Mirror of forum.nim-lang.org

700 :: compiling nimrod by tcc

• seyko2 (orginal) [2014-12-31T03:01:10+01:00] view original

  A current version of the nimrod is crashing when compiled with tcc. A version 0.8.8 of the nimrod compiler was fine. Any suggestions about what wrong with tcc?

  PS: a csource generated by nimrod must include only one file -- nimbase.h. This will allow a speed up a compiling, something like tcc -E *.c | gcc -o nim -xc -

---

• Varriount (orginal) [2014-12-31T06:08:33+01:00] view original

  If by 'tcc' you mean the 'tiny c compiler', support for that was dropped a whole ago.

---

• seyko2 (orginal) [2014-12-31T11:59:21+01:00] view original

  Yes, a question is abount a tiny C compiler. Most precisely about compiling a csource of the nimrod by tcc. Cuurrent state is difenitely depend on the nimrod new features. tcc can succesfully compile a linux kernel 2.4.26 and mamy others projects. If it fails, then there is some unusual features which triggers a bug in tcc. I think this can be triggered by a changed GC or something similar. It is very interesting why fail a nimrod compiled by tcc. Any assumptions?

---

• Araq (orginal) [2014-12-31T15:59:26+01:00] view original

  You can try the versions between 0.8.8 and 0.10.2 too. Perhaps that helps to figure out the problem.

  Any assumptions?

  No idea.

---

• seyko2 (orginal) [2015-01-01T05:31:08+01:00] view original

  I will try to bisect a git://github.com/Araq/Nim.git. May be this helps to locate a problem. Currently it looks like a problem appear after throwing away a rod code.

  PS: nim is nice as a command name. But a nimrod as a language name is better for me.

---

• seyko2 (orginal) [2015-01-03T15:04:33+01:00] view original

  csources/build.sh: using "uname -m" for ARCH detection is a _BAD thing when a 64bit kernel runs a 32bit userspace. It is better to detect ARCH of the gcc compiler

---

• def (orginal) [2015-01-03T15:40:41+01:00] view original

  csources/build.sh: using "uname -m" for ARCH detection is a _BAD thing when a 64bit kernel runs a 32bit userspace. It is better to detect ARCH of the gcc compiler

  From what I can find by googling many build systems behave in the same way. Wouldn't it be better to use "setarch linux32" on your system?

---

• Varriount (orginal) [2015-01-03T23:51:45+01:00] view original

  def-: Assuming seyko2 is running Linux, and not some other OS

---

• seyko2 (orginal) [2015-01-04T17:32:36+01:00] view original

  Nice, It is a first time I hear about this program. About compiling a nim by tinyc. If replace a -2147483648LL with -2147483647LL then this allow to get a working nim compiler. tynic fails to compare (-2147483648LL < 0) but do it right for (-2147483647LL < 0)

---

• seyko2 (orginal) [2015-01-08T08:48:53+01:00] view original

  A current git version of the tcc with patches which can be found in a tcc-mailing list (https://lists.nongnu.org/mailman/listinfo/tinycc-devel, current set of patches which I want to upload) can succesfully be used for building a nimrod compiler

---

• Araq (orginal) [2015-01-08T10:17:13+01:00] view original

  Excellent! There are plans to use as the new foundation of our (currently unusable) REPL, so Windows support is important. I have some bad memories with Tiny C on Windows, has there been any progress in the meantime?

---

• seyko2 (orginal) [2015-01-08T14:29:43+01:00] view original

  I think yes. crt1.c was rewritten and there are people on the tynicc forum which are checking a compatibility of the patches with a Windows platform.

---

• Nikki (orginal) [2015-01-09T00:07:17+01:00] view original

  @seyko2 Which versions of Windows?

---