I reduced memory fragmentation using 2 principles: Plugins should be loaded for as little time as possible, and long lasting memory allocations should be allocated before plugins are loaded. There might still be a little fragmentation left.
Note that command line settings that require plugins to be loaded don't work yet, but they didn't work (properly) before either.
svn-id: r54097
- Changed the new dithering checkbox to be more generic (i.e. relevant for all engines that may support such a feature in the future)
- Changed the checkbox to be unchecked by default. As discussed on -devel, game enhancements in ScummVM should be off by default, and opt-in
- Changed the option from "sci_undither" to "disable_dithering"
- Changed theme version style to X.Y.Z and bumped it to 0.8.2
svn-id: r54090
On my OS X machine, -Wglobal-constructors is always added by configure.
But the test runner makes heavy use of global constructors.
This change allows me to run "make test" even when using -Werror.
svn-id: r54083
* Remove DLObject virtual methods allocSegment and freeSegment.
As long as all DLObject implementations use memalign + free to
allocate/release segments, there is no point in wrapping those.
This enables further simplifications.
* Add TemplatedELFPlugin template class. Use this instead of explicit
ELFPlugin subclasses.
* Rename DLObject::discard_symtab to discardSymtab
svn-id: r54082
From our current stance, this is a game enhancement, and should be on by default, like the rest of the game enhancements (AGI mouse, save/load replacement etc)
svn-id: r54070
This generalises the workaround added for bug #3018918 i.e. UOTODDV issue, to cover the
ICVGCGT case and hopefully all future cases, by executing the variable correction during
the reload after Game Over and using the variable read by the script triggering the movement lock, rather than the movement lock variable (59) itself as this may not be set until a later script.
svn-id: r54061
The deletePointer() method approach cannot work, as it is called
by the destructor of the base class.
A possible correct solution would be to enhance ScopedPtr with a
"deleter" object like SharedPtr. But this seems overkill as long as we
need it in only one place.
svn-id: r54057
This merge was extremely difficult to carry out. It wasn't entirely SVN's fault -- there were several merges to the branch that were done by hand. Please check for any issues and regressions. Also note that the DS makefile was not copied over since the "one at a time" plugin mode currently has too much fragmentation ie. it doesn't work.
svn-id: r54051
