Repository navigation
Conversation
ebraminio
force-pushed
the
crt-free
branch
10 times, most recently
from
November 20, 2026 01:30
4e05f3d to
3b6c69d
Compare
The Windows backend currently reaches for CRT allocation, wide-string, and formatting helpers. That is fine for normal builds, but it gets in the way when HIDAPI is embedded in a DLL or plugin that is linked with /NODEFAULTLIB, or when the host wants to avoid pulling in a particular UCRT/VCRuntime. Add HIDAPI_NO_CRT to switch those operations to Win32 equivalents without changing the public API. The default build still uses the CRT, and its swprintf/wcstol/towupper paths are left alone; the replacements are compiled only for HIDAPI_NO_CRT. I kept the new path opt-in because it changes some edge behavior: allocations are zeroed, wide-string comparison is ordinal, and error formatting is bounded. I would rather not silently change every Windows consumer before this has had some use. Descriptor reconstruction is covered by the same helpers. The no-CRT compile also disables MSVC stack probes/security cookies and GCC's stack probe, since that code has a large stack frame. A matrixed Windows CI probe links a custom-entry executable against the static no-CRT library with both MSVC and clang-cl, and checks that the result does not import a CRT DLL. Assisted-by: deepseek-harness:deepseek-flash
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
HIDAPI's Windows backend uses the CRT for allocation, wide-string operations, and error formatting. That is fine for normal desktop builds, but it can matter when HIDAPI is linked into code that does not control the process runtime: ASI/Mod loader plugins, audio/video plugins, game overlays, and similar injected modules, or a native messaging host. Those hosts may build with
/NODEFAULTLIB, supply their own entry point, or pin a runtime, and a library that pulls in CRT imports can be awkward or impossible to fit. This is not a claim that the UCRT is usually missing or that/MTis not a valid alternative; the narrow case is a host that has already chosen a no-CRT or fixed-runtime model. Binary size can improve as a side effect, but it is not the main argument.This is a suggestion for an opt-in
HIDAPI_NO_CRTmode that uses Win32 allocation, string, and memory helpers instead. The normal CRT build remains the default and itsswprintf/wcstol/towupperpaths are untouched; the replacements are compiled only when the option is enabled. Report descriptor reconstruction is covered by the same helpers.A shared HIDAPI DLL still needs CRT startup unless its consumer supplies a custom entry point, so this only removes CRT function imports; the primary target is static embedding.
If this is accepted, a reasonable follow-up would be to make the internal helpers inline and use them in the default Windows path as well, so the two implementations can eventually collapse into one. The
mem*shims are the part that needs care: they exist for compiler-generated calls, not for HIDAPI's own code, and removing them requires either compiler/linker changes or keeping a small shim. I would not expand this PR to change the default path yet; the safer sequence is to let the opt-in path get some use and CI coverage first.If this does not fit HIDAPI's direction or the maintenance cost is not worth it, closing or ignoring it is completely fine.