llvm dev - Nov 2018 - Errors in configuration for LLVM with CMake GUI

I probably need GCC for Git bash.  And the one I have is MinGW's GCC, which
is the one for Windows.  I don't know why CMake is trying to find Linux
headers.  Hopefully someone who knows better what to do will weigh in here.

@Jonathan Goodwin<mailto:jondgoodwin at gmail.com> , are you sure you
didn't have to tell it where the header files for Windows are?  If I need to
specify the paths to those headers, how do I do it?
________________________________
From: Jonathan Goodwin <jondgoodwin at gmail.com>
Sent: Thursday, November 29, 2018 10:42 AM
To: Osman Zakir
Cc: llvm-dev at lists.llvm.org
Subject: Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI

I checked your CMakeError.log again. The .h include files it is marking as
missing are not LLVM include files, but are OS-related (e.g., malloc/malloc.h),
and in some cases appear to be potentially Linux ones (dlfcn.h, sys/ioctl,
malloc_np.h). I can't imagine why your build might be looking for Linux as
opposed to Windows include files. That could be a clue to your problem.

You said you installed the latest version of GCC before this attempt. This
confuses me further, since you are using Visual Studio to build with, and that
makes no use of GCC. In fact, honestly I am wondering if your GCC implementation
libraries are somehow in your search path wrt environment variables, and your
build is actually trying to use the GNU/Linux header files rather than the VS
Windows header files, which would certainly blow it up. Again, I am no expert
here, so I don't have a definite answer, but based on what little I know,
this setup doesn't sound safe. Maybe get rid of GCC completely, as I am not
sure when/how you would even use it on the Windows side.

Similarly, I don't believe Perl has anything to do with this either, as
neither CMake nor VS make any use of it. But I can't imagine it causing any
problems for you to have it.

I don't know anything about Backtrace, Libomp, Compiler-RT, so I cannot help
you there. As we have discussed before, the source for clang and lld go into the
llvm source folder's "tools" folder by those names. If they are
there, after the configure step of the CMake-Gui, options will automatically
appear for both clang and lld. You will want to mouse click the checkbox next to
CLANG_BUILD_TOOLS and LLD_BUILD_TOOLS prior to hitting the generate button.

Sorry I don't know more than that, but hopefully that will help some ... and
then others can chime in if they know more about this than me.

On Thu, Nov 29, 2018 at 2:52 PM Osman Zakir via llvm-dev <llvm-dev at
lists.llvm.org<mailto:llvm-dev at lists.llvm.org>> wrote:
I tried to configure CMake for LLVM using the GUI and I've attached the logs
for the attempt to this message.  I installed Perl and the latest version of GCC
before that attempt.  There were still errors that stopped it from being able to
generate project files, though.

I'd like to know how I can get this to work.  How do I tell it where to find
Backtrace, Libomp, Compiler-RT, LLD and Clang?  And where exactly is Backtrace? 
I copy-pasted the locations for the stuff I checked out for LLVM, so I'm not
really sure where the stuff I need to put into CMake for Compiler-RT, Backtrace
and Libomp is (it's where it should be, that's all I know).

_______________________________________________
LLVM Developers mailing list
llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20181129/d10138a8/attachment.html>

Also, what executable's path do I put in for SVN?  I just use SVN by
right-clicking and then clicking on what I need in the context menu that comes
up, since SVN is integrated into my context menus too.  But I don't know
what .exe file it's using.  And because of that I don't know what path I
need to give to it so it can find it.

For some reason, it's trying to find Nuget as well.  I'll try to find
path to it and put it there.
________________________________
From: Osman Zakir <osmanzakir90 at hotmail.com>
Sent: Thursday, November 29, 2018 8:43 PM
To: llvm-dev at lists.llvm.org; Jonathan Goodwin
Subject: Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI

I probably need GCC for Git bash.  And the one I have is MinGW's GCC, which
is the one for Windows.  I don't know why CMake is trying to find Linux
headers.  Hopefully someone who knows better what to do will weigh in here.

@Jonathan Goodwin<mailto:jondgoodwin at gmail.com> , are you sure you
didn't have to tell it where the header files for Windows are?  If I need to
specify the paths to those headers, how do I do it?
________________________________
From: Jonathan Goodwin <jondgoodwin at gmail.com>
Sent: Thursday, November 29, 2018 10:42 AM
To: Osman Zakir
Cc: llvm-dev at lists.llvm.org
Subject: Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI

I checked your CMakeError.log again. The .h include files it is marking as
missing are not LLVM include files, but are OS-related (e.g., malloc/malloc.h),
and in some cases appear to be potentially Linux ones (dlfcn.h, sys/ioctl,
malloc_np.h). I can't imagine why your build might be looking for Linux as
opposed to Windows include files. That could be a clue to your problem.

You said you installed the latest version of GCC before this attempt. This
confuses me further, since you are using Visual Studio to build with, and that
makes no use of GCC. In fact, honestly I am wondering if your GCC implementation
libraries are somehow in your search path wrt environment variables, and your
build is actually trying to use the GNU/Linux header files rather than the VS
Windows header files, which would certainly blow it up. Again, I am no expert
here, so I don't have a definite answer, but based on what little I know,
this setup doesn't sound safe. Maybe get rid of GCC completely, as I am not
sure when/how you would even use it on the Windows side.

Similarly, I don't believe Perl has anything to do with this either, as
neither CMake nor VS make any use of it. But I can't imagine it causing any
problems for you to have it.

I don't know anything about Backtrace, Libomp, Compiler-RT, so I cannot help
you there. As we have discussed before, the source for clang and lld go into the
llvm source folder's "tools" folder by those names. If they are
there, after the configure step of the CMake-Gui, options will automatically
appear for both clang and lld. You will want to mouse click the checkbox next to
CLANG_BUILD_TOOLS and LLD_BUILD_TOOLS prior to hitting the generate button.

Sorry I don't know more than that, but hopefully that will help some ... and
then others can chime in if they know more about this than me.

