3 ms·

Is this important? I treat compiler as full toolchain I run compilation and expect sane error messages - saying that "oh, it's because of linker yada yada" d

Is this important? I treat compiler as *full* toolchain I run compilation and expect sane error messages - saying that "oh, it's because of linker yada yada" doesn't solve my problems nor is valid excuse Other language (compilers) do better.

zabzonk 3y ago
> I treat compiler as full toolchain well, you may do that, but it doesn't make it so. the linker has much less information to go on than the compiler - basically (depending on how you compiled) machine code. actually, the GNU linker does quite a good job, given what it has to work with. > Other language (compilers) do better. a language is not a compiler
- tester756 3y ago
  >well, you may do that, but it doesn't make it so. the linker has much less information to go on than the compiler - basically (depending on how you compiled) machine code. actually, the GNU linker does quite a good job, given what it has to work with. but it was their decision to split the tools like that, wasn't it? nothing technically prevents you to model it in such a way that you can have access to the data you need, right? You probably don't even need to have linker at all. >> Other language (compilers) do better. >a language is not a compiler I wrote compiler in parenthesis. Also, since "language" has two definitions 1. syntax 2. whole ecosystem Then it is valid either way
  - zabzonk 3y ago
    all compiled language systems must support a separate link stage, if they are to be of any practical use
    - tester756 3y ago
      But all of them must have low quality error messages too?
      - zabzonk 3y ago
        as i said, the linker only has so much information - whatever is supplied by the (possibly various) compilers and/or assemblers. which is minimal. so a linker cannot produce very accurate error messages. i'm afraid you are going to have to live with this, and understand it.

3 ms·

Is this important? I treat compiler as full toolchain I run compilation and expect sane error messages - saying that "oh, it's because of linker yada yada" d

by tester756 3y ago

zabzonk 3y ago
> I treat compiler as full toolchain well, you may do that, but it doesn't make it so. the linker has much less information to go on than the compiler - basically (depending on how you compiled) machine code. actually, the GNU linker does quite a good job, given what it has to work with. > Other language (compilers) do better. a language is not a compiler
- tester756 3y ago
  >well, you may do that, but it doesn't make it so. the linker has much less information to go on than the compiler - basically (depending on how you compiled) machine code. actually, the GNU linker does quite a good job, given what it has to work with. but it was their decision to split the tools like that, wasn't it? nothing technically prevents you to model it in such a way that you can have access to the data you need, right? You probably don't even need to have linker at all. >> Other language (compilers) do better. >a language is not a compiler I wrote compiler in parenthesis. Also, since "language" has two definitions 1. syntax 2. whole ecosystem Then it is valid either way
  - zabzonk 3y ago
    all compiled language systems must support a separate link stage, if they are to be of any practical use
    - tester756 3y ago
      But all of them must have low quality error messages too?
      - zabzonk 3y ago
        as i said, the linker only has so much information - whatever is supplied by the (possibly various) compilers and/or assemblers. which is minimal. so a linker cannot produce very accurate error messages. i'm afraid you are going to have to live with this, and understand it.