Add tests to narrowly demonstrate defect mentioned in PR #682 #737
Conversation
If wanted, put them in separate tests.
|
This looks excellent, thank you. However, I'm not so sure about that last commit. We want |
Also, put back previously removed options.ignoreInitial in tests.
test.js
Outdated
| @@ -636,12 +636,97 @@ function runTests(baseopts) { | |||
| }); | |||
| }); | |||
| }); | |||
| it('should detect unlink unfazed by in other directory watching a second file that does not exist', function(done) { | |||
There was a problem hiding this comment.
little phrasing tweaks
should detect unlink while watching a non-existent second file in another directory
There was a problem hiding this comment.
implemented all phrase changes as requested, thank you
test.js
Outdated
| })); | ||
| }); | ||
| }); | ||
| it('should detect unlink and re-add unfazed by watching a second file', function(done) { |
There was a problem hiding this comment.
should detect unlink and re-add while watching a second file
test.js
Outdated
| })); | ||
| }); | ||
| }); | ||
| it('should detect unlink and re-add unfazed by in other directory watching a second file that does not exist', function(done) { |
There was a problem hiding this comment.
should detect unlink and re-add while watching a non-existent second file in another directory
test.js
Outdated
| })); | ||
| }); | ||
| }); | ||
| it('should detect unlink and re-add unfazed by in same directory watching a second file that does not exist', function(done) { |
There was a problem hiding this comment.
should detect unlink and re-add while watching a non-existent second file in the same directory
In this case a watched file is being unlinked and re-added, while another watched file is not unlinked, and one watched file does not exist.
|
The very last test added originated from a thought experiment. Admittedly, it didn't prove a suspicion. Nevertheless, a correct implementation should pass that test too. A good fix should pass this test without a problem, a bad fix would have to fear it. |
This one in 2.0.4 fails non-polling yet passes polling.
|
A main objective here was to make clear there was no regression. Seeing 2.0.4 released, that goal appears achieved. There is "only" a newly discovered defect. The additional test cases here I have added for sanity, as guidance for implementing a fix. I have tried implementing a fix too, as a public service, but haven't figured one. Watching individual files is not a current use case for our work, so this is not a priority for me, and I will not proceed trying to implement a fix, not at this time. |
This isn't a fix yet, but more narrowly demonstrates scope of defect mentioned in PR #682.
The referenced broadly failing test could create an impression this software library had gotten worse. In fact, it hadn't gotten worse. These tests now spell out more narrowly what still works as always and continues to work, versus what is broken and apparently had been broken for some time.