fix(http): Long https requests would be cancelled

If we had a long http request it would be cancelled potentially even before the given timeout.
This is because there was ANOTHER timeout that is specifically in getResponseFromGetRequest().

SOLUTION:

This other timeout was intended to be used for retries, and to set an upper limit in time
for retrying. It was very low (3 seconds) so we bump it to a larger time instead to be safe.

In general, the user provided Timeout object should be used instead to set the expected upper limit.
The one this commit changes is the "backup" upper limit.

Signed-off-by: nkomonen-amazon <[email protected]>

Author: nkomonen-amazon

packages/core/src/shared/resourcefetcher/httpResourceFetcher.ts

@@ -9,6 +9,7 @@ import { ResourceFetcher } from './resourcefetcher'
 import { Timeout, CancelEvent, waitUntil } from '../utilities/timeoutUtils'
 import request, { RequestError } from '../request'
 import { isUserCancelledError } from '../errors'
+import { oneMinute } from '../datetime'
 
 type RequestHeaders = { eTag?: string; gZip?: boolean }
 
@@ -117,8 +118,10 @@ export class HttpResourceFetcher implements ResourceFetcher<Response> {
                 return req.response.finally(() => cancelListener?.dispose())
             },
             {
-                timeout: 3000,
-                interval: 100,
+                // Ensures we don't abort a long request (eg: slow connection).
+                // This timeout is just backup, the given Timeout object should be relied on instead.
+                timeout: oneMinute * 30,
+                interval: 500,
                 backoff: 2,
                 retryOnFail: (error: Error) => {
                     // Retry unless the user intentionally canceled the operation.