Thanks for looking into this. I've stopped counting how often I've crashed due to this.

> Apparently, passing empty region name to the MapNameRequest makes server return
> empty name in MapBlockReply, which triggered the assertion.

Shouldn't we warn rather than assert when we receive invalid (or just useless) data from remote (whether by our own or the servers' fault)? I don't think any service should be given the power to 'shoot down' the viewer.

indra/newview/llfloaterworldmap.cpp (Diff revision 1)

Instead of returning early ...

1. Vadim ProductEngine Aug. 7, 2011, 2:58 p.m. (Aug. 7, 2011, 2:58 p.m.)

   There's no point in executing the rest of the method code, so we're returning.
   And I'd rather add a return than increase indentation, thus making hg blame useless at finding out why the indented code was written.

indra/newview/llfloaterworldmap.cpp (Diff revision 1)

... you could wrap the remainder of the method into a conditional block.

1. Vadim ProductEngine Aug. 7, 2011, 2:59 p.m. (Aug. 7, 2011, 2:59 p.m.)

   There's no point in executing the rest of the method code, so we're returning.
   And I'd rather add a return than increase indentation, thus making hg blame useless at finding out why the indented code was written.

I don't agree with your early return rationale, but let's just go with this.

> The assert will not be in issue for release builds.
> Triggering it in debug builds help catching bugs, as in this case.
> Besides, the assert was added in attempt to locate another bug (EXT-4568), so I'd just leave it.

Sounds reasonable.