Track / conform to spec encoding of reserved zero byte for memory / table index
Categories
(Core :: JavaScript: WebAssembly, enhancement, P3)
Tracking
()
Tracking | Status | |
---|---|---|
firefox67 | --- | fixed |
People
(Reporter: lth, Assigned: lth)
References
Details
Attachments
(1 file)
3.70 KB,
patch
|
jseward
:
review+
|
Details | Diff | Splinter Review |
Currently we're decoding the "reserved index" byte for source/target memories/tables as a varuint32, while the spec has u8-must-be-zero.
Once we land reftypes the table index will necessarily become a varuint32, but until then, I'm having a slightly difficult time pushing through a change so that all of them are varuint32. See https://github.com/WebAssembly/bulk-memory-operations/issues/36#issuecomment-457550639 et seq.
If my proposal is rejected then we must update the code so that (a) for memories we decode as a byte always and it must be zero and (b) if bulk-memory is defined but reftypes is not then for tables we decode ditto, but if reftypes is defined then we decode as varuint32.
Assignee | ||
Comment 1•5 years ago
|
||
The sentiment is to keep this as a zero byte, so we need to update our code (slightly) here.
Assignee | ||
Comment 2•5 years ago
|
||
Comment 3•5 years ago
|
||
Comment on attachment 9041145 [details] [diff] [review] bug1523908-zero-placeholder-index-byte.patch Review of attachment 9041145 [details] [diff] [review]: ----------------------------------------------------------------- Nicely future-proofed. Looks fine.
Pushed by lhansen@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/1431d60aedcc Require zero byte encoding for memory/table index in the absence of multi-memory/multi-table. r=jseward
Comment 5•5 years ago
|
||
bugherder |
Description
•