Time  Nick       Message
19:23 Heffalump  the fixup is a copy of the patch that was commuted with
19:18 Heffalump  the patch the rebase commuted with doesn't change
18:17 mornfall   s/up/with/
18:16 mornfall   And it is part of the rebase, part of the patch the rebase got commuted up, or a transformation of either? (Or both?)
18:15 Heffalump  the fixup then gets commuted as far down the rebase as possible, coalescing with any other fixups as it goes
18:15 Heffalump  commuting a rebase patch with a normal patch inserts a fixup into the start of the rebase
18:14 mornfall   (Diff?)
18:14 mornfall   Commute, merge, something?
18:14 mornfall   I mean, what operation creates it, in terms of patches.
18:13 Heffalump  by moving the base of the rebase patches
18:13 mornfall   How the fixup comes into being?
18:13 Heffalump  I'm not 100% certain it should, but for the fixup patches to keep them minimal
18:13 mornfall   Hm. Where is that?
18:12 Heffalump  yes
18:12 mornfall   Heffalump: Rebase uses them elsewhere than in amend?
18:01 Heffalump  can coalesce be separate from the core semantics?
17:59 Heffalump  rebase uses them
17:36 mornfall   Any other uses should probably be eliminated, since they are incorrect anyway.
17:36 mornfall   I mean, the real valid uses of coalesce are in amend, and possibly in pending, depending on how it is implemented.
17:35 mornfall   Patches without context make little sense, normally.
17:34 * Igloo    can't actually think of a case when you wouldn't OTTOMH
17:34 mornfall   Igloo: We can make it so, I suppose.
17:34 Igloo      mornfall: Do you always have the actual files when coalescing?
17:28 mornfall   (Upon amend, that is.)
17:27 mornfall   Or even record the actual user decisions in patch metadata, and simply reuse them.
17:26 mornfall   If we add "smart" differs, we can then use the existing patch and interactivity to decide whether to accept or reject their results.
17:26 mornfall   Heffalump: I would probably suggest a heuristic that would take the existing patch and use it for hints.
17:17 mornfall   A V3 Hunk, anyway.
17:16 mornfall   (Since Hunk as it is now defined has two equally feasible diffing strategies...)
17:16 mornfall   Heffalump: And can't do while coalesce is part of the core semantics.
17:16 mornfall   Heffalump: Do you think it is feasible to require a rediff for coalescing patches? I would really like to divorce diffing from core patch semantics.
13:42 mornfall   Commuting RealPatch V3.Prim doesn't work because I don't have sortCoalesceFL (yet). Smells. Bad.
13:42 mornfall   27/38 tests passed
09:42 * mornfall just got first test to pass (inverse . inverse == id, which isn't much, but better than nothing)... :-)