Feat: Rewritten Windows Bindings by Mahasvan · Pull Request #5 · Mahasvan/HWProbe

Mahasvan · 2026-08-01T12:57:31Z

No description provided.

…win-bindings-new

optional

kernel-dev · 2026-08-01T17:27:37Z

+typedef struct {
+    char values[WMI_MAX_FIELDS][WMI_FIELD_LEN];
+} WmiRow;


This could be turned into a typedef without being encapsulated in a struct. Is there a specific reason for it to stay inside of a struct?

Also, I think you meant to do char values[WMI_FIELD_LEN][WMI_MAX_FIELDS]

kernel-dev · 2026-08-01T17:41:20Z

+static void VariantToUtf8Slot(VARIANT &vt, char *dst, int dst_size) {
+    if (dst_size <= 0) return;


In general, to avoid having to check for negative numbers (especially if talking about a parameter indicating size), you can use unsigned integer types: unsigned int dst_size

The caller is also then forced to ensure the passed value is 100% a positive integer.

The check dst_size <= 0 already handles both cases in one comparison. Switching to unsigned wouldn't force callers to pass positive values in C (negative ints silently wrap to huge unsigned values, which is worse), and it would require changing the header, the ctypes binding, and all callers for no functional benefit. The current int is consistent with the rest of the ABI (WMI_FIELD_LEN, WMI_MAX_FIELDS are all int-typed).

Ah yeah, I forgot about that. We could technically use -Werror with -Wsign-conversion and -Wconversion to at least mark when these cases of wrapping are happening, but if you think it's not worth the change, then OK, can mark this as resolved.

kernel-dev · 2026-08-01T17:58:02Z

+    // VT_BSTR fast path.
+    if (bstr) {
+        int len = WideCharToMultiByte(CP_UTF8, 0, bstr, -1, nullptr, 0, nullptr, nullptr);
+        if (len > 0) {
+            if (len > dst_size) len = dst_size;
+            WideCharToMultiByte(CP_UTF8, 0, bstr, -1, dst, len, nullptr, nullptr);
+            dst[dst_size - 1] = '\0';
+        }
+    }
+}


Wouldn't it just be easier to move this into the initial if condition block?

apparently not, according to claude

Why not? It seems to be doing the exact same thing when analyzing control flow?

detection

display name

…WProbe into feat/win-bindings-new

This reverts commit 93f45a1.

kernel-dev

In general I'd like to assert a few points:

General utility/common header functionality would be a nice addition, some methods are re-used or have the potential for re-usability. Defining macros and such in private scopes is OK as long as they stay only in that scope. Otherwise, they should be moved to a common file as well.
Graceful handling of Win32/COM API calls should be implemented, i.e. the suggested changes for WideCharToMultiByte(...) second call, which is meant to allocate data to the buffer, does have the potential to fail; we should react appropriately to this.
We should strive to type data as explicitly as possible, i.e. don't use int solely in place where unsigned int fits better. For readability or ease sake, we can define typedefs like typedef unsigned int uint;
Even for our internal bindings etc. - we should still devise a system for statuses. It'll be helpful for both the backend and frontend to understand why exactly some call failed if it failed. And that way, we can also design our ABI to be as concise and feature-complete as possible.

kernel-dev · 2026-08-01T19:31:49Z

+
+// ---- wide -> UTF-8 ----
+
+static std::string WideToUtf8(const wchar_t *src) {


Could move this to a common utilities file, seems like it's used throughout the bindings frequently (along with any other functions you might have in mind)

fair enough tbh, ill refactor in a bit

kernel-dev

For some reason I couldn't add to my previous review, so I published that one and now it let me 😄

Mahasvan

Thanks for the review @kernel-dev, much appreciated.

Mahasvan · 2026-08-02T05:25:26Z

+static void VariantToUtf8Slot(VARIANT &vt, char *dst, int dst_size) {
+    if (dst_size <= 0) return;


The check dst_size <= 0 already handles both cases in one comparison. Switching to unsigned wouldn't force callers to pass positive values in C (negative ints silently wrap to huge unsigned values, which is worse), and it would require changing the header, the ctypes binding, and all callers for no functional benefit. The current int is consistent with the rest of the ABI (WMI_FIELD_LEN, WMI_MAX_FIELDS are all int-typed).

