[M3commit] [modula3/cm3] 37b167: possible techniques to be like autoconf but differ...

jaykrell jay.krell at cornell.edu
Tue Aug 11 13:36:22 CEST 2015

  Branch: refs/heads/master
  Home:   https://github.com/modula3/cm3
  Commit: 37b167b2d8fce7e38f32f33d9d1bbb8a816bcafb
  Author: jaykrell <jay.krell at cornell.edu>
  Date:   2015-08-10 (Mon, 10 Aug 2015)

  Changed paths:
    A scratch/jconf/jconf1.c
    A scratch/jconf/jconf1.sh

  Log Message:
  possible techniques to be like autoconf but different, and not nearly as general

  Commit: 56a7834bd4e023e599db1edc3c3828efd35ffcaa
  Author: jaykrell <jay.krell at cornell.edu>
  Date:   2015-08-11 (Tue, 11 Aug 2015)

  Changed paths:
    A m3-libs/m3core/src/runtime/common/RT0.h

  Log Message:
  C++ version of RT0.i3 might be useful, though I ended up not using it.

  Commit: 8eb697cca2b6c6c45226612e6ca52105d8fecf3b
  Author: jaykrell <jay.krell at cornell.edu>
  Date:   2015-08-11 (Tue, 11 Aug 2015)

  Changed paths:
    M m3-sys/cminstall/src/config-no-install/Darwin.common
    M m3-sys/m3cc/gcc-4.7/gcc/config/darwin.c

  Log Message:
  Switch Darwing config to use "as" instead of "gcc -x assembler" to assemble.
gcc on 10.0.4 Yosemite is clang and runs the clang assembler which has a big problem
if any non-lazy pointers precede any stubs -- stubs are hooked up incorrectly and
the wrong functions are called.
The non-LLVM assembler doesn't have this problem.

I also attempted dropping support for 10.4 which should have worked but didn't.

 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67183 (suggest gcc output in llvm-as-compatible order)
 https://llvm.org/bugs/show_bug.cgi?id=24428 (clearly a bug)

This configuration change might break some things. All the information around
what hosts have what assemblers supporting what targets is not well contained.
This is perhaps why gcc made a nice assembler driver.

Compare: https://github.com/modula3/cm3/compare/79655166d3ba...8eb697cca2b6

More information about the M3commit mailing list