Re: [PATCH] gdb/dwarf2: Add symbols for function declarations

[Kevin Buettner <kevinb@redhat.com>]

On Tue, 24 Jun 2025 09:12:11 -0600
Tom Tromey <tom@tromey.com> wrote:

> >>>>> "Kevin" == Kevin Buettner <kevinb@redhat.com> writes:  
> 
> Kevin> This commit was motivated by comments 3 and 4 for bug 31563:
> Kevin> https://sourceware.org/bugzilla/show_bug.cgi?id=31563#c3  
> 
> I didn't really read through this yet, but in general a patch to add a
> new (non-local) full symbol that doesn't have a corresponding change to
> the indexer seems like it will be incorrect.  The issue is that lookups
> will only work if the CU has already been expanded; and with my
> long-suffering search-via-the-index series, the lookups will fail
> entirely.

Do you think that a function declaration (where, in GDB's symbol table,
the address class is LOC_UNRESOLVED) should be added to the index?

These symbols don't have addresses associated with them, though the
address can be obtained from the minsyms.

One of the things that I had to do while looking at and fixing the
various regressions that I encountered was to make various parts of
gdb NOT use the function declaration symbols, or, at least prefer
other symbols over the ones for declarations. E.g. they had to be
removed in order for "info func" to continue working as before.

If you think it's needed, I'll take a look, but I'd guess that if
I add them for indexing purposes, other bits of code might have to
be adjusted to later remove them.

Kevin