Day changed to 2025-08-11
[01:32] nekoluce_phone: jfw: hello jfw send you a good pdf you can discuss with chad is about Panama Tech plan for 2029
[05:19] jfw: confirmed bug in busybox "grep", a pretty serious one I'd think: it stops printing any output after exceeding ~2bn lines of input, due to overflow of a 32-bit (signed!) integer line counter, used for the -n and -A/B/C features regardless of whether they're invoked.
[05:22] jfw: nekoluce_phone: if you think it's important, how about writing up a summary of those documents for us?
[11:33] lru: jfw: congrats, nice find
[12:45] nekoluce_phone: jfw: panama is starting to invest into cacao industry
[12:45] nekoluce_phone: we should sell little cacao plants :D
[15:06] jfw: bitcoind has the "honors" of being the one to produce 350GB of log file in need of compaction. with that sorted or at least diagnosed for now, our promiscuous listening node is back online, perhaps it'll help with block relay delays which have been flaring up again lately.
[15:06] sourcerer: 2025-02-03 04:01:21 (#jwrd) jfw: restarted my listening node in germany, with permissive mode for good measure, and now they all unwedge in tandem.
[15:16] jfw: been thinking it needs some kind of a "leech mode" - after all, it's mostly nothing but leeches out there so why not get with the times? - where it will receive new blocks and txns from anybody but not *serve* any requests except to known good peers, perhaps by whitelist and/or some proof of utility
[15:17] jfw: possibly combined with shorter-lived tcp connections
[19:22] nekoluce_phone: Panama aims to strengthen its Science, Technology, and Innovation (STI) system by promoting digital platforms for citizen education, continuous training for teachers and researchers, and improving public health services through better prevention, diagnosis, and treatment. Plans include increasing public investment in research and development (R&D), creating an innovation fund, reforming universities to prioritize research and knowledge transfer, and a[CUT]
[20:33] jfw: tbh I thought Panama's Sexually Transmitted Infection system was pretty strong already, lolz
[21:14] dorion: http://jfxpt.com/2025/jwrd-logs-for-Aug-2025/#14739 -- curious, do you know if that bug holds true on cat | grep the same as grep *.txt ?
[21:14] sourcerer: 2025-08-11 05:19:38 (#jwrd) jfw: confirmed bug in busybox "grep", a pretty serious one I'd think: it stops printing any output after exceeding ~2bn lines of input, due to overflow of a 32-bit (signed!) integer line counter, used for the -n and -A/B/C features regardless of whether they're invoked.
[23:30] jfw: dorion: looks like no difference, that much is done sanely, mostly the same code in both cases.
[23:33] jfw: in the case of multiple input files, the line count and thus overflow is per file rather than all together. if it gets as far as 4bn lines it'd probably start printing again as the count goes back into positive.
[23:33] dorion: calls to mind : http://trilema.com/2020/forum-logs-for-04-mar-2020/#2576219
[23:35] jfw: dunno, I never saw an awk or whatever working that stupidly but I'm sure it existed.
[23:40] jfw: easiest fix would probably be to make it a 64bit int, deferring the problem by a factor of 4bn - seems to be pretty much what all the filesystems and file position related system calls did. but it could probably also be fixed to not get confused by the overflow.
[23:41] jfw: or do both; I never used that -n to show line numbers but if you did then you wouldn't want it rolling over at least on any file that can actually fit in the machine.
Day changed to 2025-08-12
[00:02] lru: why not just make it unsigned?
[00:03] lru: a signed number doesn't make sense for a line count
[00:03] lru: and I believe overflow is not undefined behaviour for unsigned like it is for signed
[00:09] jfw: that would probably be part of the change but need to confirm there wasn't some reason however unsensible for being signed. and yes, it becomes well-defined but it's still overflow
[15:30] nekoluce_phone: hello
[18:32] jfw: !e view-tx a003219e365eabfc1bba9d4d3298b5213351a4dd5c797c824d2b81e16123b1d3
[18:32] btcexplorer: http://jfxpt.com/paste/4f6k4jhwvr 1 of 1
[18:36] jfw: !e view-block 765152
[18:36] btcexplorer: http://jfxpt.com/paste/xi5kakitvs 1 of 1
Day changed to 2025-08-13
[20:22] nekoluce: have a good wednesday people :)
[21:53] caai_: happy wednesday luz!
