Time  Nick        Message
00:06 paramat     i thought you might have some insight since it's about block loading :]
00:18 hmmmm       i see the change i'm just not sure what to make of it
00:19 hmmmm       there's a reason why i did that
00:19 paramat     thanks, i'm thinking it may be too risky to merge before release
00:20 hmmmm       isGenerated implies not isDummy
00:20 hmmmm       dummy blocks come from active objects getting stored somewhere that hasn't been generated yet
00:22 hmmmm       active object goes into position a mapblock doesn't exist for
00:22 hmmmm       activeobject gets saved and stored as staticobject
00:22 hmmmm       mapblock gets force created with no data, ergo isDummy == true, and isGenerated == false
00:23 hmmmm       is this scenario *ever* true?
00:24 hmmmm       i think it is
00:24 hmmmm       actually i'm not sure
00:24 hmmmm       i need to brush up on entity handling first
00:29 hmmmm       it's not possible for a MapBlock you fetch from Map to be a dummy
00:30 hmmmm       i think that particular check there i added just to be safe
00:32 hmmmm       but that's not relevant to the actual question at hand, d'oh
00:36 hmmmm       if you loadBlock on a block, that duplicates its objects?  really??
00:36 hmmmm       wth
00:36 hmmmm       don't the object IDs clash
00:37 hmmmm       oh wow
00:37 hmmmm       okay this fixes a huge bug
01:08 est31       paramat: its far far too risky to merge it now
01:09 paramat     hm probably
01:11 garywhite   Everyone: https://forum.minetest.net/viewtopic.php?f=3&t=16048 I think this blatantly violates the LGPL licence of MT
01:13 paramat     good grief ..
01:15 paramat     "Anti ban software which allows you to walk around on the client and shout after banned"
01:16 sofar       that's stuff we have to fix
01:16 VanessaE    and that's just stupid, too
01:16 VanessaE    I mean shit, if you're kicked from the server, why is the server continuing to trade packets?
01:16 sofar       because celeron55, lol
01:16 VanessaE    (some of the other stuff claimed there, too)
01:16 sofar       (just kidding)
01:18 paramat     a new depth of assholery
01:20 sofar       paramat: he's only an asshole for violating the GPL
01:20 sofar       I have no problems with people attempting to break minetest
01:20 sofar       if we can get the sources we can better fix the holes he found
01:21 garywhite   Best idea is to file a violation of the Digital Millenium Copyright Act of 1998
01:21 paramat     yeah
01:21 paramat     (.. to sofar)
01:21 sofar       it's hosted on AWS
01:23 garywhite   Then send a notice of Infringement to Amazon
01:23 sofar       that's quite likely possible
01:23 paramat     well i'm still unimpressed by charging large amounts of money to cause havoc on servers
01:24 sofar       HEH, lol
01:24 sofar       so this may be a false lead
01:24 sofar       but, look at the bottom of the page
01:27 sofar       I mean, that is not just a name that 2000 people in the world have
01:27 sofar       that's probably unique
01:27 sofar       and get this, there's a guy in OR with that name
01:27 sofar       ONE ... guy
01:29 paramat     the youtube channel has 'jeremiah wibberding'
01:29 sofar       yup, same guy
01:29 sofar       J is his middle
01:29 sofar       7th day advent cultist
01:30 sofar       ahhh yes! conclusive link!
01:30 sofar       I know what church he goes to and which paintball club
01:30 VanessaE    so go kick his ass :P
01:30 paramat     lol
01:31 sofar       https://www.youtube.com/watch?v=Cgx6CEuZ4Vo
01:31 sofar       go watch that, just to be informed
01:32 paramat     still, since it can't be bought yet there is not yet infringement?
01:32 garywhite   It's based on 0.4.14-dev from the comments, and celeron55: Please report copyright infringement on that video to get it pulled off YouTube
01:32 sofar       no, please don't
01:32 sofar       I strongly suggest nobody take any action yet
01:32 sofar       especially since we've identified the person doing this now
01:33 paramat     agreed, be stealthy
01:33 kaeza       post it on 4chan then?
01:33 VanessaE    that's cold, kaeza  :P
01:33 sofar       kaeza: the fuck? lol
01:34 sofar       nah, we need to decide as a team what to do
01:34 sofar       I also don't want that conversation in this channel
01:34 VanessaE    take it to #minetest-staff then
01:34 VanessaE    that one's not logged, afaik
01:35 sofar       * Cannot join #minetest-staff (Channel is invite only)
01:35 sofar       lol
01:35 garywhite   Is it devs-only?
