Conversation
|
Review requested:
|
adcda63 to
b1c015c
Compare
Codecov ReportAll modified and coverable lines are covered by tests β
Additional details and impacted files@@ Coverage Diff @@
## main #55194 +/- ##
==========================================
+ Coverage 88.39% 88.40% +0.01%
==========================================
Files 652 652
Lines 186565 186565
Branches 36046 36038 -8
==========================================
+ Hits 164916 164936 +20
+ Misses 14908 14902 -6
+ Partials 6741 6727 -14 π New features to boost your workflow:
|
|
Seemingly related test failures: |
|
I can't reproduce the failures locally |
|
Removing
These look very much like endianness issues, especially since they're failing on AIX and LinuxONE which are both big endian platforms. |
|
I don't have a big endian machine to test this one, but it's likely, there may be a bug π€. Drafting until whatever this is is resolved. |
I think it's a bug in WPT. Taking, for example, node/test/fixtures/wpt/FileAPI/blob/Blob-constructor.any.js Lines 285 to 299 in b1c015c Typed arrays are by definition
which means, e.g. that the bytes for |
|
This is a bug in node. I will give some more context in a little bit. |
|
I forgot to follow up, but here are the steps (this follows through multiple specs so I will link each one).
|
|
Thanks for the details! I'll update these files later today, and if this is still the case, I'll open an issue in the WPT repository |
@KhafraDev 's analysis is that this is a Node.js bug, not a WPT one. |
|
Oh, I misunderstood. Got it, keeping this is a draft until that's fixed |
|
Adding some info here since this issue has resurfaced by #61750: meaning there is no byte reversing happening. The main issue is simply the use of TypedArrays which are host endian dependent. We can either make use of |
git node wpt FileAPINote one new failure:
Passing a Float16Array as element of the blobParts array should work.Note that several tests now pass that previously failed.