Linking Haskell with C: The Wrong Way

Lately I've been messing with linking .o files produced by compilers... a lot. It all started when I wanted to make an SDL project in D instead of C, which required me to make bindings for the library. That mess of a project got me thinking: what other languages can I link? C and D were very easy to implement, but what about others? To test this I rewrote my statusbar script (originally written in Rust) in a combination of C, D, and Rust, and everything (mostly) worked without really that much effort. Of course, after this I got curious: how many languages could I link at once?

I created a new project, at the timed named simply "o", and started adding languages to it. C, D, C++, Rust, Objective C, everything started out pretty well, with only a few hiccups here and there. By the time I got to Ruby, I decided I wanted to try Haskell. To quickly sum up my goal, I wanted to take these two snippets of code and combine them into one executable
-- func.hs
module S where
foreign export ccall haskell :: IO ()
haskell :: IO ()
haskell = putStrLn "Sent from Haskell"
// main.c
int main() {
    return 0;
Doing a search for "embed Haskell in C" generates a bunch of results, some helpful, some not. The Haskell wiki's solution uses GHC's ForeignFunctionInterface, which sounded promising but seemed to only support compiling your C with GHC, which was out of the question for this project as I had other languages than C to add. Compiling func.hs with ghc func.hs produces a func.o and func_stub.h but I couldn't make much use of these at the time (more on this later). I found a few examples of people trying to link this object, with the best example I could find being this, which unfortunately seemed like a "Works on My Machine" moment.

