Sorry, came back late for the party... On Tue, Aug 27, 2024 at 09:34:46PM GMT, Christoph Hellwig wrote: > On Tue, Aug 27, 2024 at 07:41:12AM -0700, Darrick J. Wong wrote: > > Hmm. Here you're changing function signatures for a public library. > > attrlist_cursor and xfs_attrlist_cursor are the same object, but I > > wonder if this is going to cause downstream compilation errors for > > programs that include libattr but not xfs_fs.h? > > Oh, I forgot that jdm.h is public. Yes, this will cause the compiler > to complain, so we can't do that. I spoke with darrick on Tuesday, and it seemed IMO that libhandle is supposed to be a xfsprogs thing, not something that should be packaged separated, but I might be totally wrong here. If we ought to maintain jdm.h intact, then I'd come back to pack definitions into libfrog/attr.h, but what I don't like about this, is we'll end up with two definitions for virtually the same thing, xfs_attrlist in libfrog and attrlist in xfs_fs.h, unless we do use Darrick's idea and mimic libxfs behavior in libfrog/attr.h, something like: #define xfs_attrlist_cursor attrlist_cursor So we could use it within libfrog, without messing up with public libhandle interface?! Carlos
