Time Nick Message 00:42 Sheriff_U3 Are there any plans to remove the minetest namespace in a future Luanti version? 00:47 MTDiscord we'll do it right after the year of the linux desktop 00:48 Sheriff_U3 In other words: Any year now! lol 00:50 Sheriff_U3 But seriously, is it something that would likely be done in say Luanti v6.0.0+? 00:51 MTDiscord i suppose it is a possibility. if a hypothetical future luanti version really was completely incompatible, what point would there be in maintaining such an alias? the alias not being there would be a feature, a clear indicator that the chance of the code working is virtually zero. 00:52 MTDiscord but this is very far away because we do not have the capabilities (nor the intent) to completely overhaul the API. 00:52 MTDiscord even Luanti 6.0 is likely to only bring relatively "minor" breakages in the grand scheme of things, just look at doc/breakages.md 00:53 MTDiscord so Luanti 6.0 will certainly keep the alias. 07:50 [MatrxMT] it costs (almost?) nothing to keep it there 08:00 [MatrxMT] it would frustrate a lot of LLMs that love to still output `minetest` mind you... 08:39 MinetestBot 02[git] 040-v-0 -> 03luanti-org/luanti: Fix typos in comments (#17215) 13451ac24 https://github.com/luanti-org/luanti/commit/451ac24ac3c2089791e137ddbcd77153b0ef57b5 (152026-05-29T08:37:23Z) 08:39 MinetestBot 02[git] 04SmallJoker -> 03luanti-org/luanti: Formspec: Do not close on focus change 137701e71 https://github.com/luanti-org/luanti/commit/7701e71d6d7112c14094effe5dfd45e990465d98 (152026-05-29T08:37:50Z) 11:08 pgimeno even crazier: make half the program use little endianness and the other half use big endianness (but you can no longer use compiled code for that, I guess) 11:09 sfan5 you can compile the code as separate objects and let the linker put them together 11:09 sfan5 I think 11:36 cheapie pgimeno: I have some tests I've done like that 11:37 cheapie Like this one that switches the endianness partway through (and intentionally has some endianness-related bugs to demonstrate that it actually changed): https://cheapiesystems.com/media/images/screenshot_20260527_185443.png 11:39 cheapie It's groups of 8 characters being reversed instead of 4 like you may expect from a 32-bit processor, since RVController supports the Zilsd/Zclsd extensions for doing 64-bit loads/stores and the buggy code is using them 11:41 cheapie sfan5: Really there's nothing stopping you from putting some inline assembly or whatever partway through a C program that flips the endianness, but if you do that in a function that needed the stack at all it'll probably just crash on return 11:42 cheapie RISC-V instructions are always little-endian (the MBE bit only affects stores/loads), so while it will continue to execute instructions just fine for a little while, the moment it pops a return address off the stack it'll probably jump off into outer space and crash 11:44 cheapie Not that it's that hard to fix up the address before returning if you're expecting this (like if you're writing assembly or something), it's just a "rev8 ra,ra" (and repeat for any other registers you popped) before returning 12:16 pgimeno or use a palindrome return address :D 12:16 pgimeno (awesome that you already used mixed endianness) 12:20 cheapie Unfortunately the only possibly palindrome return address on RVController would be 0x00000000 12:21 cheapie RAM only goes up to 0x0000FFFF (and does not alias), there's just nothing above that 12:23 cheapie Reading from an invalid address (past the end of RAM) gives an error message on the monitor and halts after the current instruction finishes, but also if the read was for an instruction fetch, 0x00000000 is not a valid instruction and so that /also/ halts it 12:24 cheapie Specifically it gets interpreted as a compressed instruction, and 0x0000 is defined as an invalid 16-bit instruction. 0xFFFFFFFF is interpreted as 32-bit but also invalid 12:30 cheapie The method for detecting whether an instruction is compressed (16-bit) or not is amusingly simple, it's just... if the two least significant bits are both set then it's 32-bit, otherwise 16. Since instruction fetches are always little-endian regardless of MBE (and SBE/UBE on processors that implement those), this is always in the first byte read 14:41 cheapie In today's episode of this project going way farther than I ever thought it would, RVController now has an officially-assigned value (53) for its marchid CSR (see https://github.com/riscv/riscv-isa-manual/blob/main/marchid.md ) 15:23 Sheriff_U3 Just ran across another gun mod, this one looks interesting though. https://github.com/stujones11/shooter 15:47 Krock unfortunately his mods need adoption. Shooter yet hasn't received one. 15:52 Krock Sheriff_U3: here's one that uses raycast. Might be interesting if you'd like to write your own (CC0) 15:52 Krock https://github.com/SmallJoker/farbows/ (shameless advertisement)