Re: [PATCH 1/3] STABS: remove -gstabs and -gxcoff functionality

["Martin Liška" <mliska@suse.cz>]

On 9/2/22 10:54, Richard Biener wrote:
> On Fri, Sep 2, 2022 at 9:00 AM Martin Liška <mliska@suse.cz> wrote:
>>
>> On 9/1/22 13:18, Richard Biener wrote:
>>> I presume WarnRemoved will diagnose use of -gstabs but not fail
>>> compilation.  Will -gstabs then still enable -g (with the default debug
>>> format)?
>>
>> No, it won't set -g option.
> 
> That was the usual side-effect - I wonder if we want to emit extra
> diagnostic when one of the obsolete options is given but -g is not
> enabled in the end or whether we want to preserve the debug info
> enablement effect?

I would leave it for now and see if somebody would complain about
the current behavior.

> 
>>>
>>> Please followup with a gcc-13/changes.html entry.
>>
>> Sure.
>>
>>>
>>> I notice we have VMS_DEBUGGING_INFO left.  From a quick look
>>> it is used by alpha*-dec-* (exclusively) and ia64-hp-*vms*  (maybe
>>> also supports DWARF, it is ELF at least).  One of the goals of
>>> non-DWARF removal was to get rid of debug hooks and instead allow
>>> "free-form" early debug generation from the frontends.
>>
>> Can you please explain what you mean by the free-form and what's expected
>> to do with the VMS_DEBUGGING_INFO macro?
> 
> Well, VMS debugging would go, just like STABS.

Can go now, or shall we deprecate it in GCC 13 first?

> With "free-form" I mean
> that frontend code could call into the dwarf2out API directly, creating
> DWARF DIEs for language specific info (we probably want to export more
> and/or nicer APIs for such use).

Ok, so bypassing dwarf2_debug_hooks, right? I can leave it to you such API
improvement. What do you think?

Cheers,
Martin

> 
> Richard.
> 
>> Cheers,
>> Martin