[LLVMdev] LLVM/Clang SVN: 2 Questions

[Kevin Kelley]

On 5/31/2012 8:52 AM, Justin Holewinski wrote:
> On Thu, May 31, 2012 at 1:44 AM, MortenMacFly <mac-fly at gmx.net 
> <mailto:mac-fly at gmx.net>> wrote:
>
>
>     Windows7, Cmake 2.8.8, MinGW 4.6.1. compiler:
>
>     I got two questions seeking for help:
>
>     1.) I am trying to compile LLVM/Clang from trunk using the
>     instructions from
>     the LLVM homepage which works fine except that the llvm-config
>     tool is not
>     being built. When I do a "make" in the "tools\llvm-config" folder
>     it fails,
>     too with:
>     ../../Makefile.common:60: ../../Makefile.config: No such file or
>     directory
>     ../../Makefile.common:68: /Makefile.rules: No such file or directory
>     mingw32-make.exe: *** No rule to make target `/Makefile.rules'.  Stop.
>     Those files are there, so I am a bit lost. Is there a
>     configuration switch I
>     am missing to build this tool? I searched through the Cmake
>     options, but I
>     could not find any.
>
>
> The llvm-config tool is not built on Windows, since it requires sed 
> which is not available by default on Windows.  I actually have a patch 
> for this somewhere (build llvm-config on Windows if sed is found), 
> I'll try to find it and post it.

This "No Makefile.rules" problem is different, and recent, from what I see.

With mingw, msys, and gnuwin32, on win7, then configure and make
will normally succeed, including building llvm-config.exe.

I've run into this "No Makefile.rules" problem, and as best I know
so far, it's tied to the change in the llvm.org Getting Started procedure:
now it's called for to checkout LLVM, checkout clang, AND
checkout Compiler-RT.  I followed that path a week ago and hit
the same breakage described above.

I'm now going thru procedures one at a time;
svn checkout LLVM and configure and make, succeeds;
repeat, adding "--enable-shared" to configure; succeeds;
svn checkout clang and repeat: both static and shared succeed;

...and that's how far I've got.  Am waiting for compiler-rt
right now, to find out if it re-introduces the breakage, or
if something has been changed in trunk in the last few
days to cause different behavior.

I agree with Mikael that we do need an up-to-date, tested,
preferably buildbotted, Windows build procedure.  I think
it doesn't matter too much what tool choices are made,
most project setups start with "download this set of tools"
anyway.  But there needs to be at least one path that can
be followed to success, and that path needs to be tested,
and known to be up-to-date.


cheers,
Kevin

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20120531/8ad49f0c/attachment.html>