Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Some Mac build TODOs... #125

SDLBugzilla opened this issue Feb 10, 2021 · 0 comments

Some Mac build TODOs... #125

SDLBugzilla opened this issue Feb 10, 2021 · 0 comments


Copy link

This bug report was migrated from our old Bugzilla tracker.

Reported in version: HG 1.2
Reported for operating system, platform: Mac OS X (All), PowerPC

Comments on the original bug report:

On 2006-04-13 10:55:37 +0000, Ryan C. Gordon wrote:

I figured I'd put down some notes on how to build a Universal binary from the configure script, or at least what it would need to do.

Ideally, we need the Makefile to build each file twice on the Mac: once with gcc-3.3 for PowerPC, and once with gcc-4.0 for x86, and then use "lipo" to glue them into one object file.

so you end up with something like this:

gcc-3.3 -o myfile-ppc.o -arch ppc $OTHER_COMPILER_FLAGS myfile.c
gcc-4.0 -o myfile-x86.o -arch i386 $OTHER_COMPILER_FLAGS myfile.c
lipo -create -o myfile.o myfile-ppc.o myfile-x86.o
rm -f myfile-ppc.o myfile-x86.o

This could either be a special make rule for Mac OS X, or perhaps a shell script that wraps the compiler ("export" or something)

Also, I think we can keep 10.2.8 compatibility on the PowerPC side (and 10.4 on the x86 side, which won't always be the latest version), if we do some tapdancing with command lines. This should fix the issues with the Quicktime SDK forcing 10.4 as a baseline.

On PowerPC:

CFLAGS := -arch ppc -F/Developer/SDKs/MacOSX10.2.8.sdk/System/Library/Frameworks -I/Developer/SDKs/MacOSX10.2.8.sdk/Developer/Headers/FlatCarbon -DMAC_OS_X_VERSION_MIN_REQUIRED=1020 -I/Developer/SDKs/MacOSX10.2.8.sdk/usr/include/gcc/darwin/3.3 -I/Developer/SDKs/MacOSX10.2.8.sdk/usr/include/gcc/darwin/3.3/c++ -I/Developer/SDKs/MacOSX10.2.8.sdk/usr/include/gcc/darwin/3.3/c++/ppc-darwin -isystem /Developer/SDKs/MacOSX10.2.8.sdk/usr/include

LDFLAGS := -arch ppc -mmacosx-version-min=10.2 -L/Developer/SDKs/MacOSX10.2.8.sdk/usr/lib/gcc/darwin/3.3 -F/Developer/SDKs/MacOSX10.2.8.sdk/System/Library/Frameworks -Wl,-syslibroot,/Developer/SDKs/MacOSX10.2.8.sdk

On Intel:

CFLAGS := -arch i386 -mmacosx-version-min=10.4 -DMAC_OS_X_VERSION_MIN_REQUIRED=1040 -isysroot /Developer/SDKs/MacOSX10.4u.sdk

LDFLAGS := -arch i386 -mmacosx-version-min=10.4 -L/Developer/SDKs/MacOSX10.4u.sdk/usr/lib/gcc/i686-apple-darwin8/4.0.0

The '-F' command line is an Apple GCC thing: it tells it where to look for headers in frameworks. The rest is tapdancing to make sure all system headers use the ones from the 10.2.8 SDK, which can be installed from XCode's installer (but is not installed by default). gcc4 added the isysroot option, which cleans things up a lot.

These instructions can build on either a PowerPC or Intel Mac. The build system may need to be 10.4 or later.


On 2006-04-13 11:05:08 +0000, Sam Lantinga wrote:

If you have merged ppc and x86 objects, how do you use separate LDFLAGS?

On 2006-04-13 11:15:35 +0000, Sam Lantinga wrote:

If you show me how to build a fat .dylib, I'll see if I can integrate it into the SDL build system...

On 2006-04-13 11:17:08 +0000, Ryan C. Gordon wrote:

Whoops, I changed direction in mid-post and forgot to correct that.

"lipo" should be used on the final binaries. So you compile and link them seperately, then merge them.

Something I've been doing in Makefiles recently looks like this:

set up OBJS, PPC vs X86 compiler defines, etc up here.

make the objects to build into two seperate copies in different places...