On Thu, Nov 29, 2018 at 2:52 PM Osman Zakir via llvm-dev <llvm-dev at
lists.llvm.org<mailto:llvm-dev at lists.llvm.org>> wrote:
I tried to configure CMake for LLVM using the GUI and I've attached the logs
for the attempt to this message.  I installed Perl and the latest version of GCC
before that attempt.  There were still errors that stopped it from being able to
generate project files, though.

I'd like to know how I can get this to work.  How do I tell it where to find
Backtrace, Libomp, Compiler-RT, LLD and Clang?  And where exactly is Backtrace? 
I copy-pasted the locations for the stuff I checked out for LLVM, so I'm not
really sure where the stuff I need to put into CMake for Compiler-RT, Backtrace
and Libomp is (it's where it should be, that's all I know).

_______________________________________________
LLVM Developers mailing list
llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20181129/43d6edc2/attachment.html>

I think you are heading down the wrong path. All of those error messages
are normal. That’s just cmake doing feature detection, for example it will
test whether your system has pthread.h by compiling a program that includes
that header. So an error in your log about is missing pthread.h just means
it will set HAVE_PTHREAD_H to false which is fine.

I’m confused why we’re talking about gcc and libomp if you’re trying to do
a Windows build. You should just use msvc to build it. You don’t need any
of the things you mentioned to do a build.

I still haven’t seen the output from cmake (just a paste of stdout). That’s
going to tell you the real problem. The pog files I don’t think are very
useful and will just lead you in the wrong direction.

You said you’re using cmake gui, maybe the problem is that the gui can’t
find the Visual Studio installation. Did you run vcvarsall first? Try first
running vcvarsall from a command prompt and tgen running cmake (command
line or gui) from that same command prompt
On Thu, Nov 29, 2018 at 7:55 AM Osman Zakir via llvm-dev <
llvm-dev at lists.llvm.org> wrote:
> Also, what executable's path do I put in for SVN?  I just use SVN by
> right-clicking and then clicking on what I need in the context menu that
> comes up, since SVN is integrated into my context menus too.  But I
don't
> know what .exe file it's using.  And because of that I don't know
what path
> I need to give to it so it can find it.
>
> For some reason, it's trying to find Nuget as well.  I'll try to
find path
> to it and put it there.
> ------------------------------
> *From:* Osman Zakir <osmanzakir90 at hotmail.com>
> *Sent:* Thursday, November 29, 2018 8:43 PM
> *To:* llvm-dev at lists.llvm.org; Jonathan Goodwin
>
> *Subject:* Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI
>
> I probably need GCC for Git bash.  And the one I have is MinGW's GCC,
> which is the one for Windows.  I don't know why CMake is trying to find
> Linux headers.  Hopefully someone who knows better what to do will weigh in
> here.
>
> @Jonathan Goodwin <jondgoodwin at gmail.com> , are you sure you
didn't have
> to tell it where the header files for Windows are?  If I need to specify
> the paths to those headers, how do I do it?
> ------------------------------
> *From:* Jonathan Goodwin <jondgoodwin at gmail.com>
> *Sent:* Thursday, November 29, 2018 10:42 AM
> *To:* Osman Zakir
> *Cc:* llvm-dev at lists.llvm.org
> *Subject:* Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI
>
> I checked your CMakeError.log again. The .h include files it is marking as
> missing are not LLVM include files, but are OS-related (e.g.,
> malloc/malloc.h), and in some cases appear to be potentially Linux ones
> (dlfcn.h, sys/ioctl, malloc_np.h). I can't imagine why your build might
be
> looking for Linux as opposed to Windows include files. That could be a clue
> to your problem.
>
> You said you installed the latest version of GCC before this attempt. This
> confuses me further, since you are using Visual Studio to build with, and
> that makes no use of GCC. In fact, honestly I am wondering if your GCC
> implementation libraries are somehow in your search path wrt environment
> variables, and your build is actually trying to use the GNU/Linux header
> files rather than the VS Windows header files, which would certainly blow
> it up. Again, I am no expert here, so I don't have a definite answer,
but
> based on what little I know, this setup doesn't sound safe. Maybe get
rid
> of GCC completely, as I am not sure when/how you would even use it on the
> Windows side.
>
> Similarly, I don't believe Perl has anything to do with this either, as
> neither CMake nor VS make any use of it. But I can't imagine it causing
any
> problems for you to have it.
>
> I don't know anything about Backtrace, Libomp, Compiler-RT, so I cannot
> help you there. As we have discussed before, the source for clang and lld
> go into the llvm source folder's "tools" folder by those
names. If they are
> there, after the configure step of the CMake-Gui, options will
> automatically appear for both clang and lld. You will want to mouse click
> the checkbox next to CLANG_BUILD_TOOLS and LLD_BUILD_TOOLS prior to hitting
> the generate button.
>
> Sorry I don't know more than that, but hopefully that will help some
...
> and then others can chime in if they know more about this than me.
>
> On Thu, Nov 29, 2018 at 2:52 PM Osman Zakir via llvm-dev <
> llvm-dev at lists.llvm.org> wrote:
>
> I tried to configure CMake for LLVM using the GUI and I've attached the
> logs for the attempt to this message.  I installed Perl and the latest
> version of GCC before that attempt.  There were still errors that stopped
> it from being able to generate project files, though.
>
> I'd like to know how I can get this to work.  How do I tell it where to
> find Backtrace, Libomp, Compiler-RT, LLD and Clang?  And where exactly is
> Backtrace?  I copy-pasted the locations for the stuff I checked out for
> LLVM, so I'm not really sure where the stuff I need to put into CMake
for
> Compiler-RT, Backtrace and Libomp is (it's where it should be,
that's all I
> know).
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20181129/d02b0b6f/attachment.html>

pthread.h is the standard threads library on unix-like systems, but it
definitely doesn't exist on windows unless you are building with MinGW,
which you are not (because you said you're trying to use the Visual Studio
generator).

