5.2. It tests the behavior of the garbage collector that is not the same for in 5.1 and Lua 5.2. Moreover Lua 5.2 can "reuse a closure" but the reference manual does not specify when the runtime is able to do so. The test seems to assume it is never able to do so (and it is the case in 5.1). Can you confirm all failures are indeed when running the test on 5.2? I mean, they should all look like this: *********************** app static (5.2) ********* Test: cd src/ && @@LUA@@ ../tests/test.lua Rings 1.3.0 Hello World! app.c: ../tests/test.lua:139: Cache is not being collected The only thing allowed to change is "app static" in the first line, but 5.2 should stay there. It is the case here (but it occurs 1 time every 20 runs or so, so validating my theory here is a bit of a PITA). If it is also the case on your vm I can simply make the test optional (run only on 5.1, where it makes sense IMHO). Best, -- Enrico Tassi -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org