On Fri, 14 Dec 2018, 16:18 Julian Foad <julianf...@apache.org wrote: > Branko Čibej wrote: > > r1848943, for review. [...] > > Woo hoo! Thanks, Brane. > > Are you willing to add random-input testing for them? > > Minor comments: > * the 'relpath' one is not needed because, AFAIK, it's possible to > canonicalize any relpath and we already do; >
Possibly. I admit I was quite short on time, so I went for consistency first. * the non-canonical 'result' and scratch pool params are probably > over-kill; a subjective judgment, I know, so just a suggestion. > My thinking was that a GUI client may want to handle canonicalisation errors differently than just by displaying the error message we happen to generate. Separate result and scratch pools are just policy, I think we've learned by now that taking shortcuts with pools eventually leads to problems. -- Brane