Forgot an extra 60 in the check interval and the notification timeout.
Very unfortunate. Check has been allowed every 168 minutes instead of
every week.
Based on the previous commits by @mgmachado but simplified.
I also changed the threshold to error after a single attempt as the
connection has likely been borked and the user should be in the know if
they couldn't reconnect after 5 seconds.
Closes#1791
I disabled code-layering and code-import-patterns as I don't think we
can make them easily pass as we reference all sorts of code from both
browser and node files. At least not worth the headache now to refactor
everything.
Previously anything that wasn't "log" such as "warn" would end up doing
`logger[logger.warn]`. Would have caught this if I hadn't used `any`...
Fixes#2364.
The new fields are from vscodium and make the welcome page
documentation links work correctly.
I also renamed the distribution to "code-server" so that when you're
in a browser, it now says code-server instead of Code OSS.
It's possible that using browser storage makes more sense with settings
sync, so we might want to revisit this once/if we get settings sync
working. As it currently is though, browser storage just causes jank.
The path was also missing a `User` at the end so I added that. This
might affect the Vim extension which would have been writing to the
wrong path previously but I don't believe it should affect anything
else since they would have been writing to browser storage.
- Fixes#2208
- Fixes#2231
- Fixes#2279
- Fixes#2274
The telemetry service depends on this now. I had to move it into
invokeFunction and use accessor.get otherwise getLogger on the service
was undefined.
I also had to move some the extension management service because it
depends on the moved telemetry service. I moved a few other services as
well to better match VS Code (sharedProcessMain.ts).
I swapped some this.services.get with accessor.get since that seems to
be the correct method although for these other services either method
seems to work.
This way the connection can be initiated by either side. It looks like
sometimes the initial message from the client is lost (it never makes it
into the onControlMessage callback) but I'm still not sure why or if
that is preventable.
Also added a timeout on the server end to clean things up in case the
client never responds.
If do not update the UID within the passwd database to match whatever
uid the container is being ran as, then sudo will not work when renaming
the user to match $DOCKER_USER as it will complain about the current
user being non-existent.
Removing them just for peace of mind even though they seem to get
filtered out later. This line is meant to only add remote extensions
that aren't capable of running in the browser. If they are
browser-capable they don't need to run in our shimmed Node environment.
This makes the fetch work independently of the worker's origin which is
no longer the same as the main thread (the main problem is the inability
to send cookies without setting SameSite to None).
- The .js build files are no longer committed so they're gone.
- ParsedArgs and EnvironmentService are now NativeParsedArgs and
NativeEnvironmentService.
- Interface for environment service was moved.
- getPathFromAmdModule was deprecated.
We only want to use an old version for glibc which the centos:7
image takes care of.
The old version of git used in debian:8 was causing problems
with the uid/gid passthrough with no user in passwd.
We do not try renaming $HOME anymore as there is no good way
to do it.
We also only try to convert if the user hasn't been changed.
Finally I added usage to the docker docs in install.md
Closes#2056
VS Code has a short delay before writing storage (probably to queue up
rapid changes). In the web version of VS Code this happens on the client
which means if the page is reloaded before the delay expires the write
never happens.
Storage updates are already promises so this simply returns the promise
returned by the delayer so it won't resolve until the write actually
happens.
Fixes#2021.