Julian Waters <tanksherma...@gmail.com> writes:

> Apologies, here is the implementation with regenerated configure this time. 
> Do excuse me sending an entirely new mail instead of replying to the previous 
> one, I have to do it this way due to a quirk in my email client.
>
> best regards,
> Julian
>
> gcc/ChangeLog:
>
>         * config/i386/i386.cc (ix86_legitimate_constant_p): Handle new UNSPEC.
>         (legitimate_pic_operand_p): Handle new UNSPEC.
>         (legitimate_pic_address_disp_p): Handle new UNSPEC.
>         (ix86_legitimate_address_p): Handle new UNSPEC.
>         (ix86_tls_index_symbol): New symbol for _tls_index.
>         (ix86_tls_index): Handle creation of _tls_index symbol.
>         (legitimize_tls_address): Create thread local access sequence.
>         (output_pic_addr_const): Handle new UNSPEC.
>         (i386_output_dwarf_dtprel): Handle new UNSPEC.
>         (i386_asm_output_addr_const_extra): Handle new UNSPEC.
>         * config/i386/i386.h (TARGET_WIN32_TLS): Define.
>         * config/i386/i386.md: New UNSPEC.
>         * config/i386/predicates.md: Handle new UNSPEC.
>         * config/mingw/mingw32.h (TARGET_WIN32_TLS): Define.
>         (TARGET_ASM_SELECT_SECTION): Define.
>         (DEFAULT_TLS_SEG_REG): Define.
>         * config/mingw/winnt.cc (mingw_pe_select_section): Select proper TLS 
> section.
>         (mingw_pe_unique_section): Handle TLS section.
>         * config/mingw/winnt.h (mingw_pe_select_section): Declare.
>         * configure: Regenerate.
>         * configure.ac: New check for broken linker thread local support
>
> From fceb5113f33a950048d57a1ecde39084eaa09ffe Mon Sep 17 00:00:00 2001
> From: Julian Waters <tanksherma...@gmail.com>
> Date: Tue, 15 Oct 2024 20:56:22 +0800
> Subject: [PATCH] Implement Windows TLS
>

The commit message here should contain the ChangeLog fragment from above
(really, please use git-send-email wherever possible) and a reference to
PR80881.

It's also up to the target maintainers and release managers as to
whether this goes into GCC 15 given it's definitely not a regression fix
(we're in stage 4 now).

I'd also consider adding a Co-authored-by for some of the folks who
helped out like Eric, but that's up to you.

Reply via email to