nimforum mirror - nimrod and the C compiler disagree on the size of a pointer

fuzzthink (orginal) [2014-07-21T19:19:22+02:00] view original

Trying to emcc the generated c file, got:

./nimbase.h:382:28: error: 'assert_numbits' declared as an array with a negative size"

typedef int assert_numbits[sizeof(NI) == sizeof(void*) && NIM_INTBITS == sizeof(NI)*8 ? 1 : -1];

See in nimbase.h comment that it is because nimrod and the C compiler disagree on the size of a pointer.

So how would I go about fixing this?

(on, osx Mavericks)

Araq (orginal) [2014-07-21T20:51:25+02:00] view original

For instance you can use --cpu:i386 to make the nimrod compiler think you have a 32bit CPU. That said, I don't know if emscripten is up for the task of translating nimrod's GC. I would use the JS backend instead.

fuzzthink (orginal) [2014-07-21T21:36:26+02:00] view original

Just tried --cpu:i386 option during nimrod compile, it just generates the same error in the nimrod compilation instead.

FYI, I am using the JS backend, but wanted to see how big of a difference the emscripten code performs vs. the nimrod version.

Edit: Not a big deal if you don't think emscripten can handle nimrod's GC anyways.

Araq (orginal) [2014-07-24T17:38:08+02:00] view original

Just tried --cpu:i386 option during nimrod compile, it just generates the same error in the nimrod compilation instead.

Well you don't have to bootstrap with this switch if that's what you mean by "in the nimrod compilation".

Edit: Not a big deal if you don't think emscripten can handle nimrod's GC anyways.

Well it surely is an interesting experiment. :-)

fuzzthink (orginal) [2014-07-25T21:29:56+02:00] view original

Your "don't have to bootstrap with this switch" makes me think I may be doing something wrong. Anyways, here's what I have:

my hello.nim is just: echo("hello")

I did: nimrod c --cpu:i386 --os:linux --compile_only hello.nim

emcc hello.c (in nimcache, after copying Nimrod/lib/nimbase.h in nimcache) gives:

 warning: unresolved symbol: setStackBottom
warning: unresolved symbol: systemDatInit
warning: unresolved symbol: systemInit
warning: unresolved symbol: frameptr_14842
warning: unresolved symbol: nim_program_result

node a.out.js gives:


missing function: systemDatInit
-1
-1

/Users/john/dev/asmjs/nimcache/a.out.js:5973
      throw e;
            ^
abort() at Error
    at stackTrace (/Users/john/dev/asmjs/nimcache/a.out.js:1050:15)
    at abort (/Users/john/dev/asmjs/nimcache/a.out.js:6063:25)
    at _systemDatInit (/Users/john/dev/asmjs/nimcache/a.out.js:5488:58)
    at _PreMain (/Users/john/dev/asmjs/nimcache/a.out.js:5703:2)
    at _NimMain (/Users/john/dev/asmjs/nimcache/a.out.js:5728:2)
    at Object._main (/Users/john/dev/asmjs/nimcache/a.out.js:5773:2)
    at Object.callMain (/Users/john/dev/asmjs/nimcache/a.out.js:5954:30)
    at doRun (/Users/john/dev/asmjs/nimcache/a.out.js:6011:25)
    at run (/Users/john/dev/asmjs/nimcache/a.out.js:6026:5)
    at Object.<anonymous> (/Users/john/dev/asmjs/nimcache/a.out.js:6083:1)

Please let me know if I'm doing anything wrong.

Araq (orginal) [2014-07-26T10:37:50+02:00] view original

Well I don't know how emcc works but hello.c is not enough, you need at least stdlib_system.c (or maybe system.c) too.

fuzzthink (orginal) [2014-07-27T04:31:01+02:00] view original

Ah, I see. Including the needed c files in emcc command works. Let's see how far I can take this ;)

Mirror of forum.nim-lang.org

506 :: nimrod and the C compiler disagree on the size of a pointer