Time  Nick       Message
00:50 p_gimeno   sfan5: +1 for no-math-errno and no-trapping-math. For no-signed-zeros it's a risk to have a possible signed zero in the output and to be unable to get rid of it, but I'd say it's a risk you can live with; other than that, the semantics of minus zero are mostly irrelevant to anything but output.
00:52 p_gimeno   As for finite-math-only, that's one I don't like, especially with floats, because it's easy to get them. I've written code that relies on having proper IEEE infinity semantic in corner cases.
01:01 p_gimeno   here's a sign() function that relies on IEEE semantics with respect to infinity: https://love2d.org/forums/viewtopic.php?p=218807#p218807
06:25 nerzhul    hello, merging #9671 (thanks for the review, and enjoy the big perf improvement)
06:25 ShadowBot  https://github.com/minetest/minetest/issues/9671 -- Optimize get_objects_inside_radius calls by nerzhul
08:23 nerzhul    merging #9682
08:23 ShadowBot  https://github.com/minetest/minetest/issues/9682 -- Swap out -ffast-math for a safe subset of optimization flags by sfan5
09:24 sfan5      p_gimeno: considering it would not have worked on Windows and most likely the devs machine too (since cmake sets the flag), I don't think we have any code like that
09:25 p_gimeno   sfan5: fair enough
09:25 sfan5      I also don't know how much faster finite-math-only is, but if it happens to cause any problems in the future it can always be turned off
09:25 sfan5      so this is good enough I'd say
09:44 p_gimeno   by the way, that should really put #3943 to rest
09:44 ShadowBot  https://github.com/minetest/minetest/issues/3943 -- testStreamRead and testBufReader failures
09:45 p_gimeno   ah it says so in the commit message
09:45 sfan5      0 / 37 failed modules (0 / 204 failed individual tests).
09:45 sfan5      it does :)
09:54 Daisae     https://sourceforge.net/p/irrlicht/bugs/435/
09:55 Daisae     Irrlicht tells that Copy+Paste should work since about 4 + 1/2 years ago -- but it is not working in Minetest.
09:59 sfan5      works for me ... somewhat
09:59 sfan5      sometimes I have to press ctrl+v twice
10:02 Daisae     Oh. If I press it Ctrl+V twice, it does paste.  Shift+Insert and middle mouse click do not paste.
10:28 nephele    Sounds like X11 problems to me ;)
10:29 nephele    (tl;dr X11 has severall clipboards, and it's not obvious to users why some key combo works (by invoking one) over another)
11:20 Daisae     On X11, there is the 'clipboard' and there is the 'selection'. Sometimes there is confusion because they are separate. I synchronized both of them before attempting to paste in minetest. So, it was not due to that confusion.
11:27 rubenwardy it works in other applications, there's something wrong in Irrlicht or MT
13:35 p_gimeno   I use both 'primary' and 'clipboard' and I can say for sure that Minetest only uses 'clipboard', so that's not the problem
13:55 cheapie    Daisae: I've noticed that before, when pasting into MT (with ctrl+V) right after copying something (with ctrl+C), I get the old content on the first paste, but the correct content on any subsequent pastes.
13:56 cheapie    I've just gotten into the habit of, when copying programs from a text editor into a Luacontroller, doing ctrl+V, ctrl+A, ctrl+V.
16:26 Krock      merging #9681 in 5'
16:26 ShadowBot  https://github.com/minetest/minetest/issues/9681 -- Camera: Fix shooting line offsets by SmallJoker
16:31 Krock      merging
17:11 nerzhul    sfan5, you think we should remove the -Wabi in clang too then ?
17:11 sfan5      -Wabi does not mean anything in clang, yes
17:11 nerzhul    then my PR will finish in remove the flag
17:12 sfan5      +1
17:13 nerzhul    okay it's push and commit message is now proper :)
17:13 nerzhul    just tested on gcc 9.3 & clang 9 it's fine
17:15 nerzhul    sfan5 can you take a look at #9689, it sounds fine except i'm not sure about the style change in error message handling
17:15 ShadowBot  https://github.com/minetest/minetest/issues/9689 -- Reduce pkt->getPeerId() invocations in serverpackethandler by HybridDog
17:16 nerzhul    #9676 ready for a merge if sfan5 is always okay :)
17:17 ShadowBot  https://github.com/minetest/minetest/issues/9676 -- Fix GCC warning spam on ABI by nerzhul
17:17 nerzhul    another question, on #9677 do we keep my approach or we add the error message like Krock suggests ?
17:17 ShadowBot  https://github.com/minetest/minetest/issues/9677 -- Add an option to disable unittest build, & disable them on Docker build by nerzhul
17:17 sfan5      merging #9685
17:17 ShadowBot  https://github.com/minetest/minetest/issues/9685 -- Android: fix formspec input for AArch64 devices by MoNTE48
17:18 sfan5      nerzhul: there definitely should be an error message
17:18 sfan5      it shouldn't silently fail to run the tests
17:21 nerzhul    with my pr unittests will not fail, you will not have the option to run it because there is no code :D
17:21 nerzhul    but we can keep the message about unittest code disabled if you want
17:21 nerzhul    merging #9676
17:21 ShadowBot  https://github.com/minetest/minetest/issues/9676 -- Drop useless -Wabi flag by nerzhul
17:27 nerzhul    on unittest pr i will add the error message, and i just see that client unittest are enabled, i forgot to drop them :)
17:29 nerzhul    #9677 is now ready for a last review/merge
17:29 ShadowBot  https://github.com/minetest/minetest/issues/9677 -- Add an option to disable unittest build, & disable them on Docker build by nerzhul
17:30 nerzhul    Krock ^
17:31 Krock      neutral
17:31 Krock      will review though
17:32 nerzhul    it's just to disable them on useless systems :D
17:32 nerzhul    docker & android don't need them
17:32 nerzhul    (and will never have a way to run them)
17:32 Krock      oh! right. the README also needs documentation
17:33 nerzhul    good point, i will amend the doc to add the flag doc, i forgot it
17:33 nerzhul    sfan5, for the emerge threading issue, don't you think this warning is one part of the problem in terms of concurrency ?
17:33 nerzhul    attention: « void* memcpy(void*, const void*, size_t) » écriture dans un objet de type « struct NoiseParams » sans affectation par copie triviale; use copy-assignment or copy-initialization instead [-Wclass-memaccess]
17:33 nerzhul    i think this one sould be fixed to prevent pointer copy then strange behaviour on multiple thread access to the member pointers
17:34 sfan5      do NoiseParams contain pointers?
17:35 Krock      there are no pointers in NoiseParams
17:35 Krock      not a single one
17:36 sfan5      then it's not relevant for concurrency
17:36 Krock      they should be treated as const anyway - at runtime
17:36 nerzhul    oops :D
17:36 sfan5      nerzhul: the line that generates that warning should be replaced with "this->np = *np_;" though
17:37 nerzhul    fine :)
17:37 nerzhul    Krock i added the doc, i am not very verbose, the flag talks itself i think :D
17:39 Krock      yse
17:42 nerzhul    thanks, i will merge after CI pass
17:44 sfan5      nerzhul: did you document the flag in README.md?
17:44 sfan5      it has a list of cmake flags
17:47 Krock      sfan5: fixed 9 minutes ago :P
17:47 nerzhul    sfan5 just what Krock said before and i added it :D but double check is fine haha
18:43 nerzhul    merging #9677
18:43 ShadowBot  https://github.com/minetest/minetest/issues/9677 -- Add an option to disable unittest build, & disable them on Docker build by nerzhul