This quells warnings from CMake about reserved target names, and reduces confusion about which `test` is intended: the wesnoth executable or the standard shell command.
In most cases, it should now use less memory and be faster to construct.
In the other cases, blurring used to be incorrect because the buffer was too small.
* Some doc/formatting cleanup
* Removed some unnecessary functions
* Added some explanatory comments
* Removed some commented out code
* Simplified a few things.
Apparently nicks require at least 1 alphanumeric character, meaning the previous message
was rather confusing, since the apparently valid nickname of "-_-" did not work.
This fixes#2156.
preferences::add_acquaintance replaces the existing friend entry if the given username
has already been added to the acquaintance list, but the prefs dialog always added a
new row unconditionally. I've changed it to only add a new row if the username creates
a new entry, else replace the appropriate exiting row.
Coverity wants a cast to double. Add it and it wants a cast to long long. Add THAT and Coverity and GCC both complain.
I suppose there might be a way to rewrite things to make everyone happy. But it's easier to tell Coverity it's a false positive and force the issue closed.
This so-called header is actually just splitting a large source file into two files. I would mark these functions static but this is a header; so I'm pre-declaring them in case some other compiler complains about static functions in a header.
Hopefully someone in the future will do something about this smelliness.
This fixes#2154. For some reason, after opening help from the File menu and then quitting to desktop,
the number of labels recorded in the label map and context set weren't equal. I'm not exactly sure
why, since the code seems to indicate they should be, and the loop doesn't seem to be hit when only
invoking help with the hotkey (F1).
This just removes the predication of a label being present in the label map for removal from the context
set. I also removed two useless, unused variables that I'm surprised no one's complained about.
This commit restores redraws of entire windows, but only if a window below
has actually changed. It should fix most regressions from commit c6d8692ac6
with a much smaller CPU usage cost.
The commit also includes a check for the situation where the title screen
changes all the time, which would restore the high CPU usage (that the
developer would be unlikely to notice immediately).
This reverts commit 07b04775ff.
This code is performance-critical: adding a branch for every pixel in
the surface isn't a good idea. @GregoryLundberg was fine with reverting
the commit, as well. The same is true for the next two reverts.
Fixing Coverity Scan warnings is good and all, but we shouldn't
sacrifice performance for it.
When you fix the complaint about implicitly converting a long long to a double you get a complaint about implicity discarding the remainder from an integer division.
Mumble. Grumble. Glare at Coverity.
Closes CID 1382228
With the change from SDL_mixer 2.0.1 to 2.0.2, we cannot simply check if SDL_mixer can open an OGG/Vorbis file, we must first open an audio device. [This is probably a mistake, since the documentation implies simply initializing the codec should be enough.] To avoid future problems, upgrade the feature-test program to perform a full initialization and shut-down; everything except actually playing the sound.
Closes#2137
