All functions, that uses cf_ReadBytes(), checks return size, but don't handle potential exceptions. That leads to segfaults on reading of damaged files (i.e. broken savegames).
Before this change, cf_OpenLibrary() did something along the lines of
this:
char id[4];
fread(id, 4, 1, fp);
strncmp(id, "HOG2", 4);
If fread() finishes successfully, then that code is fine. However,
fread() might encounter an error or bump into the end of a file. In
those scenarios, the value of id will not necessarily be initialized
[1]. In other words, when fread() fails, strncmp() might operate on
uninitialized memory.
This change makes sure that the value of id only gets used if fread()
succeeds. Additionally, this change fixes a GCC warning about ignoring
fread()’s return value.
[1]: <https://en.cppreference.com/w/cpp/io/c/fread>
Before this change, cf_OpenLibrary() was inconsistent. Sometimes, it
would refer to the size of a HOG2 id using the literal 4. Other times it
would refer to the size of a HOG2 id using strlen(HOG_TAG_STR). There
was a good reason for this. Some compilers allow you to do this:
char id[strlen(HOG_TAG_STR)];
Other compilers throw an error if you try to do that [1].
This commit makes cf_OpenLibrary() more consistent. It makes it so that
cf_OpenLibrary() always uses the same constant expression when referring
to HOG_TAG_STR’s length.
[1]: <https://devblogs.microsoft.com/oldnewthing/20221114-00/?p=107393>
* Begin by marking functions and variables as static when needed.
* More work.
* More work.
* More pokes.
* More work.
* More work.
* Initial work on the netgames.
* Revert changes to the license header on source files.
* clutter.cpp poke.
* One final poke.
* Move some declarations to headers:
Move paged_in_count and paged_in_num to gamesequence.h
Move DoneLightInstance and StartLightInstance to polymodel.h
* Look over the AI script/plug-ins.
* Going over the changes one last time.
* Fix rebase errors.
* More migration from bare statics to static inlines.
The vast majority of this is fixing up `char *` that should be `const char *`
but a handful of other fixes, like potential buffer overflows that GCC
noticed, etc, were applied as well.
This removes `-Wno-write-strings` from CMakeLists.txt, as it is no longer
necessary, as there is no longer a flood of compiler warning spam when
building.
This does not fix all compiler warnings; there are still a handful, and they
are legitimate, but they can be dealt with in a future commit.
The MACINTOSH define refers to MacOS Classic (not OS X) which we do not
plan to support. Rather than carry the cruft forever, let's delete it.
NOTE: legacy/ is unused but we're keeping it around, so MACINTOSH uses
there are left alone.
Process used for this commit:
```
git rm -r mac
find . \( -iname "*.c" -or -iname "*.cpp" -or -iname "*.h" \) \
-exec unifdef -UMACINTOSH -o {} {} \;
git restore legacy
git add .
```
Test Plan:
On Mac, build both Debug and Release
```
cmake --build --preset mac --config Debug
cmake --build --preset mac --config Release
```
Other modules that depends on it, can reuse includes on linking.
There some files formally not belonging any packages (lib directory), as workaround there temporary include_directories(cfile) on root of project. After migrating all modules this can be removed.
