Pairing A C Compiler With QEMU's Code Generator

[phoronix]

Phoronix: Pairing A C Compiler With QEMU's Code Generator

Earlier this week when writing about the state of the Tiny C Compiler, I learned more about QCC. QCC is a new initiative to pair a forked version of the Tiny C Compiler (TCC) with QEMU's code generator...

http://www.phoronix.com/vr.php?view=MTIzMTQ

[Ibidem]

Landley tells me that QCC is going to be on the agenda once Aboriginal Linux switches from uclibc to musl (next release), and toybox (his BSD-licensed replacement for BusyBox, which you can extend by dropping in a single file for each applet you want to add) reaches 1.0.
Toybox is at 0.4.x right now, so I'm guessing ~1-2 years more...
QCC will be BSD-licensed; QCG is already, and he has gotten permission from Fabrice for using Fabrice's code in TCC under that license.
Much of the work on tcc 0.9.24 was based on Landley's work (http://landley.net/code/tinycc/), but some of the most recent changes are ones he may need to redo/inquire about. Specifically, he mentioned variable length arrays as something that needed to happen for the kernel to build.
As far as I can tell, tcc doesn't support weak symbols, so it doesn't work for compiling musl (what I was interested in).

And in case you're wondering why anyone wants to compile the kernel with TCC:
1-small codebase for a self-hosting environment
2-avoid GNU (for bloat and because of the FSF's attempts to rewrite history)
About the latter, here's Ulrich Drepper's comments from the glibc 2.2.4 release notes:

The glibc situation is even more frightening if one realizes the story
behind it. When I started porting glibc 1.09 to Linux (which
eventually became glibc 2.0) Stallman threatened me and tried to force
me to contribute rather to the work on the Hurd. Work on Linux would
be counter-productive to the Free Software course. Then came, what
would be called embrace-and-extend if performed by the Evil of the
North-West, and his claim for everything which lead to Linux's
success.


Which brings us to the second point. One change the SC forced to
happen against my will was to use LGPL 2.1 instead of LGPL 2. The
argument was that the poor lawyers cannot see that LGPL 2 is
sufficient. Guess who were the driving forces behind this.

The most remarkable thing is that Stallman was all for this despite
the clear motivation of commercialization. The reason: he finally got
the provocative changes he made to the license through. In case you
forgot or haven't heard, here's an excerpt:

[...] For example, permission to use the GNU C Library in non-free
programs enables many more people to use the whole GNU operating
system, as well as its variant, the GNU/Linux operating system.

This $&%$& demands everything to be labeled in a way which credits him
and he does not stop before making completely wrong statements like
"its variant". I find this completely unacceptable and can assure
everybody that I consider none of the code I contributed to glibc
(which is quite a lot) to be as part of the GNU project and so a major
part of what Stallman claims credit for is simply going away.

This part has a morale, too, and it is almost the same: don't trust
this person. Read the licenses carefully and rip out parts which give
Stallman any possibility to influence your future. Phrases like

[...] GNU Lesser General Public License as published by the Free
Software Foundation; either version 2.1 of the License, or (at your
option) any later version.

just invites him to screw you when it pleases him. Rip out the "any
later version" part and make your own decisions when to use a
different license since otherwise he can potentially do you or your
work harm.

[cb88]

I'm not sure what removal of sparc 32bit HOST support entails for TCG I would think it means TCG can no longer generate sparc instructions which would be rather bad for the cross platformness of TCG as a compiler backend then again its just a matter of adding it back,