Re: [PATCH 00/47] single-binary: implement dynamic filtering for QOM types

[Markus Armbruster <armbru@redhat.com>]

Daniel P. Berrangé <berrange@redhat.com> writes:

> Currently the QOM type name is directly derived from the machine
> type name, just with a "-machine" suffix attached. So if the
> machine name is not unique, then neither is the QOM type name.
>
> At the very least we need the QOM type name to be unique. We
> could de-couple the two to make this work though, which does
> not seem too hard.

[...]

> We don't have many examples of clashing machine names. Is there
> any beyond just the 'virt' boards ? Now that we know we want
> globally unique names, this feels easy enough to spot during
> review.

I checked -machine help for all qemu-system-T, and it seems to be just
"none", "virt" and "virt-VERSION", not counting duplicates between
architecture variants such as aarch64 and arm, x86_64 and i386, and so
forth.

>> This rests on the assumption that the only thing we need to
>> disambiguate is machine names, i.e. there aren't multiple
>> targets with user-facing (i.e. not easily renameable) device
>> names that are shared.
>
> IMHO it is best if our goal is to ensure type names are unique
> throughout the codebase, so hopefully we don't have clashing
> device names.

Yes.

Quick glance at type_new():

    if (type_table_lookup(info->name) != NULL) {
        fprintf(stderr, "Registering `%s' which already exists\n", info->name);
        abort();
    }

If we have duplicate type names, they won't survive being linked
together in a single binary.