X86OBJS := $(foreach f,$(OBJS),$(BINDIR)/x86/$(f))
PPCOBJS := $(foreach f,$(OBJS),$(BINDIR)/ppc/$(f))

Build with gcc4,x86,10.4.0 options in the "x86" dir

$(BINDIR)/x86/%.o : %.c
$(X86CC) -o $@ $(CFLAGS) $(X86CFLAGS) $<

Build with gcc3,ppc,10.2.8 options in the "ppc" dir

$(BINDIR)/ppc/%.o : %.c
$(PPCCC) -o $@ $(CFLAGS) $(PPCCFLAGS) $<

Link x86 objects into something like SDL-x86.dylib

$(X86EXE) : $(X86OBJS)
$(X86LD) -o $@ $(LDFLAGS) $(X86LDFLAGS) $^

Link ppc objects into something like SDL-ppc.dylib


merge SDL-ppc.dylib and SDL-x86.dylib into SDL.dylib

$(EXE) : $(X86EXE) $(PPCEXE)
lipo -create -o $@ $^


On 2006-04-14 01:44:02 +0000, Sam Lantinga wrote:

Ryan, I added build-scripts/ to CVS.
It's completely untested, could you check it out?

On 2006-04-15 08:11:44 +0000, Max Horn wrote:

FYI, the Apple porting guides at suggests

CFLAGS="-isysroot /Developer/SDKs/MacOSX10.4u.sdk -arch i386 -arch ppc"
LDFLAGS="-Wl,-syslibroot,/Developer/SDKs/MacOSX10.4u.sdk -arch ppc -arch i386"

But I am not sure whether -isysroot is available in gcc 3.3, probably not... Nor do I know whether one can use this to build binaries that work on 10.3 / 10.2, this'll require some experimenting.

Another potentially interesting read:

On 2006-04-16 13:56:04 +0000, Sam Lantinga wrote:

So, has anyone tried the script?

On 2006-04-16 14:06:07 +0000, Max Horn wrote:

I'll give it a try now (though I have no intel machin myself). BTW it's missing the executable bit.

On 2006-04-17 01:07:39 +0000, Sam Lantinga wrote:

I fixed a couple things, go ahead an try it again...

On 2006-04-17 02:41:55 +0000, Max Horn wrote:

It fails to compile the PPC version over here, because it mixes files from my regular /usr/include and from /Developer/SDKs/MacOSX10.2.8.sdk/

I really think using the options listed in the links I provided is necessary. Even then, when using GCC 4.0, I am not sure if the resulting binaries will actually work on 10.2/10.3...

Anyway, -nostdinc, -isysroot seem like the way to go...

On 2006-04-17 14:00:35 +0000, Ryan C. Gordon wrote:

There is no -isysroot in gcc 3.3, only gcc4, which means you have to surrender 10.2 (and 10.3?) support on PowerPC to use it.

-isystem puts that path at the start of the include directory list, so it should never favor /usr/include if it has the same file in the -isystem path.

Of course, it's probably worth noting that this will silently ignore the -isystem path if it doesn't exist...which it won't, if the 10.2.8 SDK isn't installed when installing XCode (and it isn't, by default).

Are we having fun yet? :)


On 2006-04-17 22:30:01 +0000, Sam Lantinga wrote:

What should the fat dylib be called?

$ ls -d /build/.libs/.dylib
ppc/build/.libs/libSDL- x86/build/.libs/libSDL-
ppc/build/.libs/libSDL-1.2.0.dylib x86/build/.libs/libSDL-1.2.0.dylib
ppc/build/.libs/libSDL.dylib x86/build/.libs/libSDL.dylib

$ find ppc x86 -type f -name "*.dylib"

Does the libSDL.a also need to be built as a stub for dynamic linking, or is that a static library?

On 2006-04-18 13:51:59 +0000, Ryan C. Gordon wrote:

The final files should be called what they are normally called (I think it's libSDL- and the rest are symlinks to it)...the one file will have both ppc and x86 code, and the OS and tools will pick the right half for what they need.

libSDLmain.a needs the same treatment: build both x86 and ppc, merge them with lipo, and then the linker will choose the right piece depending on what it's linking it to.

We only need them seperate when first compiling, but once everything is built, it should exist on the system as it always would: it will just contain both architectures.

I'm not sure I'm answering your question properly.


On 2006-04-18 23:51:24 +0000, Sam Lantinga wrote:

Yep, that was exactly what I needed. I think this is pretty much wrapped up.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
None yet
None yet

No branches or pull requests

1 participant