fix backtraces produced by standard library under Wine

[leroycep]

from #22484, since apparently I can't reopen the old pr?

---

Alright, I've decided to give another go. I think I understood the problem better this time, otherwise the CI will inform me of my hubris.

---

main.zig:

pub const std = @import("std");

pub fn main() !void {
    std.debug.dumpCurrentStackTrace(null);
}

Compile using zig build-exe -target x86_64-windows-msvc ./main.zig and run using wine ./main.exe

Output before these changes:

Unable to dump stack trace: Unexpected

Output after these changes:

/home/geemili/code/zig/lib/std/debug.zig:198:31: 0x140001117 in dumpCurrentStackTrace (main.exe.obj)
        writeCurrentStackTrace(stderr, debug_info, io.tty.detectConfig(io.getStdErr()), start_addr) catch |err| {
                              ^
/main.zig:4:36: 0x140001015 in main (main.exe.obj)
/home/geemili/code/zig/lib/std/start.zig:475:53: 0x1400014aa in WinStartup (main.exe.obj)
    std.os.windows.ntdll.RtlExitUserProcess(callMain());
                                                    ^
???:?:?: 0x6fffffa547f8 in ??? (kernel32.dll)
???:?:?: 0x6fffffc2f93a in ??? (ntdll.dll)

It seems like it's failing to find the location of main.zig, but that's a mystery for another time.

• Wine version: wine-9.22

---

#17535 is a related issue. Compiling Zig with these changes to the standard library fixes the original error message, but replaced it with another error (as well as a now-working backtrace 😁) that comes from an unimplemented feature in wine:

~/Downloads/zig-wine-backtraces/invoking-zig-cc-17535> wine ../zig-windows-x86_64-0.14.0-dev.2643+fb43e91b2/zig.exe build-exe main.zig
error: unable to find zig self exe path: Unexpected
~/Downloads/zig-wine-backtraces/invoking-zig-cc-17535> wine ~/code/zig/build/stage4-windows/bin/zig.exe build-exe main.zig
0184:fixme:file:NtLockFile Unimplemented yet parameter
error.Unexpected NTSTATUS=0xc0000002
/home/geemili/code/zig/lib/std/os/windows.zig:2093:40: 0x14023a46f in LockFile (zig.exe.obj)
        else => return unexpectedStatus(rc),
                                       ^
/home/geemili/code/zig/lib/std/fs/Dir.zig:1099:19: 0x1401ad675 in createFileW (zig.exe.obj)
    try w.LockFile(
                  ^
/home/geemili/code/zig/lib/std/fs/Dir.zig:968:32: 0x140055eac in createFile (zig.exe.obj)
        return self.createFileW(path_w.span(), flags);
                               ^
/home/geemili/code/zig/src/Zcu/PerThread.zig:134:33: 0x140835003 in astGenFile (zig.exe.obj)
        break zir_dir.createFile(&hex_digest, .{
                                ^
/home/geemili/code/zig/src/Compilation.zig:4253:18: 0x14052157a in workerAstGenFile (zig.exe.obj)
    pt.astGenFile(file, path_digest) catch |err| switch (err) {
                 ^
/home/geemili/code/zig/lib/std/Thread/Pool.zig:178:50: 0x140521cce in runFn (zig.exe.obj)
            @call(.auto, func, .{id.?} ++ closure.arguments);
                                                 ^
/home/geemili/code/zig/lib/std/Thread/Pool.zig:291:32: 0x1404dad5e in worker (zig.exe.obj)
            run_node.data.runFn(&run_node.data, id);
                               ^
/home/geemili/code/zig/lib/std/Thread.zig:486:13: 0x140300e6c in callFn__anon_179210 (zig.exe.obj)
            @call(.auto, f, args);
            ^
/home/geemili/code/zig/lib/std/Thread.zig:599:30: 0x1401abf1e in entryFn (zig.exe.obj)
                return callFn(f, self.fn_args);
                             ^
???:?:?: 0x6fffffa547f8 in ??? (kernel32.dll)
???:?:?: 0x6fffffc2f93a in ??? (ntdll.dll)
Z:\home\geemili\code\zig\lib\std\std.zig:1:1: error: unable to load 'Z:\home\geemili\code\zig\lib\std/std.zig': Unexpected

Running the test suite under wine also had a couple tests fail, presumably from similar issues. Hard to say without running it on Microsoft Windows.

[squeek502]

The first commit is a duplicate of #17541

[leroycep]

I see.

[mlugg]

Ha, I was wondering what was causing x86_64-windows stack traces not to work under Wine! Thanks for this work. I'll cherry-pick the first commit into my branch (I already have an equivalent diff to the second).

[squeek502]

@mlugg #17541 contains a better version of the first commit IMO

[mlugg]

Ah, okay, thanks. Since the diff will merge cleanly either way, I'll leave it out my branch for now.

[squeek502]

Thanks for working on this, but I'm going to close it as it's now obsolete. The GetFinalPathNameByHandle change is included in #17541, and the SelfInfo change is included in #25227