00:12 <wanderer> jhass: apparently, this line is causing the crash: https://github.com/manastech/crystal/blob/48461ba58d207bb5de04c9f2270eff03930beeb1/src/compiler/crystal/program.cr#L37

00:18 <wanderer> only when invoking crystal with eval rather than build, though

00:26 <wanderer> well, that's weird.. with eval it's crashing, with build it's not, however, when invoking it within gdb it's crashing for both

01:01 <wanderer> jhass: malloc is part of the intrinsics, isn't it?

01:03 <jhass> pretty sure it is

01:04 <wanderer> it seems that hash.cr is expecting it to zero out the requested memory

01:05 <wanderer> which apparently works on linux and darwin

01:05 <wanderer> however, not on windows

01:06 <wanderer> real malloc() doesn't zero the memory as you probably know, LibC.malloc is defined as well besides this intrinsics-malloc

01:09 <wanderer> Hash#clear should be called in Hash#initialize

01:15 <wanderer> realloc in Hash#rehash also probably won't zero the new memory

01:16 <wanderer> malloc and realloc should either always zero out the memory or it shouldn't be assumed that it does so

01:16 <jhass> looking forward to your patch!

01:18 <wanderer> do you even care about windows? :P

01:18 <jhass> personally? no

01:18 <wanderer> yeah, that's what I thought

01:18 <jhass> that doesn't mean I would oppose it

01:18 <jhass> just that I'm not very interested ;)

01:21 <wanderer> ah, ok, forgot crystal is using boehm gc, so malloc probably gets redirected to gc.malloc which apparently always zeros the memory, calling your function malloc and relying on it to zero memory isn't exactly safe, though, as the original malloc doesn't work like that

01:24 <jhass> I think it's cheap enough and safer in the hash case to just initialize everything

01:30 <wanderer> jhass: yes, however, there might be other malloc-calls depending on zeroed memory besides the ones in hash.cr. btw. how does eval actually work? when I invoke crystal with `eval puts "bla"` it says `undefined local variable or method 'bla'`

01:31 <jhass> that's probably due to your shell quoting

01:31 <jhass> It just passed the passed as argument to the compiler I think

01:31 <jhass> or maybe it even writes a tempfile? I never checked

01:32 <jhass> anyway, the bot here uses it

01:32 <jhass> >> puts "bla"

01:32 <DeBot> jhass: bla

01:32 <jhass> so just a shell quoting issue

01:33 <wanderer> yep, escaping it like `eval puts \"bla\"` works

01:33 <jhass> single quotes should work too

01:33 <jhass> crystal eval 'puts "bla"'

01:34 <wanderer> you're right

02:44 <ponga> >> p "bla"

02:44 <DeBot> ponga: "bla"

03:26 <ponga> btw, malloc sounds like a name for fish

03:26 <ponga> for example, "grilled malloc with mushroom sauce"

03:26 <ponga> just saying

17:52 <eli-se> hi there