[01:32] nekoluce_phone: jfw: hello jfw send you a good pdf you can discuss with chad is about Panama Tech plan for 2029
[05:19] jfw: confirmed bug in busybox "grep", a pretty serious one I'd think: it stops printing any output after exceeding ~2bn lines of input, due to overflow of a 32-bit (signed!) integer line counter, used for the -n and -A/B/C features regardless of whether they're invoked.
[05:22] jfw: nekoluce_phone: if you think it's important, how about writing up a summary of those documents for us?
[11:33] lru: jfw: congrats, nice find
[12:45] nekoluce_phone: jfw: panama is starting to invest into cacao industry
[12:45] nekoluce_phone: we should sell little cacao plants :D
[15:06] jfw: bitcoind has the "honors" of being the one to produce 350GB of log file in need of compaction. with that sorted or at least diagnosed for now, our promiscuous listening node is back online, perhaps it'll help with block relay delays which have been flaring up again lately.
[15:06] sourcerer: 2025-02-03 04:01:21 (#jwrd) jfw: restarted my listening node in germany, with permissive mode for good measure, and now they all unwedge in tandem.
[15:16] jfw: been thinking it needs some kind of a "leech mode" - after all, it's mostly nothing but leeches out there so why not get with the times? - where it will receive new blocks and txns from anybody but not *serve* any requests except to known good peers, perhaps by whitelist and/or some proof of utility
[15:17] jfw: possibly combined with shorter-lived tcp connections
[19:22] nekoluce_phone: Panama aims to strengthen its Science, Technology, and Innovation (STI) system by promoting digital platforms for citizen education, continuous training for teachers and researchers, and improving public health services through better prevention, diagnosis, and treatment. Plans include increasing public investment in research and development (R&D), creating an innovation fund, reforming universities to prioritize research and knowledge transfer, and a[CUT]
[20:33] jfw: tbh I thought Panama's Sexually Transmitted Infection system was pretty strong already, lolz
[21:14] dorion: http://jfxpt.com/2025/jwrd-logs-for-Aug-2025/#14739 -- curious, do you know if that bug holds true on cat | grep the same as grep *.txt ?
[21:14] sourcerer: 2025-08-11 05:19:38 (#jwrd) jfw: confirmed bug in busybox "grep", a pretty serious one I'd think: it stops printing any output after exceeding ~2bn lines of input, due to overflow of a 32-bit (signed!) integer line counter, used for the -n and -A/B/C features regardless of whether they're invoked.
[23:30] jfw: dorion: looks like no difference, that much is done sanely, mostly the same code in both cases.
[23:33] jfw: in the case of multiple input files, the line count and thus overflow is per file rather than all together. if it gets as far as 4bn lines it'd probably start printing again as the count goes back into positive.
[23:33] dorion: calls to mind : http://trilema.com/2020/forum-logs-for-04-mar-2020/#2576219
[23:35] jfw: dunno, I never saw an awk or whatever working that stupidly but I'm sure it existed.
[23:40] jfw: easiest fix would probably be to make it a 64bit int, deferring the problem by a factor of 4bn - seems to be pretty much what all the filesystems and file position related system calls did. but it could probably also be fixed to not get confused by the overflow.
[23:41] jfw: or do both; I never used that -n to show line numbers but if you did then you wouldn't want it rolling over at least on any file that can actually fit in the machine.
Day changed to 2025-08-12
[00:02] lru: why not just make it unsigned?
[00:03] lru: a signed number doesn't make sense for a line count
[00:03] lru: and I believe overflow is not undefined behaviour for unsigned like it is for signed
[00:09] jfw: that would probably be part of the change but need to confirm there wasn't some reason however unsensible for being signed. and yes, it becomes well-defined but it's still overflow
[15:30] nekoluce_phone: hello
[18:32] jfw: !e view-tx a003219e365eabfc1bba9d4d3298b5213351a4dd5c797c824d2b81e16123b1d3
[18:32] btcexplorer: http://jfxpt.com/paste/4f6k4jhwvr 1 of 1
[18:36] jfw: !e view-block 765152
[18:36] btcexplorer: http://jfxpt.com/paste/xi5kakitvs 1 of 1
Day changed to 2025-08-13
[20:22] nekoluce: have a good wednesday people :)
[21:53] caai_: happy wednesday luz!