On 5/12/20 6:40 PM, Dave Chinner wrote: > On Tue, May 12, 2020 at 12:06:46PM -0500, Eric Sandeen wrote: >> On 5/12/20 12:03 PM, Christoph Hellwig wrote: >>> For historical reasons xfsprogs tries to renamed xfs_* symbols used >>> by tools (but not those used inside libxfs) to libxfs_. Remove this >>> indirection to make it clear what function is being called, and to >>> avoid having to keep the renaming header uptodate. >>> >>> Signed-off-by: Christoph Hellwig <hch@xxxxxx> >> >> I'm fine with this. I'll wait a bit to see if Dave has a strong argument >> against it, but TBH I have never understood the point of the renaming. > > It was done so there was a clear delineation between the functions > that return negative errors (from the libxfs core) and anything else > that was named "xfs_<foo>" in xfsprogs that returned positive > errors. This was needed back when I had to bring xfsprogs up to date > with the kernel after years of neglect and the kernel had moved to > negative errnos.... > > If that's no longer a useful distinction, then it doesn't bother me > if we get rid of it. hm ok. Maybe another audit pass to make sure that every single xfs_* function (whether shared via kernel files or not) returns negative, and every local explicit libxfs_* function returns positive, would be wise? With some renaming as needed? I'd need to take another look I guess. -Eric
