Re: [PATCH v2] Capture warnings when writing to the index cache

[Simon Marchi <simark@simark.ca>]

On 2024-03-27 09:25, Tom Tromey wrote:
>>>>>> "Simon" == Simon Marchi <simark@simark.ca> writes:
> 
> Simon> My interpretation of this is that we bind a function marked with
> Simon> `ATTRIBUTE_PRINTF (2,0)` (deferred_warnings::operator()) to a function
> Simon> view that doesn't convey the attribute.  So I guess the compiler is
> Simon> telling us that whenever the function view is invoked, there won't be a
> Simon> format string check, and that's dangerous.
> 
> Simon> I didn't find a good fix that :(.
> 
> Please try the appended.
> 
> thanks,
> Tom
> 
> commit 8be8061ea5d0ec37289070f16717d3d0aa0c93af
> Author: Tom Tromey <tromey@adacore.com>
> Date:   Wed Mar 27 07:21:56 2024 -0600
> 
>     patch
> 
> diff --git a/gdb/utils.h b/gdb/utils.h
> index d7db1d84e2f..875a2583179 100644
> --- a/gdb/utils.h
> +++ b/gdb/utils.h
> @@ -444,10 +444,18 @@ struct deferred_warnings
>       hook; see scoped_restore_warning_hook.  Note that no locking is
>       done, so users have to be careful to only install this into a
>       single thread at a time.  */
> -  void operator() (const char *format, va_list args) ATTRIBUTE_PRINTF (2, 0)
> +  void operator() (const char *format, va_list args)
>    {
>      string_file msg (m_can_style);
> +    /* Clang warns if we add ATTRIBUTE_PRINTF to this method (because
> +       the function-view wrapper doesn't also have the attribute), but
> +       then warns again if we remove it, because this vprintf call
> +       does not use a literal format string.  So, suppress the
> +       warnings here.  */
> +    DIAGNOSTIC_PUSH
> +    DIAGNOSTIC_IGNORE_FORMAT_NONLITERAL
>      msg.vprintf (format, args);
> +    DIAGNOSTIC_POP
>      m_warnings.emplace_back (std::move (msg));
>    }
>  

It builds fine with that patch.

Thanks,

Simon