You don't need to put anything for any of the variables you specify.  You
also don't need to put anything for the linker.  As long as you run
vcvarsall.bat ahead of time, cmake should be able to find everything it
needs to compile.  FWIW, this is my cmake command line when i use the
visual studio generator using the mono-repo source-tree layout

cmake -G "Visual Studio 15 2017" -Thost=x64
-DLLVM_ENABLE_PROJECTS=clang;lld -DLLVM_TARGETS_TO_BUILD=X86
path/to/source/dir.

You can adjust the variables as necessary for your targets.

However, I never use an installed LLVM/clang.  If you're trying to build
llvm / clang from sources you shouldn't even need an installed llvm/clang,
because you can use MSVC to build it.  You can install a fresh image of
Windows, then install nothing but git, visual studio, and cmake, and you
should be able to bootstrap clang and llvm from sources.

On Thu, Nov 29, 2018 at 9:34 AM Osman Zakir <osmanzakir90 at hotmail.com>
wrote:
> After it finishes trying to configure, it says that there were errors with
> the configure step and that program files may be invalid.  And I also
don't
> know what executable to point it to for SVN.
>
> What should I put for CLANG_DEFAULT_CXX_STDLIB, CLANG_DEFAULT_LINKER,
> CLANG_DEFAULT_RTLIB, CLANG_DEFAULT_STD_C and CLANG_DEFAULT_STD_CXX?  Would
> it be okay to provide the path to the MSVC include directory for the
> CLANG_DEFAULT_CXX_STDLIB, or should I provide the path to the include
> directory in the libcxx folder in the LLVM source tree?  For the Linker I
> think I'll just put llvm_root/tools/lld, if that's good enough.
>
> As for pthread, shouldn't I have that?  I mean, it's the standard
threads
> library, isn't it?  And CMake also can't find the header that
malloc is in,
> among some other headers.
>
> And what is the CLANG_VENDOR setting asking for?  The vendor that I got
> Clang from?  And what about CLANG_VENDOR_UTI?
>
> CMAKE_INSTALL_OLDINCLUDEDIR is set to the Linux-specific /usr/local for
> some reason.  What should I set it to?
>
> Note: @Zachary Turner <zturner at google.com> I wrote this before I
saw your
> latest message.  I'll try that now.  Thanks.
> ------------------------------
> *From:* Zachary Turner <zturner at google.com>
> *Sent:* Thursday, November 29, 2018 9:49 PM
>
> *To:* Osman Zakir
> *Cc:* llvm-dev at lists.llvm.org
> *Subject:* Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI
> I think you are heading down the wrong path. All of those error messages
> are normal. That’s just cmake doing feature detection, for example it will
> test whether your system has pthread.h by compiling a program that includes
> that header. So an error in your log about is missing pthread.h just means
> it will set HAVE_PTHREAD_H to false which is fine.
>
> I’m confused why we’re talking about gcc and libomp if you’re trying to do
> a Windows build. You should just use msvc to build it. You don’t need any
> of the things you mentioned to do a build.
>
> I still haven’t seen the output from cmake (just a paste of stdout).
> That’s going to tell you the real problem. The pog files I don’t think are
> very useful and will just lead you in the wrong direction.
>
> You said you’re using cmake gui, maybe the problem is that the gui can’t
> find the Visual Studio installation. Did you run vcvarsall first? Try first
> running vcvarsall from a command prompt and tgen running cmake (command
> line or gui) from that same command prompt
> On Thu, Nov 29, 2018 at 7:55 AM Osman Zakir via llvm-dev <
> llvm-dev at lists.llvm.org> wrote:
>
> Also, what executable's path do I put in for SVN?  I just use SVN by
> right-clicking and then clicking on what I need in the context menu that
> comes up, since SVN is integrated into my context menus too.  But I
don't
> know what .exe file it's using.  And because of that I don't know
what path
> I need to give to it so it can find it.
>
> For some reason, it's trying to find Nuget as well.  I'll try to
find path
> to it and put it there.
> ------------------------------
> *From:* Osman Zakir <osmanzakir90 at hotmail.com>
> *Sent:* Thursday, November 29, 2018 8:43 PM
> *To:* llvm-dev at lists.llvm.org; Jonathan Goodwin
>
> *Subject:* Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI
>
> I probably need GCC for Git bash.  And the one I have is MinGW's GCC,
> which is the one for Windows.  I don't know why CMake is trying to find
> Linux headers.  Hopefully someone who knows better what to do will weigh in
> here.
>
> @Jonathan Goodwin <jondgoodwin at gmail.com> , are you sure you
didn't have
> to tell it where the header files for Windows are?  If I need to specify
> the paths to those headers, how do I do it?
> ------------------------------
> *From:* Jonathan Goodwin <jondgoodwin at gmail.com>
> *Sent:* Thursday, November 29, 2018 10:42 AM
> *To:* Osman Zakir
> *Cc:* llvm-dev at lists.llvm.org
> *Subject:* Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI
>
> I checked your CMakeError.log again. The .h include files it is marking as
> missing are not LLVM include files, but are OS-related (e.g.,
> malloc/malloc.h), and in some cases appear to be potentially Linux ones
> (dlfcn.h, sys/ioctl, malloc_np.h). I can't imagine why your build might
be
> looking for Linux as opposed to Windows include files. That could be a clue
> to your problem.
>
> You said you installed the latest version of GCC before this attempt. This
> confuses me further, since you are using Visual Studio to build with, and
> that makes no use of GCC. In fact, honestly I am wondering if your GCC
> implementation libraries are somehow in your search path wrt environment
> variables, and your build is actually trying to use the GNU/Linux header
> files rather than the VS Windows header files, which would certainly blow
> it up. Again, I am no expert here, so I don't have a definite answer,
but
> based on what little I know, this setup doesn't sound safe. Maybe get
rid
> of GCC completely, as I am not sure when/how you would even use it on the
> Windows side.
>
> Similarly, I don't believe Perl has anything to do with this either, as
> neither CMake nor VS make any use of it. But I can't imagine it causing
any
> problems for you to have it.
>
> I don't know anything about Backtrace, Libomp, Compiler-RT, so I cannot
> help you there. As we have discussed before, the source for clang and lld
> go into the llvm source folder's "tools" folder by those
names. If they are
> there, after the configure step of the CMake-Gui, options will
> automatically appear for both clang and lld. You will want to mouse click
> the checkbox next to CLANG_BUILD_TOOLS and LLD_BUILD_TOOLS prior to hitting
> the generate button.
>
> Sorry I don't know more than that, but hopefully that will help some
...
> and then others can chime in if they know more about this than me.
>
> On Thu, Nov 29, 2018 at 2:52 PM Osman Zakir via llvm-dev <
> llvm-dev at lists.llvm.org> wrote:
>
> I tried to configure CMake for LLVM using the GUI and I've attached the
> logs for the attempt to this message.  I installed Perl and the latest
> version of GCC before that attempt.  There were still errors that stopped
> it from being able to generate project files, though.
>
> I'd like to know how I can get this to work.  How do I tell it where to
> find Backtrace, Libomp, Compiler-RT, LLD and Clang?  And where exactly is
> Backtrace?  I copy-pasted the locations for the stuff I checked out for
> LLVM, so I'm not really sure where the stuff I need to put into CMake
for
> Compiler-RT, Backtrace and Libomp is (it's where it should be,
that's all I
> know).
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20181129/c6fa4b64/attachment-0001.html>

