On Fri, Mar 28, 2025 at 10:24:25AM -0700, Cary Coutant wrote:
> >
> > Yeah. Strings are more expensive than just numbers (which can be done
> > through DW_FORM_implicit_const) and for strings you'd need some agreed way
> > how to compare what is newer and what is older.
> > strverscmp, rpmvercmp, ... (many choices, what is the right segmented
> > string
> > comparison)?
> >
>
> For an attribute that is used maybe once per compilation unit, I don't
> think the space cost is all that significant. But specifying how to compare
Partial units can be quite small and there it could matter.
> Once a version scheme is assigned, it's not that simple to change it. We'd
> have to impose a requirement that in moving from one scheme to another, the
> producer guarantees that newer versions always compare greater than earlier
> versions (e.g., moving from VVMMPP to a hypothetical VVVMMMPPP would work
> as long as VVV > 0).
I didn't mean to change versioning scheme when one has been assigned
already, just to consider this when assigning the versioning scheme if
something has such wild minor/patchlevel numbers and doesn't want to use
dates. Note, it isn't a big deal if the major goes over 100 in the VVMMPP
schemes (as long as it still fits into numbers with reasonable precision).
Jakub
--
Dwarf-discuss mailing list
Dwarf-discuss@lists.dwarfstd.org
https://lists.dwarfstd.org/mailman/listinfo/dwarf-discuss
