On 11-03-2020 01:08, Tom Tromey wrote:
>>>>>> "Tom" == Tom de Vries <tdevries@suse.de> writes:
> 
> Tom> When loading the executable in gdb, we create a partial symbol for foo, but
> Tom> after expansion into a full symbol table no actual symbol is created,
> Tom> resulting in a maint check-psymtab failure:
> Tom> ...
> Tom> (gdb) maint check-psymtab
> Tom> Static symbol `foo' only found in inline.c psymtab
> Tom> ...
> 
> Tom> Fix this by preventing the creation of this type of partial symbol.
> 
> With this patch, if there is a function which is only inlined (there is
> no "out-line" copy), will "break function" still work?
> 
> It seems to me that it wouldn't always, because there wouldn't be a
> partial symbol to match.  Maybe it might work accidentally if the inline
> function appears in a CU that is already expanded, like main's.  So the
> test would have to be an always-inline function that isn't used in
> main's CU.
> 
> If it does work, then I wonder how, since my mental model is that all
> the paths to expansion require a matching partial symbol.
> 
> If it doesn't work, then I think a different fix is needed.

Indeed, a different fix is needed.

This patch fixes things in the check itself instead.

OK for trunk?

Thanks,
- Tom