AddressSanitizer

[Hyxogen]

This PR implements (basic) AddressSanitizer functionality for cheerp-wasm.

For now I'll leave this as a draft, as I'm still planning some minor changes, need to setup some other stuff for the ci, and would also like to get some feedback.

Getting started

Docker

If you want to use it right away, you can use the following docker image: hyxogen/cheerp-asan:latest. I'll update this docker image with the latest changes from my development branch

Cheerp will be located in /opt/cheerp, and you'll also have the following aliases set:

alias cheerp="/opt/cheerp/bin/clang"
alias cheerp++="/opt/cheerp/bin/clang++"

Building it yourself

If you want to build it on your own system, then build the entire cheerp toolchain like normal with compiler-rt at the end. Please note that you'll need the latest versions from all the cheerp repositories. And if you haven't been keeping up with them, you'll have to do a rebuild of them all, lest you want a lot of errors.

building compiler-rt
To be able to use asan, you'll have to compile and install the asan runtime library.

Build instructions:

cd compiler-rt
cmake -DCMAKE_INSTALL_PREFIX="$CHEERP_DEST" -B build -C CheerpCmakeConf.cmake \
	-DCMAKE_TOOLCHAIN_FILE="$CHEERP_DEST/share/cmake/Modules/CheerpWasmToolchain.cmake" .
make -C build install # NOTE: Parallel builds do NOT work

If for some reason the above command does not work, you're probably using an old version of clang (<14.0). You should
be able to still setup with the following (assuming that you've already built cheerp, and that the cheerp build
directory is build at the root of the repo)

cmake -DCMAKE_INSTALL_PREFIX="$CHEERP_DEST" -B build -C CheerpCmakeConf.cmake \
    -DLLVM_DIR="$PWD/../build/lib/cmake/llvm"
	-DCMAKE_TOOLCHAIN_FILE="$CHEERP_DEST/share/cmake/Modules/CheerpWasmToolchain.cmake" .

Usage

Once installed, address sanitizer should work (pretty much) like on native. An example:

# cat test.c
int main()
{
  int *p = 0;
  return *p;
}
# "$CHEERP_DEST/bin/clang" -fsanitize=address -cheerp-pretty-code test.c
# node a.out
=================================================================
==1==ERROR: AddressSanitizer: unknown-crash on address 0x00000000 at pc 0x000192b6 bp 0x00000000 sp 0x00100ffc
READ of size 4 at 0x00000000 thread T0
    #0 0x192b6 in main (main+0x192b6)
    #1 0x1f2c3 in _start (main+0x1f2c3)
    #2 0x80000145 in <unknown function> main:325

Address 0x00000000 is a wild pointer inside of access range of size 0x00000004.
SUMMARY: AddressSanitizer: unknown-crash (main+0x192b6) in main
Shadow bytes around the buggy address:
=>0x00000000:[fe]fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe
  0x00000080: fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe
  0x00000100: fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe
  0x00000180: fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe
  0x00000200: fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe
  0x00000280: fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe
Shadow byte legend (one shadow byte represents 8 application bytes):
  Addressable:           00
  Partially addressable: 01 02 03 04 05 06 07
  Heap left redzone:       fa
  Freed heap region:       fd
  Stack left redzone:      f1
  Stack mid redzone:       f2
  Stack right redzone:     f3
  Stack after return:      f5
  Stack use after scope:   f8
  Global redzone:          f9
  Global init order:       f6
  Poisoned by user:        f7
  Container overflow:      fc
  Array cookie:            ac
  Intra object redzone:    bb
  ASan internal:           fe
  Left alloca redzone:     ca
  Right alloca redzone:    cb
==1==ABORTING

Two step compilation

If you're doing compiling and linking separately, you'll have to specify the -fsanitize=address flag for both steps

Notes

LeakSanitizer

For LeakSanitizer to trigger, you'll have to manually call exit, as code compiled with cheerp doesn't call the atexit
callbacks when returning from main.

LeakSanitizer is not perfect.
It simply scans the memory if there are any pointers still pointing to allocated memory. Thus it might not find leaks
for cheerp-wasm which it can find for native.

Function "x" does not detect memory errors

ASan can only detect memory errors in instrumented code or intercepted functions. If you're using a library, you'll have
to compile it with ASan as well for it to be able to pick up these errors.

As for the intercepted functions, currently ASan only intercepts a handful of libc functions, thus it might be the case
you're using one that isn't intercepted.

More reading

If you want to find out how AddressSanitizer works in depth:
https://github.com/google/sanitizers/wiki/AddressSanitizer

EDIT: Implemented asan flags and exceptions support

[Hyxogen]

I'm sure that something in the CI will not work, but this PR should be ready