Time  Nick       Message
02:12 ANAND      !tell Krock https://krock-works.uk.to/u/patches/set_rotation_autorot.diff doesn't seem to work. The object auto-rotates only along the global Y-axis, even if had a rotation applied.
02:12 ShadowBot  ANAND: O.K.
08:49 nerzhul    sfan5, #9641 has been fixed and required a last view :)
08:49 ShadowBot  https://github.com/minetest/minetest/issues/9641 -- Replace travis with github actions by nerzhul
11:03 nerzhul    oops i closed it by error ? :D
15:36 nerzhul    sfan5 i have a question about the buildbot windows scripts. What is the interest to pull or clone the repository in this script ? We are in the CI there is no interest to switch commit id whereas we want to test it
15:37 nerzhul    we can just remove the pull/clone part i think
15:37 nerzhul    if it's a end user need it should not be in the code repository unless it's used for a real good reason in the CI
15:37 sfan5      good question, historically the script always did this since it was standalone
15:37 sfan5      I don't think it should be removed though
15:38 nerzhul    i had some scripts like this at work historicall, and it was mor error prone than nothing when you use them and forgot they do git operation when you don't want :D
15:39 nerzhul    maybe we can keep the existing behaviour and change the EXISTING_MINETEST_DIR with a if [ "x$NO_CODE_UPDATE" = "x" ] or something like this, okay ?
15:40 nerzhul    it permits to keep update for who wants it, and remove the ugly symlink CI hack :D
15:40 nerzhul    does that sounds okay to you ?
15:40 sfan5      hm
15:41 sfan5      instead of symlinking, the buildbot.sh scripts could maybe just cd $EXISTING_MINETEST_DIR
15:42 nerzhul    it can be another behaviour but requires extra setup on the CI side i want to avoid to keep the shorter ci. Short CI make it easy to understand for whoever wants to build on its machine
15:43 sfan5      inside the buildbot.sh I mean, no extra code needed for the CI
15:44 nerzhul    i should define the EXISTING_MINETEST_DIR and detect it on the CI
15:44 nerzhul    it's extra setup :D
15:45 nerzhul    the only goal of this variable is to have a clean workdir and the current workdir
15:45 nerzhul    (as far as i understand when you switch the variable value)
15:46 nerzhul    it's why if you want to keep the update part i prefer to add a toggle (default true) to do the clone or build what we have currently
15:47 nerzhul    our travis setup is a little bit crazy, you are in the folder, you register, you go upper, you symlink and you build
15:47 sfan5      nerzhul: here's what I mean http://sprunge.us/nqoRhC
15:48 sfan5      then you can call EXISTING_MINETEST_DIR=$PWD ./util/buildbot/buildwin32.sh cmakebuild from the CI
15:48 nerzhul    exact
15:49 nerzhul    i need to change the cd on minetest game, they will be impacted too (the cd ../.. is specific when you clone the mtg and should not work like you want if you clone or don't update+clone)
15:50 sfan5      huh?
15:50 nerzhul    cd games + cd GAME_NAME if exists  + cd ../...
15:50 nerzhul    in the case the folder doesn't exists
15:50 nerzhul    cd games + clone + cd ../..
15:51 nerzhul    we don't finish in the same folder, but it's recovered by the cd CORE_NAME just after
15:52 sfan5      oh then replace cd ../.. with cd ..
15:52 nerzhul    i will do some tiny changes and test it, let me a second :)
15:58 nerzhul    sfan5 just updated the PR, it sounds fine and proper now :d
16:13 nerzhul    sfan5, green, i hope you like it this time, after that i can return to real cpp life :D
16:14 sfan5      yes I'll take a look soon
17:42 sfan5      nerzhul: can you change the on.push.paths to be more closely what the travis.yml had?
17:42 sfan5      e.g. 'builtin/**.lua' for the lua check
17:43 nerzhul    i can yes
17:43 nerzhul    i'm testing that all tasks are appearing but i'm close :D
17:43 nerzhul    okay for lua, other concerns ?