I opened up an elevated Developer Command Prompt for VS2017, ran vcvarsall with
argument amd64, and then ran the command "cmake-gui" in that command
prompt.  Then in that gui window, I clicked on "Configure" without
doing anything else.  At the end, I got an error window saying that there were
errors, and that project files may be invalid.

Here's the output from CMake:
"

Could NOT find Backtrace (missing: Backtrace_LIBRARY Backtrace_INCLUDE_DIR)

Native target architecture is X86

Threads enabled.

Doxygen disabled.

Go bindings disabled.

Could NOT find OCaml (missing: OCAMLFIND OCAML_VERSION OCAML_STDLIB_PATH)

OCaml bindings disabled.

Could NOT find Python module pygments

Could NOT find Python module pygments.lexers.c_cpp

Could NOT find Python module yaml

LLVM host triple: x86_64-pc-windows-msvc

LLVM default target triple: x86_64-pc-windows-msvc

Using Debug VC++ CRT: MDd

Using Release VC++ CRT: MD

Using MinSizeRel VC++ CRT: MD

Using RelWithDebInfo VC++ CRT: MD

Constructing LLVMBuild project information

CMake Error at C:/Program
Files/CMake/share/cmake-3.13/Modules/FindSubversion.cmake:123 (message):
Command "C:/Program Files/TortoiseSVN/bin/ info C:/llvm" failed with
output:

Call Stack (most recent call first):
cmake/modules/VersionFromVCS.cmake:19 (subversion_wc_info)
CMakeLists.txt:757 (add_version_info_from_vcs)


LLVMHello ignored -- Loadable modules not supported on this platform.

Targeting AArch64

Targeting AMDGPU

Targeting ARM

Targeting BPF

Targeting Hexagon

Targeting Lanai

Targeting Mips

Targeting MSP430

Targeting NVPTX

Targeting PowerPC

Targeting Sparc

Targeting SystemZ

Targeting WebAssembly

Targeting X86

Targeting XCore

Compiler-RT supported architectures: x86_64

Builtin supported architectures: x86_64

Builtin supported architectures: x86_64

CMake Error at C:/Program
Files/CMake/share/cmake-3.13/Modules/FindPackageHandleStandardArgs.cmake:137
(message):
Could NOT find Perl (missing: PERL_EXECUTABLE)
Call Stack (most recent call first):
C:/Program
Files/CMake/share/cmake-3.13/Modules/FindPackageHandleStandardArgs.cmake:378
(_FPHSA_FAILURE_MESSAGE)
C:/Program Files/CMake/share/cmake-3.13/Modules/FindPerl.cmake:75
(FIND_PACKAGE_HANDLE_STANDARD_ARGS)
projects/openmp/runtime/cmake/config-ix.cmake:147 (find_package)
projects/openmp/runtime/CMakeLists.txt:234 (include)


Configuring incomplete, errors occurred!

See also "C:/llvm/build/CMakeFiles/CMakeOutput.log".

See also "C:/llvm/build/CMakeFiles/CMakeError.log".

"

If I want Clang to use its own C++ standard libraries, with LLD as the linker,
instead of MSVC's variants, what should I do?  Tell me that too please. 
Thanks.
________________________________
From: Osman Zakir <osmanzakir90 at hotmail.com>
Sent: Thursday, November 29, 2018 10:33 PM
To: Zachary Turner
Subject: Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI

After it finishes trying to configure, it says that there were errors with the
configure step and that program files may be invalid.  And I also don't know
what executable to point it to for SVN.

What should I put for CLANG_DEFAULT_CXX_STDLIB, CLANG_DEFAULT_LINKER,
CLANG_DEFAULT_RTLIB, CLANG_DEFAULT_STD_C and CLANG_DEFAULT_STD_CXX?  Would it be
okay to provide the path to the MSVC include directory for the
CLANG_DEFAULT_CXX_STDLIB, or should I provide the path to the include directory
in the libcxx folder in the LLVM source tree?  For the Linker I think I'll
just put llvm_root/tools/lld, if that's good enough.

As for pthread, shouldn't I have that?  I mean, it's the standard
threads library, isn't it?  And CMake also can't find the header that
malloc is in, among some other headers.

And what is the CLANG_VENDOR setting asking for?  The vendor that I got Clang
from?  And what about CLANG_VENDOR_UTI?

