aboutsummaryrefslogtreecommitdiff
path: root/engines/tinsel/coroutine.h
diff options
context:
space:
mode:
Diffstat (limited to 'engines/tinsel/coroutine.h')
-rw-r--r--engines/tinsel/coroutine.h4
1 files changed, 2 insertions, 2 deletions
diff --git a/engines/tinsel/coroutine.h b/engines/tinsel/coroutine.h
index 8cdbd44b19..30d90970b6 100644
--- a/engines/tinsel/coroutine.h
+++ b/engines/tinsel/coroutine.h
@@ -35,7 +35,7 @@ namespace Tinsel {
* <http://www.chiark.greenend.org.uk/~sgtatham/coroutines.html>.
* However, many improvements and tweaks have been made, in particular
* by taking advantage of C++ features not available in C.
- *
+ *
* Why is this code here? Well, the Tinsel engine apparently used
* setjmp/longjmp based coroutines as a core tool from the start, and
* so they are deeply ingrained into the whole code base. When we
@@ -45,7 +45,7 @@ namespace Tinsel {
* would have meant a major restructuring of the entire code base, a
* rather daunting task. Also, it would have very likely introduced
* tons of regressons.
- *
+ *
* So instead of getting rid of the coroutines, we chose to implement
* them in an alternate way, using Simon Tatham's trick as described
* above. While the trick is dirty, the result seems to be clear enough,