17:44 sfan5      similar for the build workflow
17:44 sfan5      other than that, no concerns
17:44 nerzhul    i will filter on lib & src
17:44 nerzhul    (we need to rebuild if lib change then)
17:45 sfan5      there's a few more paths https://github.com/minetest/minetest/blob/master/util/travis/common.sh#L47
17:45 nerzhul    ah right, the cmake files too :D
17:45 nerzhul    and the scripts, okay
17:46 nerzhul    i note that travis don't rebuild on lib change but it's required
17:47 nerzhul    okay PR is ready, this time i hope it can be merged haha
17:47 nerzhul    builds are running
17:48 nerzhul    i don't know why when i started to use actions i failed to use 2 workflow files, it's why i did only one, but this time is the good one haha
17:55 sfan5      pushed some small changes to your branch
17:55 sfan5      if the build works now the PR is finally finished
17:56 nerzhul    if i'm correct the windows build will fail
17:56 nerzhul    i tested one time to use an env var befor ethe script call and it failed, but we will see
17:56 sfan5      it doesn't actually use that toolchain file, util/buildbot/ has one for x86_64
17:56 sfan5      oh
17:57 sfan5      https://github.com/minetest/minetest/pull/9641/checks?check_run_id=602690072 looks like it's working
17:57 nerzhul    oh yeah, it's fine, maybe it was due to another concern at the beginning of my gh action discovery :D
17:57 nerzhul    gh actions are powerful but far more harder to use than gitlab ci
18:07 nerzhul    here we are :) will merge after tidy finished
18:09 sfan5      hm I forgot one thing
18:09 sfan5      the removal of util/travis/toolchain_mingw.cmake.in might break gitlab-ci
18:09 sfan5      but you can fix that afterwards
18:09 sfan5      ..or before merge actually
18:12 nerzhul    yep i will restore it
18:12 nerzhul    and merge :)
18:12 sfan5      no, don't restore the file
18:13 sfan5      https://github.com/minetest/minetest/blob/master/.gitlab-ci.yml#L222 remove this line
18:13 nerzhul    okay like github, perfect
18:14 nerzhul    i added a commit, if it's fine i merge :)
18:18 nerzhul    merging #9641
18:18 ShadowBot  https://github.com/minetest/minetest/issues/9641 -- Replace travis with github actions by nerzhul
18:19 sfan5      the badge in the README still points to "validate", oops
18:19 sfan5      should be changed to "build" I think
18:20 nerzhul    fuck
18:20 nerzhul    i forgot that, and what about the 3 workflows we have now... i put the build on the cpp part only ? :D
18:20 nerzhul    or i do 3 badges ? :p
18:21 nerzhul    okay to push build instead of validate on master sfan5 on the readme ?
18:21 sfan5      sure
18:21 sfan5      3 badges is too much
18:22 nerzhul    okay it'sp ushed to master
18:36 sfan5      reminder about #9617, I've had it running on a server for a week without any issues now
18:36 ShadowBot  https://github.com/minetest/minetest/issues/9617 -- Improve protocol-level receiving code by sfan5
18:44 Krock      couldn't invalid packets (too short) hang the server endlessly due to the while() if "getFromBuffers" throws an exception?
18:45 Krock      that was also in the previous code, so no change in behaviour there
18:46 Krock      actually that would throw an InvalidIncomingDataException
18:47 Krock      looks good. Will quickly test
18:48 sfan5      tip: enable no-whitespace diff
18:48 Krock      already happened
18:51 Krock      singleplayer, local server and mutiplayer work.
20:40 rubenwardy What was the original reason for moving from flat files to SQLite? User is asking
21:10 sfan5      performance?
21:10 sfan5      merging #9617 in 10m
21:10 ShadowBot  https://github.com/minetest/minetest/issues/9617 -- Improve protocol-level receiving code by sfan5
21:12 rubenwardy that's what I thought, from the overhead of getting handles from the kernel right? The filesystem in a directory soft limit is unlikely to be exceeded
21:13 sfan5      file i/o even with lots of small ones is fast on Linux
21:14 sfan5      on Windows not so much, you can even observe this when extracting MT from zip: the I/O graph goes down as soon as it reaches the games folders with lots of small files (lua, png)