After trying what felt like everything, I compiled a Hello, World! Haskell program with the -v option and noticed that it actually calls cc at the end with an absolutely enormous mess of compiler flags:
cc '-fuse-ld=gold' -Wl,--no-as-needed -o hw -no-pie -fno-PIC -Wl,--gc-sections hw.o -L/usr/lib/ghc-9.0.2/text- -L/usr/lib/ghc-9.0.2/integer-gmp-1.1 -L/usr/lib/ghc-9.0.2/ghc-9.0.2 -L/usr/lib/ghc-9.0.2/terminfo- -L/usr/lib/ghc-9.0.2/process- -L/usr/lib/ghc-9.0.2/hpc- -L/usr/lib/ghc-9.0.2/ghci-9.0.2 -L/usr/lib/ghc-9.0.2/ghc-heap-9.0.2 -L/usr/lib/ghc-9.0.2/ghc-boot-9.0.2 -L/usr/lib/ghc-9.0.2/exceptions-0.10.4 -L/usr/lib/ghc-9.0.2/template-haskell- -L/usr/lib/ghc-9.0.2/pretty- -L/usr/lib/ghc-9.0.2/ghc-boot-th-9.0.2 -L/usr/lib/ghc-9.0.2/stm- -L/usr/lib/ghc-9.0.2/mtl-2.2.2 -L/usr/lib/ghc-9.0.2/transformers- -L/usr/lib/ghc-9.0.2/directory- -L/usr/lib/ghc-9.0.2/unix- -L/usr/lib/ghc-9.0.2/time-1.9.3 -L/usr/lib/ghc-9.0.2/filepath- -L/usr/lib/ghc-9.0.2/binary- -L/usr/lib/ghc-9.0.2/containers- -L/usr/lib/ghc-9.0.2/bytestring- -L/usr/lib/ghc-9.0.2/deepseq- -L/usr/lib/ghc-9.0.2/array- -L/usr/lib/ghc-9.0.2/base- -L/usr/lib/ghc-9.0.2/ghc-bignum-1.1 -L/usr/lib/ghc-9.0.2/ghc-prim-0.7.0 -L/usr/lib/ghc-9.0.2/rts /tmp/ghc104289_0/ghc_5.o /tmp/ghc104289_0/ghc_8.o -Wl,-u,base_GHCziTopHandler_runIO_closure -Wl,-u,base_GHCziTopHandler_runNonIO_closure -Wl,-u,ghczmprim_GHCziTuple_Z0T_closure -Wl,-u,ghczmprim_GHCziTypes_True_closure -Wl,-u,ghczmprim_GHCziTypes_False_closure -Wl,-u,base_GHCziPack_unpackCString_closure -Wl,-u,base_GHCziWeak_runFinalizzerBatch_closure -Wl,-u,base_GHCziIOziException_stackOverflow_closure -Wl,-u,base_GHCziIOziException_heapOverflow_closure -Wl,-u,base_GHCziIOziException_allocationLimitExceeded_closure -Wl,-u,base_GHCziIOziException_blockedIndefinitelyOnMVar_closure -Wl,-u,base_GHCziIOziException_blockedIndefinitelyOnSTM_closure -Wl,-u,base_GHCziIOziException_cannotCompactFunction_closure -Wl,-u,base_GHCziIOziException_cannotCompactPinned_closure -Wl,-u,base_GHCziIOziException_cannotCompactMutable_closure -Wl,-u,base_GHCziIOPort_doubleReadException_closure -Wl,-u,base_ControlziExceptionziBase_nonTermination_closure -Wl,-u,base_ControlziExceptionziBase_nestedAtomically_closure -Wl,-u,base_GHCziEventziThread_blockedOnBadFD_closure -Wl,-u,base_GHCziExceptionziType_divZZeroException_closure -Wl,-u,base_GHCziExceptionziType_underflowException_closure -Wl,-u,base_GHCziExceptionziType_overflowException_closure -Wl,-u,base_GHCziConcziSync_runSparks_closure -Wl,-u,base_GHCziConcziIO_ensureIOManagerIsRunning_closure -Wl,-u,base_GHCziConcziIO_interruptIOManager_closure -Wl,-u,base_GHCziConcziIO_ioManagerCapabilitiesChanged_closure -Wl,-u,base_GHCziConcziSignal_runHandlersPtr_closure -Wl,-u,base_GHCziTopHandler_flushStdHandles_closure -Wl,-u,base_GHCziTopHandler_runMainIO_closure -Wl,-u,ghczmprim_GHCziTypes_Czh_con_info -Wl,-u,ghczmprim_GHCziTypes_Izh_con_info -Wl,-u,ghczmprim_GHCziTypes_Fzh_con_info -Wl,-u,ghczmprim_GHCziTypes_Dzh_con_info -Wl,-u,ghczmprim_GHCziTypes_Wzh_con_info -Wl,-u,base_GHCziPtr_Ptr_con_info -Wl,-u,base_GHCziPtr_FunPtr_con_info -Wl,-u,base_GHCziInt_I8zh_con_info -Wl,-u,base_GHCziInt_I16zh_con_info -Wl,-u,base_GHCziInt_I32zh_con_info -Wl,-u,base_GHCziInt_I64zh_con_info -Wl,-u,base_GHCziWord_W8zh_con_info -Wl,-u,base_GHCziWord_W16zh_con_info -Wl,-u,base_GHCziWord_W32zh_con_info -Wl,-u,base_GHCziWord_W64zh_con_info -Wl,-u,base_GHCziStable_StablePtr_con_info -Wl,-u,hs_atomic_add8 -Wl,-u,hs_atomic_add16 -Wl,-u,hs_atomic_add32 -Wl,-u,hs_atomic_add64 -Wl,-u,hs_atomic_sub8 -Wl,-u,hs_atomic_sub16 -Wl,-u,hs_atomic_sub32 -Wl,-u,hs_atomic_sub64 -Wl,-u,hs_atomic_and8 -Wl,-u,hs_atomic_and16 -Wl,-u,hs_atomic_and32 -Wl,-u,hs_atomic_and64 -Wl,-u,hs_atomic_nand8 -Wl,-u,hs_atomic_nand16 -Wl,-u,hs_atomic_nand32 -Wl,-u,hs_atomic_nand64 -Wl,-u,hs_atomic_or8 -Wl,-u,hs_atomic_or16 -Wl,-u,hs_atomic_or32 -Wl,-u,hs_atomic_or64 -Wl,-u,hs_atomic_xor8 -Wl,-u,hs_atomic_xor16 -Wl,-u,hs_atomic_xor32 -Wl,-u,hs_atomic_xor64 -Wl,-u,hs_cmpxchg8 -Wl,-u,hs_cmpxchg16 -Wl,-u,hs_cmpxchg32 -Wl,-u,hs_cmpxchg64 -Wl,-u,hs_xchg8 -Wl,-u,hs_xchg16 -Wl,-u,hs_xchg32 -Wl,-u,hs_xchg64 -Wl,-u,hs_atomicread8 -Wl,-u,hs_atomicread16 -Wl,-u,hs_atomicread32 -Wl,-u,hs_atomicread64 -Wl,-u,hs_atomicwrite8 -Wl,-u,hs_atomicwrite16 -Wl,-u,hs_atomicwrite32 -Wl,-u,hs_atomicwrite64 -lHStext- -lHSinteger-gmp-1.1 -lHSghc-9.0.2 -lHSterminfo- -lHSprocess- -lHShpc- -lHSghci-9.0.2 -lHSghc-heap-9.0.2 -lHSghc-boot-9.0.2 -lHSexceptions-0.10.4 -lHStemplate-haskell- -lHSpretty- -lHSghc-boot-th-9.0.2 -lHSstm- -lHSmtl-2.2.2 -lHStransformers- -lHSdirectory- -lHSunix- -lHStime-1.9.3 -lHSfilepath- -lHSbinary- -lHScontainers- -lHSbytestring- -lHSdeepseq- -lHSarray- -lHSbase- -lHSghc-bignum-1.1 -lHSghc-prim-0.7.0 -lHSrts -ltinfo -lgmp -lc -lm -lm -lrt -ldl -lffi -lnuma

