Re: boost 1.53 breaks monotone build

Richard Hopkins <[email protected]> writes:

>> 
>> On my local machine, resolve_conflicts_dropped_modified_1
>> fails
>> constantly.
> Same here! Glad I'm not the only one; I was seriously starting to doubt
> my hardware, OS, or config, especially as I had it succeed once or twice,
> the backtrace kept changing, and that it hasn't failed yet on Windows.

In 79e73d6882ea68c70abd5891f47806eabe5d0f10 on 16 March 2013 (3 days
ago), I fixed a bug related to that; there was this snippet of code in
merge_content.cc:

      else
        {
          parents.erase (i);
          set<revision_id> more_parents;
          db.get_revision_parents(*i, more_parents);
          parents.insert(more_parents.begin(), more_parents.end());
        }

This has *i after erase (i), which can have random effects depending on
the stack/heap content.

It is now:

      else
        {
          set<revision_id> more_parents;
          db.get_revision_parents(*i, more_parents);
          parents.erase (i);
          parents.insert(more_parents.begin(), more_parents.end());
        }

This fixed a hang on Windows for me.

Let me know if it fixes your problems.

-- 
-- Stephe