01:35 VanessaE    devs and forum staff
01:35 garywhite   oh ok. Then I won't even ask for an invite
01:37 paramat     no more talk of this subject in this channel
01:38 paramat     as it's publicly logged
01:39 garywhite   Good point paramat
01:56 paramat     devs, there are still a number of milestone issues #4528 #4638 #4692 #4810 #4909
01:57 ShadowBot   https://github.com/minetest/minetest/issues/4528 -- minetest 0.4.14 64bit crashes at startup on Windows 10 Pro 64bit: debug.txt permissions/path problem
01:57 ShadowBot   https://github.com/minetest/minetest/issues/4638 -- Don't ship with broken languages
01:57 ShadowBot   https://github.com/minetest/minetest/issues/4692 -- Mod security: Accessing mod paths in callbacks is broken
01:57 ShadowBot   https://github.com/minetest/minetest/issues/4810 -- Unittests crash on Release build
01:57 ShadowBot   https://github.com/minetest/minetest/issues/4909 -- Mod security: Still broken with different --world
02:07 est31       paramat: the mod security issues can be fixed 1-2 days before release simply by disabling mod security
02:09 paramat     heh ok, oh yeah that was the plan
02:09 paramat     i forgot
03:30 hmmmm       maybe it's just me, but i feel like the 4847 is okay to merge
03:31 hmmmm       let's just put it this way; there's no drawback to skipping the re-load of blocks in memory
03:31 hmmmm       I don't get why loadBlock doesn't check to see if it's already in memory or not either.. WTF is up with that?
03:47 garywhite   Well, I can confirm that #4528 is an issue only in Windows 10, I use 64bit Windows 7 Pro, and it runs fine...So it's either an issue with security in Win10 or an issue with the WinNT kernel now
03:47 ShadowBot   https://github.com/minetest/minetest/issues/4528 -- minetest 0.4.14 64bit crashes at startup on Windows 10 Pro 64bit: debug.txt permissions/path problem
04:34 est31       well
04:34 est31       its a bugfix
04:35 est31       removing my -1 to get it merged before release
04:35 est31       so if the pr has the appropriate approval, please merge it
04:35 garywhite   est31: Is 0.4.14 still gonna be released before Christmas?
04:36 est31       garywhite: yes
05:04 paramat     #4914 oops something i missed
05:04 ShadowBot   https://github.com/minetest/minetest/issues/4914 -- Make mgv7 the default mapgen in UI. by sofar
05:21 paramat     on that subject i'll make a PR to expand the mapgen list box so that mgvalleys is visible
06:40 sofar       urgh, so we added get_look_horizontal but no conversion function from facedir to/from
07:11 paramat     hm looks like the mapgen dropdown list is of fixed height unfortunately
07:13 paramat     ok nevermind then
07:15 paramat     we'll have to live with the crappy scrollbar
12:27 paramat     will merge #4914 in a moment
12:27 ShadowBot   https://github.com/minetest/minetest/issues/4914 -- Make mgv7 the default mapgen in UI. by sofar
12:36 Zeno`       paramat, I have a team working on the groin bump
12:36 Zeno`       jordach and some others are working on it
12:37 paramat     excellent!
12:37 paramat     merging
12:41 paramat     merged
12:58 nrzkt       est31: i added some precisions and possible fixes which should be done, but warn some real changes needs protocol compat breakage
12:58 nrzkt       https://forum.minetest.net/viewtopic.php?f=3&t=16048&p=242830#p242830
12:58 nrzkt       half of the mentionned points could be fixed
15:33 Fixer       found odd problem
15:42 juhdanad    nrzkt: this means there will be a protocol bump in the near future?
16:22 sfan5       juhdanad: fixing those does not need a protocol bump
16:22 sfan5       though it would be nice if the client knew that it won't have to calculate damage on it's own
16:22 sfan5       for that a protobump might be useful
16:24 est31       for the damage calculation there should be an option
16:24 est31       otherwise youll get issues with lag
20:45 ShadowNinja For #4909: The fix is just copying the component_start != 0 condition to the ".." case as seen in the last fix: https://github.com/minetest/minetest/commit/e4ee6548afd01040046ee3780d0fbb121d141251  I might also tweak the code a bit and I'll add a test case since RemoveRelativePathComponents("../foo") obviously shouldn't be "foo" since that's an entirely different directory.
20:45 ShadowBot   https://github.com/minetest/minetest/issues/4909 -- Mod security: Still broken with different --world
20:45 ShadowNinja I'll push it as trivial in a bit unless there are comments.
20:46 garywhite   Has anyone heard of snapcraft?
20:46 garywhite   Sorry, wrong channel
20:48 est31       ShadowNinja: please show your actual change before pushing
20:49 est31       ShadowNinja: also, can you amend your pr according to paramat's comments?
