gollvm: build fails in combination with -fcf-protection=branch

[sitilge]

Does this issue reproduce with the latest release?

Yes (with latest commit to master).

What operating system and processor architecture are you using (go env)?

Ubuntu Focal Fossa, amd64

What did you do?

I followed the instructions on the gollvm landing page.

What did you expect to see?

The project builds cleanly.

What did you see instead?

I got this output

-- Could NOT find LibXml2 (missing: LIBXML2_LIBRARY LIBXML2_INCLUDE_DIR) 
-- Native target architecture is X86
-- Threads enabled.
-- Doxygen disabled.
-- Go bindings disabled.
-- Ninja version: 1.10.0
-- Could NOT find OCaml (missing: OCAMLFIND OCAML_VERSION OCAML_STDLIB_PATH) 
-- Could NOT find OCaml (missing: OCAMLFIND OCAML_VERSION OCAML_STDLIB_PATH) 
-- OCaml bindings disabled.
-- LLVM host triple: x86_64-unknown-linux-gnu
-- LLVM default target triple: x86_64-unknown-linux-gnu
-- Building with -fPIC
-- Constructing LLVMBuild project information
-- Targeting AArch64
-- Targeting AMDGPU
-- Targeting ARM
-- Targeting AVR
-- Targeting BPF
-- Targeting Hexagon
-- Targeting Lanai
-- Targeting Mips
-- Targeting MSP430
-- Targeting NVPTX
-- Targeting PowerPC
-- Targeting RISCV
-- Targeting Sparc
-- Targeting SystemZ
-- Targeting WebAssembly
-- Targeting X86
-- Targeting XCore
CMake Error at tools/gollvm/cmake/modules/AddGollvm.cmake:37 (message):
  C compiler does not support -fsplit-stack
Call Stack (most recent call first):
  tools/gollvm/CMakeLists.txt:21 (include)


-- starting libgo configuration.
CMake Error at tools/gollvm/cmake/modules/LibbacktraceUtils.cmake:19 (message):
  Support for mmap() is required -- setup failed.
Call Stack (most recent call first):
  tools/gollvm/libgo/CMakeLists.txt:45 (setup_libbacktrace)


CMake Error at tools/gollvm/cmake/modules/LibffiUtils.cmake:27 (message):
  Support for mmap() is required -- setup failed.
Call Stack (most recent call first):
  tools/gollvm/libgo/CMakeLists.txt:49 (setup_libffi)


-- Libgo: creating stdlib package targets
-- Libgo: generating check targets
-- libgo configuration complete.
-- starting gotools configuration.
-- gotools: generating check targets
-- gotools configuration complete.
-- Registering Bye as a pass plugin (static build: OFF)
-- Failed to find LLVM FileCheck
-- Version: 0.0.0
-- Performing Test HAVE_GNU_POSIX_REGEX -- failed to compile
-- Performing Test HAVE_POSIX_REGEX -- success
-- Performing Test HAVE_STEADY_CLOCK -- success
-- Configuring incomplete, errors occurred!
See also "/home/gopher/workarea/build-debug/CMakeFiles/CMakeOutput.log".
See also "/home/gopher/workarea/build-debug/CMakeFiles/CMakeError.log".
ninja: error: loading 'build.ninja': No such file or directory

The output of CMakeError.log is around 2.5k lines. Here is a snippet:

Determining if the include file malloc/malloc.h exists failed with the following output:
Change Dir: /home/gopher/workarea/build-debug/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/ninja cmTC_4ce15 && [1/2] Building C object CMakeFiles/cmTC_4ce15.dir/CheckIncludeFile.c.o
FAILED: CMakeFiles/cmTC_4ce15.dir/CheckIncludeFile.c.o 
/usr/bin/cc    -o CMakeFiles/cmTC_4ce15.dir/CheckIncludeFile.c.o   -c CheckIncludeFile.c
CheckIncludeFile.c:1:10: fatal error: malloc/malloc.h: No such file or directory
    1 | #include <malloc/malloc.h>
      |          ^~~~~~~~~~~~~~~~~
compilation terminated.
ninja: build stopped: subcommand failed.

Determining if the include file valgrind/valgrind.h exists failed with the following output:
Change Dir: /home/gopher/workarea/build-debug/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/ninja cmTC_d0b71 && [1/2] Building C object CMakeFiles/cmTC_d0b71.dir/CheckIncludeFile.c.o
FAILED: CMakeFiles/cmTC_d0b71.dir/CheckIncludeFile.c.o 
/usr/bin/cc    -o CMakeFiles/cmTC_d0b71.dir/CheckIncludeFile.c.o   -c CheckIncludeFile.c
CheckIncludeFile.c:1:10: fatal error: valgrind/valgrind.h: No such file or directory
    1 | #include <valgrind/valgrind.h>
      |          ^~~~~~~~~~~~~~~~~~~~~
compilation terminated.
ninja: build stopped: subcommand failed.

Has anyone managed to build gollvm? Or have a Docker image? The Docker images publicly available are somewhat old and also gave me bugs.

[ALTree]

Hi,

in theory the issue tracker is not for questions or support requests (just Go bugs), but this could be a bug in the makefiles I guess? So leaving open for investigation

cc @thanm @cherrymui

[thanm]

As mentioned previously, this sort of question is better off sent to the go-nuts mailing list, as opposed creating an issue.

Looking at your cmake output, it seems likely that there is some issue with the C compiler you are using. What happens when you do this:

$ cat > /tmp/example.c <<EOF
#include <sys/mman.h>
int x;
EOF
$ cc -c -o /tmp/example.o /tmp/example.c
$

Does this succeed for you?

[sitilge]

@thanm Yes, it succeeds.

cc (Ubuntu 9.3.0-10ubuntu2) 9.3.0

Edit: I accidentally submitted the wrong cc version before. The real version is 9.3.0. The script succeeds.

[thanm]

Update: with help of @sitilge , who ran some experiments on the build system, it was eventually determined that the system GCC is hard-wiring on the "-fcf-protection=branch" option (or equivalent). This results in function prologs that can't be recognized by ld.gold. Typical function prolog without -cf-protection=branch:

   0: 64 48 3b 24 25 70 00 cmp    %fs:0x70,%rsp
   7: 00 00
   9: 73 12                 jae    1d <main+0x1d>

and prolog with -cf-protection=branch:

   0:	f3 0f 1e fa          	endbr64 
   4:	64 48 3b 24 25 70 00 	cmp    %fs:0x70,%rsp
   b:	00 00 
   d:	73 16                	jae    25 <main+0x25>

The additional 'endbr64' is confusing gold's pattern match.

Next step is to file an upstream issue for the ld.gold problem. Will hold this issue open until I have more info.

[thanm]

Sent upstream binutils bug here.

[hdost]

Does it mean that this issue should be closed?

[thanm]

@hdost not yet, I still have a couple of CLs to submit. Should be able to close it out in a day or two.