[EMAIL PROTECTED] wrote: > On Mar 14, 6:38 am, Benjamin Smedberg <[EMAIL PROTECTED]> wrote: >> [EMAIL PROTECTED] wrote: >>> I may have missed this, but do we have a dehydra script to look for >>> objects that are ref-counted, that contain what appear to be member >>> pointers to other ref-counted types, that are not users of the >>> NS_CYCLE_COLLECTION_*. We'd want this soon, in case something turns up >>> that's easy to fix for Gecko 1.9. Thanks, >> This analysis doesn't sound hard, but you're going to end up with lots of >> false-positives: i.e. all the necko classes, which certainly have >> member-comptrs but can't be cycle-collected because they are multi-threaded >> objects. > > This is no objection, it's a trivial tweak to my (very non- > algorithmic) sketch! ;-) > > Of course, we have some bogo-threadsafe XPCOM classes that are main- > thread only and should be considered in the scan. Never said it was > easy, but the decision tree to reduce false positives beckons already.
To get fewer false positives we could only warn for members that hold references to objects that participate in cycle collection. / Jonas _______________________________________________ Dev-static-analysis mailing list [email protected] https://lists.mozilla.org/listinfo/dev-static-analysis
