| commit | 0bf1231b88c95b474ed00a6471a407fe588c6255 | |
| author | Douglas Katzman <dougk@google.com> | |
| Thu, 14 Sep 2017 14:11:05 +0000 (14 10:11 -0400) | ||
| committer | Douglas Katzman <dougk@google.com> | |
| Thu, 14 Sep 2017 14:25:21 +0000 (14 10:25 -0400) | ||
| tree | 284b245eac9019d8173eb2c9f2cee69263d2f38b | treesnapshot (tar.gz zip) |
| parent | 8408fe5a71b5c52b2394408fa9d083594bf5fbb9 | commitdiff |
Fix #+(and immobile-space (not compact-instance-header)) some more
* set_instance_layout always stored a layout into the high bits of an instance
header even though it should be in slots[0]. Using it was seemingly harmless,
but the object got a bad pointer in its header, confusing if naught else.
Defining instance_layout() to be an lvalue makes it clearer, and exposed a
bug that fix_object_layout() before the call in instance_scan() during
defragmentation touched the stored instance layout more than once.
* Rename fin_layout() to funinstance_layout()
* Add another assertion into verify_range()
* Don't use instance_layout() to mean function_layout()
* set_instance_layout always stored a layout into the high bits of an instance
header even though it should be in slots[0]. Using it was seemingly harmless,
but the object got a bad pointer in its header, confusing if naught else.
Defining instance_layout() to be an lvalue makes it clearer, and exposed a
bug that fix_object_layout() before the call in instance_scan() during
defragmentation touched the stored instance layout more than once.
* Rename fin_layout() to funinstance_layout()
* Add another assertion into verify_range()
* Don't use instance_layout() to mean function_layout()