CMAKE_INSTALL_OLDINCLUDEDIR is set to the Linux-specific /usr/local for some
reason.  What should I set it to?

Note: @Zachary Turner<mailto:zturner at google.com> I wrote this before I
saw your latest message.  I'll try that now.  Thanks.
________________________________
From: Zachary Turner <zturner at google.com>
Sent: Thursday, November 29, 2018 9:49 PM
To: Osman Zakir
Cc: llvm-dev at lists.llvm.org
Subject: Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI

I think you are heading down the wrong path. All of those error messages are
normal. That’s just cmake doing feature detection, for example it will test
whether your system has pthread.h by compiling a program that includes that
header. So an error in your log about is missing pthread.h just means it will
set HAVE_PTHREAD_H to false which is fine.

I’m confused why we’re talking about gcc and libomp if you’re trying to do a
Windows build. You should just use msvc to build it. You don’t need any of the
things you mentioned to do a build.

I still haven’t seen the output from cmake (just a paste of stdout). That’s
going to tell you the real problem. The pog files I don’t think are very useful
and will just lead you in the wrong direction.

You said you’re using cmake gui, maybe the problem is that the gui can’t find
the Visual Studio installation. Did you run vcvarsall first? Try first running
vcvarsall from a command prompt and tgen running cmake (command line or gui)
from that same command prompt
On Thu, Nov 29, 2018 at 7:55 AM Osman Zakir via llvm-dev <llvm-dev at
lists.llvm.org<mailto:llvm-dev at lists.llvm.org>> wrote:
Also, what executable's path do I put in for SVN?  I just use SVN by
right-clicking and then clicking on what I need in the context menu that comes
up, since SVN is integrated into my context menus too.  But I don't know
what .exe file it's using.  And because of that I don't know what path I
need to give to it so it can find it.

For some reason, it's trying to find Nuget as well.  I'll try to find
path to it and put it there.
________________________________
From: Osman Zakir <osmanzakir90 at hotmail.com<mailto:osmanzakir90 at
hotmail.com>>
Sent: Thursday, November 29, 2018 8:43 PM
To: llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>;
Jonathan Goodwin

Subject: Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI

I probably need GCC for Git bash.  And the one I have is MinGW's GCC, which
is the one for Windows.  I don't know why CMake is trying to find Linux
headers.  Hopefully someone who knows better what to do will weigh in here.

@Jonathan Goodwin<mailto:jondgoodwin at gmail.com> , are you sure you
didn't have to tell it where the header files for Windows are?  If I need to
specify the paths to those headers, how do I do it?
________________________________
From: Jonathan Goodwin <jondgoodwin at gmail.com<mailto:jondgoodwin at
gmail.com>>
Sent: Thursday, November 29, 2018 10:42 AM
To: Osman Zakir
Cc: llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>
Subject: Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI

I checked your CMakeError.log again. The .h include files it is marking as
missing are not LLVM include files, but are OS-related (e.g., malloc/malloc.h),
and in some cases appear to be potentially Linux ones (dlfcn.h, sys/ioctl,
malloc_np.h). I can't imagine why your build might be looking for Linux as
opposed to Windows include files. That could be a clue to your problem.

You said you installed the latest version of GCC before this attempt. This
confuses me further, since you are using Visual Studio to build with, and that
makes no use of GCC. In fact, honestly I am wondering if your GCC implementation
libraries are somehow in your search path wrt environment variables, and your
build is actually trying to use the GNU/Linux header files rather than the VS
Windows header files, which would certainly blow it up. Again, I am no expert
here, so I don't have a definite answer, but based on what little I know,
this setup doesn't sound safe. Maybe get rid of GCC completely, as I am not
sure when/how you would even use it on the Windows side.

Similarly, I don't believe Perl has anything to do with this either, as
neither CMake nor VS make any use of it. But I can't imagine it causing any
problems for you to have it.

I don't know anything about Backtrace, Libomp, Compiler-RT, so I cannot help
you there. As we have discussed before, the source for clang and lld go into the
llvm source folder's "tools" folder by those names. If they are
there, after the configure step of the CMake-Gui, options will automatically
appear for both clang and lld. You will want to mouse click the checkbox next to
CLANG_BUILD_TOOLS and LLD_BUILD_TOOLS prior to hitting the generate button.

Sorry I don't know more than that, but hopefully that will help some ... and
then others can chime in if they know more about this than me.

On Thu, Nov 29, 2018 at 2:52 PM Osman Zakir via llvm-dev <llvm-dev at
lists.llvm.org<mailto:llvm-dev at lists.llvm.org>> wrote:
I tried to configure CMake for LLVM using the GUI and I've attached the logs
for the attempt to this message.  I installed Perl and the latest version of GCC
before that attempt.  There were still errors that stopped it from being able to
generate project files, though.

I'd like to know how I can get this to work.  How do I tell it where to find
Backtrace, Libomp, Compiler-RT, LLD and Clang?  And where exactly is Backtrace? 
I copy-pasted the locations for the stuff I checked out for LLVM, so I'm not
really sure where the stuff I need to put into CMake for Compiler-RT, Backtrace
and Libomp is (it's where it should be, that's all I know).

_______________________________________________
LLVM Developers mailing list
llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
_______________________________________________
LLVM Developers mailing list
llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20181129/1969c4a5/attachment.html>

This output is much more interesting.  Can you show the full cmake command
line or set of variables that you configured?

1. I don't know why it's saying it requires Perl.  Perl is absolutely
not a
requirement
2. I don't know why it's trying to build openmp or compiler-rt.  I guess
it
depends on how you configured CMake.  openmp I don't think is well
supported (perhaps not even supported at all) on Windows
3. I also don't know why it would be trying to find subversion.  However,
even if it is trying to find subversion, "C:/Program Files/TortoiseSVN/bin/
info C:/llvm" is a meaningless command line, because there is no executable
there.  It looks like it is trying to "run" a directory.

