3 ms·

Unsafe is not a problem in my experience. The same way a kernel might have some part written in asm (for freebsd it's like 1% I believe) your rust code will hav

Unsafe is not a problem in my experience. The same way a kernel might have some part written in asm (for freebsd it's like 1% I believe) your rust code will have 1% of Unsafe code. Might be more depending on what you are doing. However the same way few bugs happen in the asm code of a kernel, your unsafe will not have many bugs.

saagarjha 6y ago
Actually, a lot of bugs happen in the assembly part of the kernel. Processors are hard ;)
- loeg 6y ago
  Do you have a source for that claim?
  - saagarjha 6y ago
    A couple examples off the top of my head: all the recent speculative execution stuff, OpenBSD's buggy SMAP implementation, some recent bugs ARM's PAN.
    - adamnemecek 6y ago
      OK thats how many 10? How many are not asm bugs? Like it's not many considering the total number of bugs.
    - loeg 6y ago
      You specifically responded to a claim that there were relatively few assembler bugs due to it be a relatively small body of code. So in that context, I would like you to show that the handful you've highlighted is "a lot" against the backdrop of bugs in non-assembly code. My semi-informed (but not quantified) opinion is that you are broadly wrong and that there are not a lot of assembler bugs relative to C/C++ bugs.
- adamnemecek 6y ago
  Relative to what?
  - saagarjha 6y ago
    Relative to zero.
    - adamnemecek 6y ago
      No its relative to the total amount of bugs.
      - saagarjha 6y ago
        In the context I was replying to, it's not. My response is to the claim that writing code in assembly has "few" bugs.
        adamnemecek 6y ago
        Right, comparatively.
        loeg 6y ago
        No, it's pretty clear from the full context of the comment you are replying to that it is expressing a relative measure[1]: > Unsafe is not a problem in my experience. The same way a kernel might have some part written in asm (for freebsd it's like 1% I believe) your rust code will have 1% of Unsafe code. Might be more depending on what you are doing. However the same way few bugs happen in the asm code of a kernel, your unsafe will not have many bugs. [1]: https://news.ycombinator.com/item?id=23510862 https://news.ycombinator.com/item?id=23510862
snkIaq 6y ago
Proven asm code like in TAOCP is safer than any other language. For example you know exactly what the FPU will do. Good luck with any compiler. I know that is not what people mean with "unsafe", but it is getting tiresome.
- ohgodplsno 6y ago
  TAOCP was written in different times, namely one where microcode was not a widespread thing. As it stands, your asm "proof" relies on having no µcode update. CPUs are no longer easy to understand like 8086s.

saagarjha 6y ago

Actually, a lot of bugs happen in the assembly part of the kernel. Processors are hard ;)

loeg 6y ago
Do you have a source for that claim?
- saagarjha 6y ago
  A couple examples off the top of my head: all the recent speculative execution stuff, OpenBSD's buggy SMAP implementation, some recent bugs ARM's PAN.
  - adamnemecek 6y ago
    OK thats how many 10? How many are not asm bugs? Like it's not many considering the total number of bugs.
  - loeg 6y ago
    You specifically responded to a claim that there were relatively few assembler bugs due to it be a relatively small body of code. So in that context, I would like you to show that the handful you've highlighted is "a lot" against the backdrop of bugs in non-assembly code. My semi-informed (but not quantified) opinion is that you are broadly wrong and that there are not a lot of assembler bugs relative to C/C++ bugs.
adamnemecek 6y ago
Relative to what?
- saagarjha 6y ago
  Relative to zero.
  - adamnemecek 6y ago
    No its relative to the total amount of bugs.
    - saagarjha 6y ago
      In the context I was replying to, it's not. My response is to the claim that writing code in assembly has "few" bugs.
      - adamnemecek 6y ago
        Right, comparatively.
      - loeg 6y ago
        No, it's pretty clear from the full context of the comment you are replying to that it is expressing a relative measure[1]: > Unsafe is not a problem in my experience. The same way a kernel might have some part written in asm (for freebsd it's like 1% I believe) your rust code will have 1% of Unsafe code. Might be more depending on what you are doing. However the same way few bugs happen in the asm code of a kernel, your unsafe will not have many bugs. [1]: https://news.ycombinator.com/item?id=23510862 https://news.ycombinator.com/item?id=23510862

snkIaq 6y ago

Proven asm code like in TAOCP is safer than any other language. For example you know exactly what the FPU will do. Good luck with any compiler. I know that is not what people mean with "unsafe", but it is getting tiresome.

ohgodplsno 6y ago
TAOCP was written in different times, namely one where microcode was not a widespread thing. As it stands, your asm "proof" relies on having no µcode update. CPUs are no longer easy to understand like 8086s.