Comment #4 on issue 2692 by [email protected]: Crankshaft produces
wrong code for plus arithmetics.
http://code.google.com/p/v8/issues/detail?id=2692
I've looked into this today. Summary of my findings so far:
- the provided repro case doesn't reproduce the issue with plain
bleeding_edge any more.
- but it still fails with the 3.18 branch.
- it also fails with --noanalyze-environment-liveness on bleeding_edge.
- it still fails with --hydrogen-filter='$$._SHA1._updateHash$1'
- it's somewhat flaky; --noparallel-recompilation makes it reliable.
- --noopt-safe-uint32-operations makes it pass.
- --nouse-inlining makes it pass (maybe due to different register choices?).
- looks like it's some bad interaction of inlining, spilling, and uint32.
- I'm pretty sure that r13534 hasn't introduced it but rather exposed it.
- I've been able to trim down the repro case a little by commenting out
most of the "if (foo) throw ..." checks in _updateHash$1(); if someone
finds an even smaller repro case that would be appreciated.
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings
--
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
---
You received this message because you are subscribed to the Google Groups "v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/groups/opt_out.
