Follow up to #6508. Adds a request "counter" that increments for every API request, matching how `front` does it.
Also tweaks the `output.time()` output a bit by rendering the timestamp in gray, and with `[]` brackets. If the time diff is < 1000ms then the full milliseconds will be printed intstead of the seconds.
<img width="710" alt="Screen Shot 2021-07-22 at 2 45 39 PM" src="https://user-images.githubusercontent.com/71256/126713843-db70ed9c-4752-4ca9-8f54-313c4cb44914.png">
Updates the `output.time()` function to accept a function as the "label" parameter. When a function is passed, it will be invoked twice, once at the beginning of the promise being executed, and a second time after the promise has resolved. The second time, the return value of the promise is passed to the label function, so that the resolved value may be used to create the label.
Drops usage of Node's `console.time()` and `console.timeEnd()` since we were using it in a hacky and unnecessary way.
Adds an "out-of-band" login mode to the `vc login` command, for use in headless environments like inside a Docker container or an SSH session. In these situations, spawning a localhost HTTP server won't work, because the HTTP server will not be accessible to the web browser that is doing the authentication.
In "out-of-band" mode, the login URL is simply presented to the user, and they must visit the login URL in their web browser and complete the authentication. The callback URL is a page on front that shows the verification token to the user and instructs them to copy + paste the token back into the CLI, at which point the login can complete.
Docker/SSH sessions are attempted to be automatically detected, and "out-of-band" mode is enabled implicitly. But there is also an `--oob` flag in case that detection logic were to fail (or if you'd like to run OOB mode locally).
This PR also adds shortcut flags for the Git provider login mechanisms, i.e. `vc login --github`.
https://user-images.githubusercontent.com/71256/125129304-1f5f3380-e0b4-11eb-8139-d1109811bd86.mov
This is a continuation of https://github.com/vercel/vercel/pull/6382 (thank you @markozxuu!) with merge conflicts addressed and pushed to a local branch, so that CI tests run.
Fixes an edge case bug when the user has a current scope that is
different than the owner of the project that is being deployed.
When this was the case, the API call to get the certs for a domain at
the end of the `vc deploy` command was using the incorrect `teamId`,
potentially causing a 403 error (when the current auth token does not
have access to the selected scope).
`vc deploy` doesn't care about the current "scope" that the CLI has selected, since it uses the `.vercel` directory to determine the ownerId and projectId.
Therefore, it should not be fetching the team details of the selected scope during `vc deploy`, since it's possible that the token might not have access to the current scope and returns a 403, even though that API call is unnecessary for the `vc deploy` command to complete.
Closing keep-alive HTTP connections was causing the `server.close()` call
to take a few seconds before completing, so set the `Connection: close`
response header in order to make the connections close immediately,
so that `server.close()` is fast.
Considering that it's a development environment, it's not important to wait for ongoing HTTP request connections to complete. Sometimes it takes a long time for all the shutdown operations to complete, which makes the command feel sluggish. So let's just `process.exit()` and exit quickly.
In some cases (i.e. when SSH'd to a remote machine) the `open` command will not work reliably. So we need to print the URL to the user as a fallback for those cases when the web browser is not automatically opened.
This also moves where `tokenName` is specified to be in the "verify" endpoint, so that it does not need to be part of the URL that gets printed to the user.
<img width="738" alt="Screen Shot 2021-06-07 at 2 12 47 PM" src="https://user-images.githubusercontent.com/71256/121089239-b5452d00-c79b-11eb-85b2-0e45b817dff0.png">
* refactor: Remove unncessary file
* feat(cli/alias): Migration to TS
* refactor(cli/alias): Add a line break
* refactor(cli/alias): Remove unnecesary code
* feat(cli): Add `Paginationoptions` type to `pagination`
* feat(cli/commands): Rewrite th `alias ls` command to TS
* refactor: Remove unncessary code
* feat: Create helper function for `getSafeAlias`
* refactor: Remove unnecessary code
* feat: Remove parameter generic, "null" for the fetch
* feat(cli/alias): Rewritten in full TS
* feat: Add Partial to opts
* refactor: Remove comment @ts-ignore
* feat: Add Partial to opts
* feat: Only should be return `alias.uid`
* refactor: Remove `Alias` type from of the parameter `id`
* refactor: Remove destructuring from alias object
* refactor: Remove unnecessary code
* feat: Rename `created` property to `createdAt` of number type
* refactor: Move getSafeAlias function in the same file
* refactor: Simplifying code
* refactor: Intentation did not affect diff on git
* Add null back to type
Co-authored-by: Steven <steven@ceriously.com>
The call to `GET /projects/info` is used to check existence but it can cause a race condition if the project was removed before the `DELETE /v2/projects` is called.
Instead, we rely on the response from `DELETE /v2/projects` to determine if the project exists or not.
This will also allow us to remove a legacy API endpoint in the future (see related API PR)
