(#326) - Update install and fixed polling problem #329
Conversation
.gitignore - A bunch of files it would be good not to check in :) test-setup.sh - To adapt to the new way of building PouchDB I now clone PouchDB into its own separate directory, pouchdbclone and then do an install and build, symbolically link to packages/pouchdb and install that as well. test-pouchdb-minimum.sh - The tests (as opposed to the live code) now lives in pouchdbclone test-pouchdb.sh - Same as previous changes - As explained in pouchdb#326 I tried to hook all the obvious events and literally nothing fired when a connection was lost. So instead when the heartbeat runs I check if we have lost the socket and if so then I clean up. I also made cleanup able to cancel any live changes objects and call res.end.
|
I'm tracking Travis but I actually expect this to fail because of #328. But I have spent all day on this and I need to get back to coding my day job. So if it fails as expected then I'll have to see if I can find time to fix. But as I noted in #328 the failure was always there, this checkin didn't create. It's just that we didn't re-run the tests after putting in the hook to detect unhandled promise rejections. |
|
Looks like i spoke way too soon. It's dying for a completely different reason. But what the heck did I possibly do to break jshint? |
|
@yaronyg you need to lock jshint down to 2.8.0. They have a breaking change once they are above 2.9 |
|
@marten-de-vries, above @garrensmith makes the point that we shouldn't have JSHint above 2.8.0 because of breaking changes. I checked and you were the one who approved Greenkeeper bumping us up. I wanted to therefore check with you first that there isn't some specific reason why we went above 2.8.0? How did you get tests to pass? |
|
Just leaving myself a note that I need to update this PR based on the work @garrensmith does in pouchdb/pouchdb-server#140 |
|
@yaronyg Something weird is going on with the test setup then: https://travis-ci.org/pouchdb/express-pouchdb/jobs/124334134#L263 was green. Maybe it used a different jshint install (it definitely installed the new one, as you can see on the highlighted line), although I wouldn't know where it could find another one. Anyway, I approved that just as a routine update, so feel free to revert it if it's causing trouble now. |
|
@marten-de-vries O.k. I'll experiment and see if changing the version fixes things. |
|
I've got a PR #342 that should fix all these issues. |
|
@yaronyg could you either close this PR and open another one with the changes work or adjust this one to just be the changes fixes. |
|
@garrensmith Makes sense. I'm going to keep this open for now so I can track and then close when I have the new PR ready. |
|
In the midst of preparing the new PR using the pouchdb-server monorepo. |
#326
.gitignore - A bunch of files it would be good not to check in :)
test-setup.sh - To adapt to the new way of building PouchDB I now clone
PouchDB into its own separate directory, pouchdbclone and then do an
install and build, symbolically link to packages/pouchdb and install that
as well.
test-pouchdb-minimum.sh - The tests (as opposed to the live code) now
lives in pouchdbclone
test-pouchdb.sh - Same as previous
changes - As explained in #326 I tried to hook all the obvious events and
literally nothing fired when a connection was lost. So instead when
the heartbeat runs I check if we have lost the socket and if so then
I clean up. I also made cleanup able to cancel any live changes objects
and call res.end.