It turns out that ZFS doesn't do any normalization when storing files,
but does do normalization "as part of any comparison process".
In practice, this seems to mean that if you LStat a normalized filename,
ZFS will return the FileInfo for the un-normalized version of that
filename.
This meant that our test to see whether a separate file with a
normalized version of the filename already exists was failing, as we
were detecting the same file.
The fix is to use os.SameFile, to see whether we're getting the same
FileInfo from the normalized and un-normalized versions of the same
filename.
One complication is that ZFS also seems to apply its magic to os.Rename,
meaning that we can't use it to rename an un-normalized file to its
normalized filename. Instead we have to move via a temporary object. If
the move to the temporary object fails, that's OK, we can skip it and
move on. If the move from the temporary object fails however, I'm not
sure of the best approach: the current one is to leave the temporary
file name as-is, and get Syncthing to syncronize it, so at least we
don't lose the file. I'm not sure if there are any implications of this
however.
As part of reworking normalizePath, I spotted that it appeared to be
returning the wrong thing: the doc and the surrounding code expecting it
to return the normalized filename, but it was returning the
un-normalized one. I fixed this, but it seems suspicious that, if the
previous behaviour was incorrect, noone ever ran afoul of it. Maybe all
filesystems will do some searching and give you a normalized filename if
you request an unnormalized one.
As part of this, I found that TestNormalization was broken: it was
passing, when in fact one of the files it should have verified was
present was missing. Maybe this was related to the above issue with
normalizePath's return value, I'm not sure. Fixed en route.
Kindly tested by @khinsen on the forum, and it appears to work.
GitHub-Pull-Request: https://github.com/syncthing/syncthing/pull/4646
The Ping event is important, as it means that requests complete within
a sensible time. The disk events API didn't have the Ping event, so
if there were no disk events, the request would keep taking forever.
Unless, of course, there's a reverse proxy which times the request out
after a suitably large interval (or something else aborts it), in which
case Syncthing isn't very happy.
GitHub-Pull-Request: https://github.com/syncthing/syncthing/pull/3929
Events API consumers rely on being able to detect that events were skipped
by the fact that the event ID has increased by more than 1. This is
documented, and is absolutely necessary when trying to maintain a local
model of Syncthing's state.
With the introduction of LocalChangeDetected, which is not exposed to the
Events API, this contract was broken.
This commit introduces separate concepts of a "Global ID" and a
"Subscription ID". The Global ID of an event is unique across all
subscriptions. The Subscription ID is local to a particular subscription,
and always increments by 1. They are both exposed over the Events API, but
the Subscription ID uses the key "id" for backwards compatibility, and
the "?since=xx" parameter refers to the Subscription ID (making the Global
ID for information only).
GitHub-Pull-Request: https://github.com/syncthing/syncthing/pull/3351
LGTM: calmh
With the previous setup, browsers were free to use a local cache for any
length of time they pleased: we didn't set an 'Expires' header (or max-age
directive), and Cache-Control just said "you're free to cache this".
Therefore be more explicit: we don't mind if browsers cache things, but they
MUST revalidate everything on every request.
GitHub-Pull-Request: https://github.com/syncthing/syncthing/pull/3221
The log bar is hidden by CSS, but will appear briefly while the page is
loading (after the html is fetched, but before dev.css is fetched).
Hide it by using an inline style instead, so this does not happen.
GitHub-Pull-Request: https://github.com/syncthing/syncthing/pull/3220
This is the same issue as #2014/#2062. Bootstrap doesn't like having two dialogs
open at once: it marks the body has having no dialogs open when the first dialog
is closed, regardless of whether the second dialog is still open.
This means that scrolling doesn't happen properly, and the user cannot
scroll to the dialog's 'close' button.
Work around this by making sure the first dialog (the settings page) is fully closed
before the second dialog (usage preview) is opened.
- Move to ipinfo.io for geoip, rather than Telize. Telize has been closed
down. ipinfo.io has apparently got decent availability, and allows
1,000 requests per day on the free tier. Since requests are made by the
client, this should be more than enough (and the total across all clients
should still be less than this).
- Fix issue where one nonresponsive relay would cause 'data unavailable'
to be shown for many relays. This was caused by the relay status
promise not being correctly added to the list of things being waited
for before the map was rendered. Any delayed relay status requests
would therefore occur after the map was rendered, which was too late.
Knowing why a relay server failed to join the pool can be important. This
is typically an issue which must be investigated after it occurred, so
having logs available is useful.
Running with -debug permanently enabled is impractical, due to the amount
of traffic that is generated, particularly when data is being transferred.
Logging is limited to at most one message per minute, although one message
per hour is more likely.
This allows relaysrv to listen on an unprivileged port, with port
forwarding directing traffic from 443, thus providing an alternative
to using setcap cap_net_bind_service=+ep
class modal-open is applied to <body>, which ultimately means that the
browser will scoll to the modal's content. However, #editFolder was
finishing its close animation (and removing this modal-open class)
after #editIgnores had set modal-open (and had started its open
animation). The end result is that <body> ends up without modal-open
when #editIgnores is open, and so the browser doesn't properly scroll.
Instead, only open the #editIgnores once #editFolder has finished closing.
Upgrading to bootstrap 3.3.5 meant that checkboxes inside a div with
column-count:3 set would be unclickable in Chrome: in fact, the entire
div appears to sit on top of its contents, making interaction impossible.
This affected both the 'show folder with these devices' and 'these devices
can access this folder' sections of the UI.
I'm not sure what the the underlying cause is, but moving to Bootstrap's
grid system appears work around the issue. Devices/folders have to be
explicitly split out into rows, otherwise the final element appears offset.
To do this grouping by row, a new filter (groupFilter) has been added, which
turns an input of e.g. [1, 2, 3, 4, 5] with a groupSize of 3 into
[[1, 2, 3], [4, 5]]. However altering the collection in this way throws
Angular into an infinite watch loop, terminating in infdig. m59peacemaker's
pmkr.filterStabilize (MIT) was added to work around this issue.
This also has the nice side-effect of wrapping the list of devices/folders
when the screen width decreases.
See also:
- #2027 (bootstrap update which triggered this issue)
- #1121 (last time it happened)