I instantly noticed that it was compiling a lot of files in my /tmp so I made a very hacky script with fswatch to copy my entire tmp folder into the local directory, then realized that ghc's -tmpdir and -keep-tmp-files exist. The directory it generates looks like this:
├── ghc_2.s
├── ghc_3.c
├── ghc_4.s
├── ghc_5.rsp
├── ghc_6.o
├── ghc_7.o
└── ghc_8.ldscript
1 directory, 7 files
Compiling the objects here with our main.c with the flags that GHC specifies generates us a valid executable! Great!

...Except not really. If we run the resulting binary, we get an error:
newBoundTask: RTS is not initialised; call hs_init() first
As you would expect, we can call hs_init(&argc, &argv); and our code runs without error. However, this wasn't good enough for me, so I used nm to list the symbols, export those symbols into a C file while also calling hs_init(), then add them both to a static archive. In my amalgamation project (Haskell is language #19), the Makefile target is this:
build/s.a: src/s.hs 
	ghc -tmpdir helpers -keep-tmp-files -no-keep-hi-files -no-keep-o-files src/s.hs 
	{ \
		printf "_hs_init = 0;"; \
		nm src/s.o | \
			grep -E '^[0-9a-f]{16} T' | \
			awk '{print $$3}' | \
			xargs -I {} echo 'void {}() {  if(!_hs_init){_hs_init=1;int argc = 1;char **argv = (char **)malloc(sizeof(char *));argv[0] = strdup("hi");hs_init(&argc, &argv);}export_{}();}'; \
	} > helpers/funcs.c 
	gcc -c -o helpers/funcs.o helpers/funcs.c
	ar r build/s.a helpers/ghc*/*.o
	nm src/s.o | grep -E '^[0-9a-f]{16} T' | awk '{print $$3}' | xargs -I {} objcopy --redefine-sym {}=export_{} build/s.a build/s.a
	ar r build/s.a helpers/funcs.o
	rm -f helpers/ghc* src/*.o src/*.h src/*.hi
Yes, it's a chaotic mess but it works! You can find the rest of the Makefile here

After I had this solution implemented into my project and working well, I realized you could also just directly compile the object generated by GHC when creating an executable... oops :)

Anyways that's my journey of linking Haskell into C; I hope I never have to do this again