Mahasvan · 2026-08-02T05:30:30Z

+    // VT_BSTR fast path.
+    if (bstr) {
+        int len = WideCharToMultiByte(CP_UTF8, 0, bstr, -1, nullptr, 0, nullptr, nullptr);
+        if (len > 0) {
+            if (len > dst_size) len = dst_size;
+            WideCharToMultiByte(CP_UTF8, 0, bstr, -1, dst, len, nullptr, nullptr);
+            dst[dst_size - 1] = '\0';
+        }
+    }
+}


apparently not, according to claude

Check return value of all WideCharToMultiByte write calls. On failure, return empty string or null-terminate the buffer instead of leaving uninitialized contents.

tests

Mahasvan and others added 17 commits August 1, 2026 18:24

WIP: WMI wrapper DLL

c5c62a0

WIP: rename win_new to win, and win to win_old

cccaa78

Update wmi binding

7243b8a

Update wmi.cpp

9780709

build: update Windows native binding [skip ci]

d081f14

docs: add reference in docstring for win cpu

39354f5

Use WMI for Windows CPU data

e128bec

wip: add gpu info bindings

e2c97d6

wip: working GPU detection on Windows

ec3e2d9

Update gpu_info.cpp

cdeabf0

Merge branch 'main' of https://github.com/Mahasvan/HWProbe into feat/…

7a0e15e

…win-bindings-new

Delete device_info.dll

3d44843

build: update Windows native bindings [skip ci]

16d175f

feat: rewrite windows memory discovery

43f33eb

fix: refactor windows memory discovery, make MemoryModuleSlot fields

9e7bcf6

optional

Use new WMI for Windows Storage

1001eb7

feat: rewrite windows network retrieval with new wmi wrapper

5727a44

kernel-dev reviewed Aug 1, 2026

View reviewed changes

Comment thread src/hwprobe/interops/win/src/gpu_info.cpp Outdated

Comment thread src/hwprobe/interops/win/src/gpu_info.cpp Outdated

Comment thread src/hwprobe/interops/win/src/gpu_info.cpp Outdated

Mahasvan and others added 11 commits August 2, 2026 00:25

feat: Add display info bindings for windows, revamp windows display

cfe5444

detection

build: update Windows native bindings [skip ci]

93f45a1

fix: build correct acpi path in win display; potentially fix missing

c856be4

display name

Merge branch 'feat/win-bindings-new' of https://github.com/Mahasvan/H…

6f9cc0f

…WProbe into feat/win-bindings-new

wip: fix monitor name retrieval on windows

33f97cc

build: update Windows native bindings [skip ci]

c54e961

improve display detection by combining wmi, dxgi, and setupapi

2a76c5d

Merge branch 'feat/win-bindings-new' of https://github.com/Mahasvan/H…

9e3123d

…WProbe into feat/win-bindings-new

Update display_info.cpp

646a843

build: update Windows native bindings [skip ci]

ecd17bb

Revert "build: update Windows native bindings [skip ci]"

79c5006

This reverts commit 93f45a1.

kernel-dev reviewed Aug 1, 2026

View reviewed changes

Comment thread src/hwprobe/core/windows/cpu.py Outdated

Comment thread src/hwprobe/interops/win/include/gpu_info.h

kernel-dev added the enhancement New feature or request label Aug 1, 2026

Mahasvan added 2 commits August 2, 2026 10:58

fix: harden VariantToUtf8Slot against null dst and redundant termination

2de52c8

Fix WideCharToMultiByte logic

01d458a

Mahasvan commented Aug 2, 2026

View reviewed changes

Mahasvan added 6 commits August 2, 2026 11:06

handle edge case in MultiByteToWideChar

cd51ab6

fix: handle WideCharToMultiByte failures in gpu_info.cpp

80751b1

Check return value of all WideCharToMultiByte write calls. On failure, return empty string or null-terminate the buffer instead of leaving uninitialized contents.

more descriptive default manufacturer namein windows gpu

b5005b6

misc: remove magic numbers from windows cpu instruction check; update

518f179

tests

misc: align dashes in code comments

1756b6c

fix: Improve virtual network adapter detection; update tests

2caee82

kernel-dev assigned kernel-dev and Mahasvan Aug 2, 2026

		static void VariantToUtf8Slot(VARIANT &vt, char *dst, int dst_size) {
		if (dst_size <= 0) return;


		// ---- wide -> UTF-8 ----

		static std::string WideToUtf8(const wchar_t *src) {

Conversation

Mahasvan commented Aug 1, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

kernel-dev left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

kernel-dev left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Mahasvan left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

kernel-dev left a comment •

edited

Loading