So these are the problems you should be looking at I think.  Can you try it
from a command line, and not the cmake gui?  For example, you might start
by trying the exact command line I posted earlier.  Even if it's not
suitable for your needs, I know for certain that command line works for me,
so see if it works for you.  If it does not work, there's a problem with
your installation, system, cmake version, etc.  If it does work, then it's
more likely a problem with the cmake command line you're running.

On Thu, Nov 29, 2018 at 9:45 AM Osman Zakir <osmanzakir90 at hotmail.com>
wrote:
> I opened up an elevated Developer Command Prompt for VS2017, ran vcvarsall
> with argument amd64, and then ran the command "cmake-gui" in that
command
> prompt.  Then in that gui window, I clicked on "Configure"
without doing
> anything else.  At the end, I got an error window saying that there were
> errors, and that project files may be invalid.
>
> Here's the output from CMake:
> "
>
> Could NOT find Backtrace (missing: Backtrace_LIBRARY
> Backtrace_INCLUDE_DIR)
>
> Native target architecture is X86
>
> Threads enabled.
>
> Doxygen disabled.
>
> Go bindings disabled.
>
> Could NOT find OCaml (missing: OCAMLFIND OCAML_VERSION OCAML_STDLIB_PATH)
>
> OCaml bindings disabled.
>
> Could NOT find Python module pygments
>
> Could NOT find Python module pygments.lexers.c_cpp
>
> Could NOT find Python module yaml
>
> LLVM host triple: x86_64-pc-windows-msvc
>
> LLVM default target triple: x86_64-pc-windows-msvc
>
> Using Debug VC++ CRT: MDd
>
> Using Release VC++ CRT: MD
>
> Using MinSizeRel VC++ CRT: MD
>
> Using RelWithDebInfo VC++ CRT: MD
>
> Constructing LLVMBuild project information
>
> CMake Error at C:/Program
> Files/CMake/share/cmake-3.13/Modules/FindSubversion.cmake:123 (message):
> Command "C:/Program Files/TortoiseSVN/bin/ info C:/llvm" failed
with
> output:
>
> Call Stack (most recent call first):
> cmake/modules/VersionFromVCS.cmake:19 (subversion_wc_info)
> CMakeLists.txt:757 (add_version_info_from_vcs)
>
> LLVMHello ignored -- Loadable modules not supported on this platform.
>
> Targeting AArch64
>
> Targeting AMDGPU
>
> Targeting ARM
>
> Targeting BPF
>
> Targeting Hexagon
>
> Targeting Lanai
>
> Targeting Mips
>
> Targeting MSP430
>
> Targeting NVPTX
>
> Targeting PowerPC
>
> Targeting Sparc
>
> Targeting SystemZ
>
> Targeting WebAssembly
>
> Targeting X86
>
> Targeting XCore
>
> Compiler-RT supported architectures: x86_64
>
> Builtin supported architectures: x86_64
>
> Builtin supported architectures: x86_64
>
> CMake Error at C:/Program
>
Files/CMake/share/cmake-3.13/Modules/FindPackageHandleStandardArgs.cmake:137
> (message):
> Could NOT find Perl (missing: PERL_EXECUTABLE)
> Call Stack (most recent call first):
> C:/Program
>
Files/CMake/share/cmake-3.13/Modules/FindPackageHandleStandardArgs.cmake:378
> (_FPHSA_FAILURE_MESSAGE)
> C:/Program Files/CMake/share/cmake-3.13/Modules/FindPerl.cmake:75
> (FIND_PACKAGE_HANDLE_STANDARD_ARGS)
> projects/openmp/runtime/cmake/config-ix.cmake:147 (find_package)
> projects/openmp/runtime/CMakeLists.txt:234 (include)
>
> Configuring incomplete, errors occurred!
>
> See also "C:/llvm/build/CMakeFiles/CMakeOutput.log".
>
> See also "C:/llvm/build/CMakeFiles/CMakeError.log".
> "
>
> If I want Clang to use its own C++ standard libraries, with LLD as the
> linker, instead of MSVC's variants, what should I do?  Tell me that too
> please.  Thanks.
> ------------------------------
> *From:* Osman Zakir <osmanzakir90 at hotmail.com>
> *Sent:* Thursday, November 29, 2018 10:33 PM
> *To:* Zachary Turner
>
> *Subject:* Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI
> After it finishes trying to configure, it says that there were errors with
> the configure step and that program files may be invalid.  And I also
don't
> know what executable to point it to for SVN.
>
> What should I put for CLANG_DEFAULT_CXX_STDLIB, CLANG_DEFAULT_LINKER,
> CLANG_DEFAULT_RTLIB, CLANG_DEFAULT_STD_C and CLANG_DEFAULT_STD_CXX?  Would
> it be okay to provide the path to the MSVC include directory for the
> CLANG_DEFAULT_CXX_STDLIB, or should I provide the path to the include
> directory in the libcxx folder in the LLVM source tree?  For the Linker I
> think I'll just put llvm_root/tools/lld, if that's good enough.
>
> As for pthread, shouldn't I have that?  I mean, it's the standard
threads
> library, isn't it?  And CMake also can't find the header that
malloc is in,
> among some other headers.
>
> And what is the CLANG_VENDOR setting asking for?  The vendor that I got
> Clang from?  And what about CLANG_VENDOR_UTI?
>
> CMAKE_INSTALL_OLDINCLUDEDIR is set to the Linux-specific /usr/local for
> some reason.  What should I set it to?
>
> Note: @Zachary Turner <zturner at google.com> I wrote this before I
saw your
> latest message.  I'll try that now.  Thanks.
> ------------------------------
> *From:* Zachary Turner <zturner at google.com>
> *Sent:* Thursday, November 29, 2018 9:49 PM
> *To:* Osman Zakir
> *Cc:* llvm-dev at lists.llvm.org
> *Subject:* Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI
>
> I think you are heading down the wrong path. All of those error messages