20:49 est31       https://github.com/minetest/minetest/pull/4849
20:51 ShadowNinja Actually, there might be a bit more to this...
20:52 ShadowNinja est31: Uh, in a minute...
20:56 ShadowNinja Hmmm, RemoveRelativePathComponents currently bails on "../foo" and returns foo.  It'll have to be changed to just remove as many relative components as possible, which means more changes.
22:09 ShadowNinja So, it turns out the fix is fairly complicated.
22:14 ShadowNinja RemoveRelativePathComponents is completely broken, since "../foo" is transformed into "/foo".
22:15 ShadowNinja But it's also all-or-nothing.  "../foo" *should* return "" since the ouput isn't allowed to have any remaining relative components due to the way it works.
22:19 ShadowNinja Now, in the security code we have to basically iterate up the directory structure using the path given to us.  So given "bar/../worlds/foo/noexist" we want to try "bar/../worlds/foo", "bar/../worlds", and "bar/..", but not "bar" (otherwise we would have an issue with "/minetest_world/../etc/passwd".
22:19 sfan5       well minetestserver should probably resolve relative paths before passing them to something else
22:20 ShadowNinja Currently the path is just passed to RemoveRelativePathComponents to handle this, but if someone tries to open "../worlds/foo/auth.txt" it will fail because RRPC will return "".
22:20 ShadowNinja So, I've fixed RRPC and I've changed the security code to no longer use it.
22:21 ShadowNinja The security code now simply removes the components one-by-one and bails as soon as one of them is "..".
22:23 ShadowNinja This way "../worlds/foo/auth.txt" is handled properly.  It's a bit conservative (e.g., "worlds/foo/players/../auth.txt" will fail), but handling the remaining cases will require a new function or a change in how RRPC works.
22:24 ShadowNinja sfan5: Normally AbsolutePath is used for everything, but AP only works if the path exists, which isn't the case with mkdir().
22:25 sfan5       i wonder whether --world /../some/dir works with mt
22:27 ShadowNinja Actually, nevermind, "worlds/foo/players/../auth.txt" will work since it's resolvable with AP.  What won't work is something like "worlds/foo/noexist/../noexist".
22:28 ShadowNinja In fact, since mkdir() doesn't work recursively IIRC, we could just check the immediate parent, which would only break "worlds/foo/noexist/noexist/..".
22:29 ShadowNinja sfan5: Probably not, seeing as that can't possible exist.
22:30 sfan5       linux disagrees
22:30 sfan5       % file /../usr/local
22:30 sfan5       /../usr/local: directory
22:30 ShadowNinja Oh, didn't know /.. == /
22:35 ShadowNinja sfan5: It works.
22:36 sfan5       huh
22:38 ShadowNinja Alternatively I could make RRPC more oportunistic, so "../foo/bar/../baz" becomes "../foo/baz" instead of "".
22:38 ShadowNinja Right now "" means "error", but that's a reasonable return value for ".".
22:40 ShadowNinja That would basucally just mean replacing `return "" if dotdot_count > 0` with `path = ("../" * dotdot_count) + path` (in pseudocode).
22:42 est31       I think this is something which should get lots of unit tests
22:42 ShadowNinja I added one for "../a".
22:42 Fixer       i want to add suggestion on github related to mouse tricks while working with inventory, it is engine part?
22:42 ShadowNinja Fixer: Yep.
22:43 est31       Fixer: which mouse trick do you mean
22:44 Fixer       est31: my previous mouse scroll wheel died because of this transfer 10 blocks clicking, so i have an idea
22:44 est31       yeah scroll wheels are the first thing to die in a mouse
22:45 Fixer       est31: you transfer block, press right mouse button, it transfers 1 block and after short delay starts transfering more and more progressively increasing its speed of transfer
22:45 est31       yeah that is an engine feature request
22:45 Fixer       it is pain in the ass currently to work with lots of blocks, i miss some of minecraft tricks, that are quite cool
22:53 sofar       est31: my wife had a very expensive mouse: left click microswitch died first (and that mouse type had many users complain about it)
22:55 kaeza       <Fixer> est_31: you transfer block, press right mouse button, it transfers 1 block and after short delay starts transfering more and more progressively increasing its speed of transfer <-- sounds like Terraria
22:55 Fixer       maybe
22:55 celeron55   i've never seen scroll wheels die; it's always the left click switch
22:55 Fixer       maybe we need to put ctrl to use with a mouse
22:56 kaeza       for me it's always been right one
22:56 Fixer       celeron55: that mouse was like 1 year old, ultracheap one, still I was quite annoyed
23:58 sofar       do we have a standard invisible texture that's 1x1px in mtg?