Prevent loss of characters in case of false end-partial-match #454
Conversation
|
Thanks! Nice work. The tests for this area are here: https://github.com/ef4/ember-auto-import/blob/a88004f656df9e08d3d5e7e3eb337e0cd5727b9e/packages/ember-auto-import/ts/tests/analyzer-test.ts#L379-L490 If you can get a test case in here that fails without your fix and passes with it, we'll be good to go. |
|
@ef4 thanks for the pointer. Test added. |
| assert.ok( | ||
| meta.slice(MARKER.length, -MARKER.length).indexOf(MARKER[0]) > -1, | ||
| 'serialized sample data must contain first character of MARKER somewhere between boundary markers for test to have meaning' | ||
| ); |
There was a problem hiding this comment.
This assertion isn't really checking the result of running code -- its just checking that our sample data is sufficient for this test case. If MARKER or the sample data were changed in the future, such that the first character of MARKER didn't show up somewhere in the serialized data, this test would become invalid.
|
Thanks, this is going out today as 2.2.1. |
Attempted fix for #449
I believe I ran into the same issue, and built on the debugging that @vlascik had done. I don't have a deep understanding of how this code fits into the overall build pipeline, but was able to wrap my head around the code in this module enough to propose a fix.
This addresses the case where we found a partial match for
MARKERat the end of the last chunk, but now that we're looking in the next chunk we've realized it wasn't really a match. The fix is to make sure the characters we sliced off of the last chunk don't get lost.