Reimplement resolve_path in JS NFC

[ryanking13]

As discussed in #23872, this PR reimplements resolve_path function, which is used to locate the library using LD_LIBRARY_PATH in JavaScript, so that it can be reused inside libdylink.js.

Currently, this code doesn't work in thread.

[ryanking13]

@hoodmane @sbc100 This is a split off from #23872, but I am having trouble making the function work in thread. Can I get some help with this? It seems like FS.lookupPath (JS) fails to locate the file in thread environment, while stat (C) does.

[hoodmane]

Is there a ci failure for this problem? Or which test should I run? I only see the sigs job failing which needs an update to the expected output.

[ryanking13]

This one fails

./test/runner other.test_ld_library_path_pthread

while the same test (without pthread)

./test/runner other.test_ld_library_path

success

[hoodmane]

Well applying this change, I can see that stat finds the file but lookupPath does not:

diff --git a/test/test_other.py b/test/test_other.py
index ad240f4a1..d33fdf0b4 100644
--- a/test/test_other.py
+++ b/test/test_other.py
@@ -7605,12 +7605,29 @@ Module.preRun = () => {
 #include <stdlib.h>
 #include <string.h>
 #include <dlfcn.h>
+#include "emscripten.h"
+#include <sys/stat.h>
+
+EM_JS(int, check_path, (char* cpath), {
+  const path = UTF8ToString(cpath);
+  console.log("check_path", path);
+  try {
+    const res = FS.lookupPath(path);
+    console.log("res", res);
+  } catch(e) {
+    console.log(e);
+  }
+});
 
 int main() {
   void *h;
   void (*f)();
   double (*f2)(double);
 
+  check_path("/lib/libhello1.wasm");
+  struct stat statbuf;
+  printf("res: %d\n", stat("/lib/libhello1.wasm", &statbuf));
+
   h = dlopen("libhello1.wasm", RTLD_NOW);
   assert(h);
   f = dlsym(h, "hello1");

[hoodmane]

Okay I think I fixed the problem in b805542.

[ryanking13]

Okay I think I fixed the problem in b805542.

Indeed. Thanks for the fix! Didn't know about the proxy parameter. The remaining issue now seems to be handling BigInt in a MEMORY64 environment.

[hoodmane]

The memory64 test now passes for me locally.

[ryanking13]

The memory64 test now passes for me locally.

@hoodmane Awesome, thanks for helping me fix the issues.

[hoodmane]

Slightly shorter:

Suggested change

	if (typeof FS.lookupPath !== 'function') {
	if (!FS.lookupPath) {

[hoodmane]

shorter:

Suggested change

	if (filename.charAt(0) === '/') { // abs path
	if (filename[0] === '/') { // abs path

[hoodmane]

It looks good to me, thanks @ryanking13. Let's wait for @sbc100 to review as well.

[hoodmane]

Hmm it seems the test I added breaks in CI. When I run it locally it works fine...

[hoodmane]

I hit the merge button on this, but now looking back at the history I'm thinking it might have been premature. Ah well, we could either revert and reland or fix any problems in followups.

[sbc100]

Yes, lets revert this pending some more discussion.

[sbc100]

Can you re-open this PR? (or perhaps a new one is needed?)

[ryanking13]

@hoodmane @sbc100 Made a new PR #23958

[hoodmane]

Sorry about that!