Re: [PATCH] [gdb/symtab] Add set/show always-read-ctf on/off

[Tom de Vries <tdevries@suse.de>]

On 2/25/23 13:34, Tom Tromey wrote:
> Tom> Hmm, I'm starting to wonder now, perhaps a way to cover both scenarios
> Tom> (the one described above and the non-overlapping dwarf+ctf one I'm
> Tom> trying to fix) is to check where the dwarf came from:
> Tom> - if the dwarf resides in the same object file as the ctf, read both
> Tom> - if the dwarf doesn't come from the same object file (so, it comes from
> Tom>   a debug info package or debug info server), skip the ctf
> 
> Tom> I suppose that'll work, but even so I still think it makes sense to
> Tom> make it configurable.  For instance, the added test-case in
> Tom> combination with target board cc-with-gnu-debuglink will need
> Tom> convincing to read both ctf an dwarf, even if the dwarf has been split
> Tom> off.
> 
> Tom> WDYT?
> 
> I think your original approach is fine as well, and I'd hate to make it
> more complicated for you.
> 

OK, then let's go with that.

As for syntax, I went now first for the simplest approach, an on/off 
setting:
...
set always-read-ctf <on|off>
...
which itself for instance could be reformulated as read-ctf-always or 
ctf-read-always, I'm not sure if there's a preferred scheme for things 
like that.

Furthermore, I wonder, if we want to keep the possibility open of adding 
additional ways to handle this in the future, should we go instead with 
something like:
...
set read-ctf <no-dwarf2|yes>
...
or:
...
set ctf-read <no-dwarf2|yes>
...
or:
...
set ctf-read <no-dwarf2|always>
...
such that we could add the strategy proposed above as say 
no-external-dwarf2.

Or am I unnecessarily complicating things?

Thanks,
- Tom