> > I think for an arch_has_* #ifdef it is normal to make them uppercase.
> > Anyway, those naming details can get upstream feedback when you post.
>
> OK. ARCH_HAS_SINGLE_STEP_INFO ?
I think the convention is ARCH_HAS_FOO where foo is the exact name sans
"arch_" of the actual function name. Ano
I think this answers my questions in the previous email I sent.
On 11/09, Roland McGrath wrote:
>
> I wouldn't make the behavior conditional on the arch hook's definition. I
> think a consistent change to a real SIGTRAP signal is better for any arch.
> Until each arch defines the hook, it can jus
I wouldn't make the behavior conditional on the arch hook's definition. I
think a consistent change to a real SIGTRAP signal is better for any arch.
Until each arch defines the hook, it can just get the default siginfo_t
contents of 0-fill.
This arch hook is purely for the single-step case. For
Preparation, no changes in compiled code.
Currently tracehook_report_syscall_exit() ignores "int step" argument.
Each architecture has to implement the stepping itself in the low-level
code, and the behaviour is not consistent across different machines.
Change tracehook_report_syscall_exit() to h