Hello Samuel, Thank you for the detailed review. You are right that the series does not present a sufficiently concrete plan and that it introduces abstractions without first demonstrating that the existing GNU Mach facilities are insufficient.
The current series should not be considered ready for review. In particular, the freestanding headers, time abstraction, thread layer, and string implementation duplicate existing facilities or compiler-provided headers. I will not send a revised version of this series in its current form. Regarding authorship, I need to clarify that the patches were prepared with substantial assistance from an AI system. I therefore should not present the series as if every line had been independently written by me without that assistance. I will review the contribution and determine whether I can rewrite and maintain a smaller patch whose implementation I fully understand and can take responsibility for. I also understand that I must not add an FSF copyright notice without the appropriate copyright assignment. I will remove or correct such notices in any future work and will follow the FSF procedure before submitting files whose copyright is intended to be assigned to the FSF. My revised plan is: 1. discard the proposed `modern/` framework; 2. identify one concrete problem in the existing GNU Mach code; 3. reuse GCC-provided freestanding headers and existing Mach facilities; 4. prepare a short RFC describing the problem, scope, alternatives, and validation plan; 5. submit only a small, independently reviewable patch after the design is agreed upon. If appropriate, please consider the current series withdrawn from review while I rework the proposal. Regards, Breno Rodrigues Alves Em dom., 8 de nov. de 2026 às 20:24, Samuel Thibault < [email protected]> escreveu: > Hello, > > You need to explain what your plans are. As it is it just looks like a > pile of AI-generated code that makes no actual sense. > > > Breno Rodrigues Alves, le dim. 08 nov. 2026 18:47:07 -0300, a ecrit: > > Introduce the freestanding C17 vocabulary, exact-width kernel types, > alignment helpers, compiler built-ins, memory primitives, and the initial > task/thread platform layer. Preserve hardware-specific assembly outside > this portable layer. > > > > Signed-off-by: Breno Rodrigues Alves <[email protected]> > > Do you actually own copyright on these changes? As in: you have written > them, and not AI-generated them. > > Also, to make the files > > > + Copyright (C) 2026 Free Software Foundation, Inc. > > You need to assign copyright to the FSF, with the assignment form at the > end of this mail. > > > modern/Makefile | 57 +++ > > Don't use the "modern" word, it is meaningless. You will find people who > find C17 already ancient and not modern like C23. > > > modern/freestanding/limits.h | 29 ++ > > modern/freestanding/stdalign.h | 14 + > > modern/freestanding/stdarg.h | 17 + > > modern/freestanding/stdatomic.h | 138 +++++++ > > modern/freestanding/stdbool.h | 14 + > > modern/freestanding/stddef.h | 17 + > > modern/freestanding/stdint.h | 99 +++++ > > modern/freestanding/stdnoreturn.h | 11 + > > We don't want to reimplement gcc-provided files, and rather just reuse > them. > > > modern/freestanding/time.h | 23 ++ > > There is no need for this. Mach already has its own notion of time. > > > modern/include/mach/c17/base.h | 333 +++++++++++++++++ > > That's duplicating a lot of existing stuff for no reason. > > > modern/include/threads.h | 104 ++++++ > > modern/src/core/threads.c | 594 ++++++++++++++++++++++++++++++ > > That's reinventing threads for no reason. > > > modern/src/core/mem.c | 140 +++++++ > > We already have the strings functions. > > With regards, > Samuel >
