554649 - enabling AVMFEATURE_INTERIOR_POINTERS=1 will fail to compile

Status: RESOLVED FIXED

(Tamarin Graveyard :: Garbage Collection (mmGC), defect, P4)

Description

[Edwin Smith]

Due to misspelling kHasInteriorPointers in this code in GC-inlines.h

 REALLY_INLINE GCWorkItem::GCWorkItem(const void *p, uint32_t s, GCWorkItemType workItemType)
     : ptr(p)
 #ifdef MMGC_INTERIOR_PTRS
     , _size(s | uint32_t(kHasInteriorPointers) | uint32_t(workItemType))
 #else
     , _size(s | uint32_t(workItemType))
 #endif

the code probably isn't well tested.

possible fixes:
1. fix the misspelling and hope for the best.
2. remove the feature entirely, less configuration surface to worry about.

Comment 1

[Edwin Smith]

Lars's comments in bug 541338 make a case for keeping the feature ifdef, but it was a while ago and the facts on the ground might have changed since then.

Comment 2

[Lars T Hansen]

Attachment: Obvious patch

Comment 3

[Lars T Hansen]

(In reply to comment #1)
> Lars's comments in bug 541338 make a case for keeping the feature ifdef, but it
> was a while ago and the facts on the ground might have changed since then.

We should keep this as a debugging option for the time being, it is known to be useful when there are strange GC bugs.  There isn't much of a cpu cost to the feature, but it increases the chance of false retention of storage.  Once we move more in the direction of exact marking we should probably remove the feature.  I've lobbied in bug #557629 that we should demote it to a tweak.

Comment 4

[Felix S. Klock II [:pnkfelix, :fklock]]

Comment on attachment 437601 [details] [diff] [review]
Obvious patch

Confirmed that patch fixes compilation with ../configure.py --enable-interior-pointers

Comment 5

[Lars T Hansen]

tamarin-redux changeset:   4350:b4aa6a0f4c9a