summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorSimon Howard2007-06-21 11:32:04 +0000
committerSimon Howard2007-06-21 11:32:04 +0000
commitee2a6fda9fd817c191d74d0f1a00a4c05b4ed433 (patch)
tree7bcfdbd7969e12248887e026843576debdb28189
parentbec32345ba1b4a32497d23af18e09c15b5e812f8 (diff)
downloadchocolate-doom-ee2a6fda9fd817c191d74d0f1a00a4c05b4ed433.tar.gz
chocolate-doom-ee2a6fda9fd817c191d74d0f1a00a4c05b4ed433.tar.bz2
chocolate-doom-ee2a6fda9fd817c191d74d0f1a00a4c05b4ed433.zip
Add portability section to HACKING file.
Subversion-branch: /trunk/chocolate-doom Subversion-revision: 921
-rw-r--r--HACKING31
1 files changed, 31 insertions, 0 deletions
diff --git a/HACKING b/HACKING
index ee2436f7..699d42e8 100644
--- a/HACKING
+++ b/HACKING
@@ -113,6 +113,37 @@ void FunctionName(int argument, int arg2, int arg3, int arg4, int arg5,
} while (condition);
}
+Portability
+===========
+
+Chocolate Doom is designed to be cross-platform and work on different
+Operating Systems and processors. Bear this in mind when writing code.
+
+Do not use the long type (its size differs across platforms; use
+int or int64_t depending on which you want).
+
+Use Doom's byte data type for byte data. 'int' is assumed to be a
+32-bit integer, and 'short' is a 16-bit integer. You can also use the
+ISO C99 data types: intN_t and uintN_t where N is 8,16,32,64.
+
+Be careful with platform dependencies: do not use Windows API
+functions, for example. Use SDL where possible.
+
+Preprocessor #defines are set that can be used to identify the OS
+if necessary: _WIN32 for Windows and __MACOSX__ for MacOS X. Others
+are set through SDL. Try to avoid this if possible.
+
+Be careful of endianness! Doom has SHORT() and LONG() macros that
+do endianness conversion. Never assume that integer types have a
+particular byte ordering. Similarly, never assume that fields
+inside a structure are aligned in a particular way. This is most
+relevant when reading or writing data to a file or a network pipe.
+
+For signed integers, you shouldn't assume that (i >> n) is the same as
+(i / (1 << n)). However, most processors handle bitshifts of signed
+integers properly, so it's not a huge problem.
+
+
GNU GPL and licensing
=====================