> are normal. That’s just cmake doing feature detection, for example it will
> test whether your system has pthread.h by compiling a program that includes
> that header. So an error in your log about is missing pthread.h just means
> it will set HAVE_PTHREAD_H to false which is fine.
>
> I’m confused why we’re talking about gcc and libomp if you’re trying to do
> a Windows build. You should just use msvc to build it. You don’t need any
> of the things you mentioned to do a build.
>
> I still haven’t seen the output from cmake (just a paste of stdout).
> That’s going to tell you the real problem. The pog files I don’t think are
> very useful and will just lead you in the wrong direction.
>
> You said you’re using cmake gui, maybe the problem is that the gui can’t
> find the Visual Studio installation. Did you run vcvarsall first? Try first
> running vcvarsall from a command prompt and tgen running cmake (command
> line or gui) from that same command prompt
> On Thu, Nov 29, 2018 at 7:55 AM Osman Zakir via llvm-dev <
> llvm-dev at lists.llvm.org> wrote:
>
> Also, what executable's path do I put in for SVN?  I just use SVN by
> right-clicking and then clicking on what I need in the context menu that
> comes up, since SVN is integrated into my context menus too.  But I
don't
> know what .exe file it's using.  And because of that I don't know
what path
> I need to give to it so it can find it.
>
> For some reason, it's trying to find Nuget as well.  I'll try to
find path
> to it and put it there.
> ------------------------------
> *From:* Osman Zakir <osmanzakir90 at hotmail.com>
> *Sent:* Thursday, November 29, 2018 8:43 PM
> *To:* llvm-dev at lists.llvm.org; Jonathan Goodwin
>
> *Subject:* Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI
>
> I probably need GCC for Git bash.  And the one I have is MinGW's GCC,
> which is the one for Windows.  I don't know why CMake is trying to find
> Linux headers.  Hopefully someone who knows better what to do will weigh in
> here.
>
> @Jonathan Goodwin <jondgoodwin at gmail.com> , are you sure you
didn't have
> to tell it where the header files for Windows are?  If I need to specify
> the paths to those headers, how do I do it?
> ------------------------------
> *From:* Jonathan Goodwin <jondgoodwin at gmail.com>
> *Sent:* Thursday, November 29, 2018 10:42 AM
> *To:* Osman Zakir
> *Cc:* llvm-dev at lists.llvm.org
> *Subject:* Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI
>
> I checked your CMakeError.log again. The .h include files it is marking as
> missing are not LLVM include files, but are OS-related (e.g.,
> malloc/malloc.h), and in some cases appear to be potentially Linux ones
> (dlfcn.h, sys/ioctl, malloc_np.h). I can't imagine why your build might
be
> looking for Linux as opposed to Windows include files. That could be a clue
> to your problem.
>
> You said you installed the latest version of GCC before this attempt. This
> confuses me further, since you are using Visual Studio to build with, and
> that makes no use of GCC. In fact, honestly I am wondering if your GCC
> implementation libraries are somehow in your search path wrt environment
> variables, and your build is actually trying to use the GNU/Linux header
> files rather than the VS Windows header files, which would certainly blow
> it up. Again, I am no expert here, so I don't have a definite answer,
but
> based on what little I know, this setup doesn't sound safe. Maybe get
rid
> of GCC completely, as I am not sure when/how you would even use it on the
> Windows side.
>
> Similarly, I don't believe Perl has anything to do with this either, as
> neither CMake nor VS make any use of it. But I can't imagine it causing
any
> problems for you to have it.
>
> I don't know anything about Backtrace, Libomp, Compiler-RT, so I cannot
> help you there. As we have discussed before, the source for clang and lld
> go into the llvm source folder's "tools" folder by those
names. If they are
> there, after the configure step of the CMake-Gui, options will
> automatically appear for both clang and lld. You will want to mouse click
> the checkbox next to CLANG_BUILD_TOOLS and LLD_BUILD_TOOLS prior to hitting
> the generate button.
>
> Sorry I don't know more than that, but hopefully that will help some
...
> and then others can chime in if they know more about this than me.
>
> On Thu, Nov 29, 2018 at 2:52 PM Osman Zakir via llvm-dev <
> llvm-dev at lists.llvm.org> wrote:
>
> I tried to configure CMake for LLVM using the GUI and I've attached the
> logs for the attempt to this message.  I installed Perl and the latest
> version of GCC before that attempt.  There were still errors that stopped
> it from being able to generate project files, though.
>
> I'd like to know how I can get this to work.  How do I tell it where to
> find Backtrace, Libomp, Compiler-RT, LLD and Clang?  And where exactly is
> Backtrace?  I copy-pasted the locations for the stuff I checked out for
> LLVM, so I'm not really sure where the stuff I need to put into CMake
for
> Compiler-RT, Backtrace and Libomp is (it's where it should be,
that's all I
> know).
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20181129/e683fdf6/attachment.html>

