This reverts commit 2c201a162639b54f58e441f53edec4198edfc7d3, reversing
changes made to 76b9976b75842519acdd9150dd9b6463cb91aec8.
I accidentally merged a topic branch into master, I am attempting to go back
to the merge PR 94 status of master just previous to my accidental commit.
This is intended to address bug #3856.
Following discussion on IRC, this is considered a feature request
and not a bug, and so cannot be addressed during feature freeze
for 1.12. Hence I am pushing it to a topic branch.
I can't push a changelog change either as we don't know what version
this will go in, but the changelog entry should be something like:
"Fix bug #3856: The turn dialog used in hotseat MP play now applies
a blindfold to the map for the duration of the dialog."
this is intended to make linux dbus notifications a bit more useful, as it appears on my machine that the current notifications only display the messages in the most distant past...
the is_local() function should return true if a side is idle. if not this causes multiple side init events when a side disconnects and is assigned idle, and thus makes the server unhappy.
Two changes might warrant further explanation:
1) No need to set a property of an object (current_label->second)
that you are about to erase.
2) Setting res to NULL when it is already NULL is a no-op. (This
line exists because at one time res was set to something other
than NULL.)
In two places, the code from get_label() was basically duplicated,
presumably because the public get_label() function returns a pointer
to a const object. I added a private version that returns a pointer
to a non-const object, and now set_label() is easier to read.
The variable 'n' was declared within a scope that had already
declared 'n'. While each variable is only used in the line in which
it is defined (they are loop control variables for 'for' statements),
it's just bad style to have that potential confusion.
Plus, 'n' is not a descriptive name. (Not that the new names are
much better.)
