[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [MiNT] GCC 4.2.2 - Keith's Latest Progress (or lack thereof)



Hello Miro,

I think we should wait until we get 100% working (in term of comparing
to gcc 2.95) gcc4 and then make RPM from that.

The big question is do I create an RPM that just goes ahead and replaces
the installed GCC 2.95?

I think this shouldn't be problem, just create gcc with the suffix
-4.2.2 and everybody is free to choose (and link to /usr/bin/gcc the
one he prefers). But of course, this assumes we'll solve that problem
with gcc2-mintlib and gcc4 compilation.

What I'll try to do is build an RPM to share that installs it like that, which is what I do already. I have GCC 2.95.3 installed in the default location, and GCC 4.1.2 (now 4.2.2) installed under /usr/local. I renamed the 2.95.3 binaries, and already made gcc4 with --program-suffix=-4.2.2. So, all a user would need to do, if I could build an RPM, would be to manually rename the 2.95.3 bins, and create 2 scripts maybe, gcc2 and gcc4, which would just setup the symlinks from gcc, g++, etc to point to the version of GCC they want.

I am building ScummVM natively (to see if the thunk problems still exist
with GCC 4.2.2, already had one wierd issue with configure, their custom
configure did not like the cross compiled g++ (doing g++ -dumpversion,
reported bad version when looking for 2.95 or > up to 4.9.9) but liked the
native compiled one, really wierd, but thats something I'll look into more
later.  I might start with binutils tonight when it completes.

huh that's pretty strange. so you say your only problem with c++ was
you used old binutils patch? i don't remember now, how g++4 worked
with binutils2.13? I mean with native build.


I still hit the thunk problem I was having with ScummVM with the newer g++. I will look into this further after I decide if I can easily build an RPM for people to test. The fix (or cheap hack) I had to do in my version of GCC (4.1.2) was done in mint.h:

/* This is the assembler directive to equate two values.  */
/* Fix for setlTHUNK ?
#undef SET_ASM_OP
#define SET_ASM_OP    ".set"
*/

Vincent refined this some in his patch (I just glanced thru and noticed it was different), so I will look into that further in the next few days/week. I'll try to share my native build before I go ahead and totally break it! :)

Keith