Hi Radek,
On Jun 24 19:21, Radek Barton via Cygwin-patches wrote:
> Hello
>
> Finally we've managed to rule out that the regressions were actually
> introduced by
> https://sourceware.org/pipermail/cygwin-patches/2025q2/013832.html, Thiru
> will send the fix soon.
>
> Radek
>
> ---
> >From c33f2e1b0037f9e5a3dbae4a0c82070db851cb33 Mon Sep 17 00:00:00 2001
> From: =?UTF-8?q?Radek=20Barto=C5=88?= <[email protected]>
> Date: Thu, 5 Jun 2025 13:15:22 +0200
> Subject: [PATCH v3] Cygwin: stack base initialization for AArch64
> MIME-Version: 1.0
> Content-Type: text/plain; charset=UTF-8
> Content-Transfer-Encoding: 8bit
>
> Signed-off-by: Radek BartoĊ <[email protected]>
> ---
> winsup/cygwin/dcrt0.cc | 8 +++++++-
> 1 file changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/winsup/cygwin/dcrt0.cc b/winsup/cygwin/dcrt0.cc
> index f4c09befd..3dceae654 100644
> --- a/winsup/cygwin/dcrt0.cc
> +++ b/winsup/cygwin/dcrt0.cc
> @@ -1030,14 +1030,20 @@ _dll_crt0 ()
> PVOID stackaddr = create_new_main_thread_stack (allocationbase);
> if (stackaddr)
> {
> -#ifdef __x86_64__
> /* Set stack pointer to new address. Set frame pointer to
> stack pointer and subtract 32 bytes for shadow space. */
> +#if defined(__x86_64__)
The comment in terms of the 32 byte shadow space is now in the wrong spot.
Splitting out this part would probably help the reader.
> __asm__ ("\n\
> movq %[ADDR], %%rsp \n\
> movq %%rsp, %%rbp \n\
> subq $32,%%rsp \n"
> : : [ADDR] "r" (stackaddr));
> +#elif defined(__aarch64__)
> + __asm__ ("\n\
> + mov fp, %[ADDR] \n\
> + mov sp, fp \n"
> + : : [ADDR] "r" (stackaddr)
> + : "memory");
> #else
> #error unimplemented for this target
> #endif
> --
> 2.49.0.vfs.0.4
Thanks,
Corinna