"  are you sure you didn't have to tell it where the header files for
Windows are?"

You should not have to and I do not. The point of CMake's configure step is
that it probes your system to find out what tools to use and where to find
libraries and include files. Because you specify use of the Visual Studio
generator, it uses Microsoft's conventions to locate where the compiler,
linker, libraries and include files are stored. That should be an automatic
part of the Cmake 'configure' step. Which is why I was confused by it
even
looking for non-Windows include files, and then failing the configure step.

With regard to the executable path for SVN or Nuget (or anything else), I
cannot think of why you need to specify it for CMake to build LLVM. I see
the error messages you posted, but in reality the LLVM build process should
not need access to SVN to access the files, as all of the LLVM files should
be viewable by CMake or any other program as normal files in the file
system. You only need SVN-aware tools when doing change management
activities or looking at earlier history. CMake is only working on the
current version of the files which should be 100% viewable as normal files.
Thus, CMake needs no knowledge of the SVN executable. As someone else said,
you might be better off using the git repo version of LLVM instead.

With regard to Perl, that appears to be failing because you are trying to
build OpenMP as a project as part of the LLVM build. My strong
recommendation is not to complicate your problematic build process out of
the gate. I would recommend you get a vanilla build working first (with
clang and lld), before trying to handle projects like backtrace, compiler-rt
and openmp. And you may want to build them separately anyway, following
their respective build processes.

Wrt needing gcc for Git bash, I don't understand why that would be
necessary. I don't have it and I use Git bash just fine. I understand that
gcc can be made to work on Windows by jumping through some hoops, but it
seems like you plan to have three very different toolsets that do
essentially the same things:  Visual Studio, LLVM w/ clang/lld, and GCC.
Personally, that's too complicated for my tastes as there is too much
potential for colliding include files in PATHs and such, but you may have
your reasons. When I want to play with GCC, I usually just do it on a Linux
box.

You said you use this command:

cmake -G "Visual Studio 15 2017" -Thost=x64
-DLLVM_ENABLE_PROJECTS=clang;lld
-DLLVM_EXPERIMENTAL_TARGETS_TO_BUILD=WebAssembly
-DLLVM_TARGETS_TO_BUILD=X86 .

To build clang and lld, I would think you would need something like this,
as I mentioned earlier:

cmake -G "Visual Studio 15 2017" -Thost=x64 -D CLANG_BUILD_TOOLS -DLLD
_BUILD_TOOLS  -DLLVM_EXPERIMENTAL_TARGETS_TO_BUILD=WebAssembly
-DLLVM_TARGETS_TO_BUILD=X86 .

Hopefully by using the Git version of the LLVM source and stripping out the
irrelevant extra projects, you will get further along ...


On Fri, Nov 30, 2018 at 1:43 AM Osman Zakir <osmanzakir90 at hotmail.com>
wrote:
> I probably need GCC for Git bash.  And the one I have is MinGW's GCC,
> which is the one for Windows.  I don't know why CMake is trying to find
> Linux headers.  Hopefully someone who knows better what to do will weigh in
> here.
>
> @Jonathan Goodwin <jondgoodwin at gmail.com> , are you sure you
didn't have
> to tell it where the header files for Windows are?  If I need to specify
> the paths to those headers, how do I do it?
> ------------------------------
> *From:* Jonathan Goodwin <jondgoodwin at gmail.com>
> *Sent:* Thursday, November 29, 2018 10:42 AM
> *To:* Osman Zakir
> *Cc:* llvm-dev at lists.llvm.org
> *Subject:* Re: [llvm-dev] Errors in configuration for LLVM with CMake GUI
>
> I checked your CMakeError.log again. The .h include files it is marking as
> missing are not LLVM include files, but are OS-related (e.g.,
> malloc/malloc.h), and in some cases appear to be potentially Linux ones
> (dlfcn.h, sys/ioctl, malloc_np.h). I can't imagine why your build might
be
> looking for Linux as opposed to Windows include files. That could be a clue
> to your problem.
>
> You said you installed the latest version of GCC before this attempt. This
> confuses me further, since you are using Visual Studio to build with, and
> that makes no use of GCC. In fact, honestly I am wondering if your GCC
> implementation libraries are somehow in your search path wrt environment
> variables, and your build is actually trying to use the GNU/Linux header
> files rather than the VS Windows header files, which would certainly blow
> it up. Again, I am no expert here, so I don't have a definite answer,
but
> based on what little I know, this setup doesn't sound safe. Maybe get
rid
> of GCC completely, as I am not sure when/how you would even use it on the
> Windows side.
>
> Similarly, I don't believe Perl has anything to do with this either, as
> neither CMake nor VS make any use of it. But I can't imagine it causing
any
> problems for you to have it.
>
> I don't know anything about Backtrace, Libomp, Compiler-RT, so I cannot
> help you there. As we have discussed before, the source for clang and lld
> go into the llvm source folder's "tools" folder by those
names. If they are
> there, after the configure step of the CMake-Gui, options will
> automatically appear for both clang and lld. You will want to mouse click
> the checkbox next to CLANG_BUILD_TOOLS and LLD_BUILD_TOOLS prior to hitting
> the generate button.
>
> Sorry I don't know more than that, but hopefully that will help some
...
> and then others can chime in if they know more about this than me.
>
> On Thu, Nov 29, 2018 at 2:52 PM Osman Zakir via llvm-dev <
> llvm-dev at lists.llvm.org> wrote:
>
> I tried to configure CMake for LLVM using the GUI and I've attached the
> logs for the attempt to this message.  I installed Perl and the latest
> version of GCC before that attempt.  There were still errors that stopped
> it from being able to generate project files, though.
>
> I'd like to know how I can get this to work.  How do I tell it where to
> find Backtrace, Libomp, Compiler-RT, LLD and Clang?  And where exactly is
> Backtrace?  I copy-pasted the locations for the stuff I checked out for
> LLVM, so I'm not really sure where the stuff I need to put into CMake
for
> Compiler-RT, Backtrace and Libomp is (it's where it should be,
that's all I
> know).
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20181130/2cfa7e70/attachment.html>

With regard to the executable path for SVN or Nuget (or anything else), I cannot
think of why you need to specify it for CMake to build LLVM.

CMake uses the version control system (SVN or git) to derive part of the version
string for the build.  Whichever one you use, the location of the tool should be
on the PATH so that simple "svn status" or "git status"
commands will work.

FTR I'm using Visual Studio 2015, with the git "mono repo" and
this is my CMake invocation:

cmake -G "Visual Studio 14 Win64" -Thost=x64
-DLLVM_ENABLE_ASSERTIONS=ON ^
  -DLLVM_ENABLE_PROJECTS="llvm;clang;lld" path\to\llvm

As previously noted, if you are following the SVN instructions, and put your
clang and lld checkouts under llvm\tools, then you should omit the
LLVM_ENABLE_PROJECTS definition.
--paulr
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20181129/f198e3ab/attachment.html>