2024-07-29 10:33:46 <veltas      > I think Forth is actually easy to debug, just need to debug more than you would other things
2024-07-29 10:33:55 <veltas      > But at least you've got a REPL to work through it
2024-07-29 13:22:39 <zelgomer    > you certainly have to pace yourself and test incrementally as you go. if you write a screen full and then try to load it and it segfaults, then you start having to insert prints and binary search for the cause
2024-07-29 13:24:41 <veltas      > The same has to happen with C
2024-07-29 13:25:02 <veltas      > The difference is C is less likely to produce these errors, and more likely to also warn or error before you have to debug
2024-07-29 13:25:21 <veltas      > But when you get mystery segfaults it's the same bag
2024-07-29 13:25:39 <veltas      > Especially when you're using STC and can debug with gdb, as I've done with Forth sometimes
2024-07-29 13:26:12 <veltas      > So you will have more debug cost in Forth, but I don't think the debug cost in Forth per bug is particularly high
2024-07-29 13:26:22 <veltas      > You can just expect more bugs
2024-07-29 13:26:46 <veltas      > Just my impression from fixing my own broken code
2024-07-29 13:26:54 <veltas      > Even with Forth I tend to write a screenful
2024-07-29 13:27:13 <veltas      > I'm not afraid of debugging so don't try everything out as I go, although that does feel nicer sometimes
2024-07-29 13:37:01 <olle        > 
ACTION looking at vau calculi

2024-07-29 13:38:42 <veltas      > Debugging's one of those important costs people fail to consider.  It's one of the reasons I don't use C++
2024-07-29 13:39:16 <veltas      > Because when something goes wrong and there's no obvious solution, suddenly your whole stack is under scrutiny.  You have a probability distribution of many things.
2024-07-29 13:39:26 <veltas      > And the distribution for C++-isms is worse than e.g. C
2024-07-29 13:40:06 <veltas      > The foot-shotgun thing
2024-07-29 13:45:59 <veltas      > It's why zero-code, AI, etc all fall flat in practice.  Stuff doesn't actually work and then you need a real programmer/engineer to fix it.
2024-07-29 13:46:13 <veltas      > It's great.... until it doesn't work!
2024-07-29 13:46:39 <veltas      > I prefer a compromise, understand as much as you can but sure use more declarative programming to make life easier
2024-07-29 14:01:47 <zelgomer    > i can't think of many times my c compiler has segfaulted just loading the code
2024-07-29 14:11:54 <veltas      > It might if it could execute code
2024-07-29 14:12:21 <veltas      > I've seen C compilers generate incorrect code, and C++ compilers segfault
2024-07-29 14:12:40 <veltas      > I saw MSVC++ segfault on code in the C++ standard document
2024-07-29 14:12:59 <veltas      > And I've seen SDCC generate incorrect code
2024-07-29 14:13:25 <veltas      > I've only seen one questionable generation from GCC although it was defended by the devs and is a bit of a grey matter
2024-07-29 14:36:13 <zelgomer    > lol i've seen gcc run out of memory
2024-07-29 14:55:01 <veltas      > I've seen GCC run out of memory because of exponentially large macro substitutions
2024-07-29 20:53:12 <MrMobius    > what was the incorrect SDCC code?
2024-07-29 20:53:29 <MrMobius    > just curious. I used it briefly for 8051 before fleeing in terror