On Tue, Feb 6, 2018 at 7:55 PM, Duy Nguyen <pclouds@xxxxxxxxx> wrote: > On Tue, Feb 6, 2018 at 7:27 PM, Duy Nguyen <pclouds@xxxxxxxxx> wrote: >> On Tue, Feb 6, 2018 at 8:48 AM, Ben Peart <peartben@xxxxxxxxx> wrote: >>> With the new behavior, making a change in dir1/, then calling status would >>> update the dir1/ untracked cache entry but not write it out. On the next >>> status, git would detect the change in dir1/ again and update the untracked >> >> Thing only missing piece here I would add is, this dir1/ detection is >> done by watchman. We have to contact watchman and ask the set of >> changed paths since $TIME where $TIME is the last time we updated >> untracked cache and invalidate those paths in core. Then it should >> work correctly. I checked the watchman query in the fsmonitor hook and >> I think it's correct so far. > > No I think I'm wrong. And worse, I think the interaction between FSNM > and UNTR extension is broken. This is partly from me guessing how > fsmonitor works so I may be double-wrong. > > UNTR extension is supposed to cover the untracked state at timestamp > $X. Where $X is stored in FSNM extension. Correct? > > When fsmonitor is used and read_directory() is executed (at timestamp > $Y), we ask watchman "what's new on worktree since $X?"). We get the > list, we invalidate relevant directories so we can see new untracked > entries (or lack of old untracked entries). We write FSNM with > timestamp $Y down. We may or may not write UNTR down because of this > patch, but let's suppose we do. All is good. UNTR now records the > state at $Y. FSNM says $Y. This is how it works (when there's no bugs) > > UNTR extension is only updated when read_directory() is called. It > does not always happen. FSNM is updated all the time (almost at every I was indeed doubly wrong. When FSMN is read, it does make calls to invalidate stuff in untracked cache data structure. If the index is written down (with updated FSMN extension and timestamp) then the UNTR extension, which is generated from in-core untracked data structure, also reflects the damages by the changed paths so next time even if those changed paths are not reported again (between $Y and $Z below), it's fine. All is good in the world again :) > git command since most of them needs to read index, many write it > down) with new timestamp. Suppose FSNM now records timestamp $Z, UNTR > still records the state at $Y because in the last index update, > read_directory() is not executed. 4 files have been added between $Y > and $Z. When we ask watchman "what's new since $Z?" we will not see > those files, so we don't invalidate relevant directories and > read_directory() will not see those files. -- Duy
