[kaffe] Cygwin Build Error--kaffe 1.1.5--/kaffevm/jit/methodcalls.c:162
robilad at kaffe.org
Wed Sep 14 07:01:50 PDT 2005
4C/Gilbert T. Williams wrote:
> Someone who has built kaffe 1.1.5 on cygwin may be able to help me. TIA!
I've just gone through making the CVS head build and work (mostly) on
> I am evaluating Kaffe for use on an embedded platform. As a starting
> point, I would like to get a clean build from source
> on my cygwin installation using:
> 1. Windows XP (Home Edition Service Pack 2)
> 2. CYGWIN_NT-5.1
> 3. gcc 3.4.4 (cygming special)
> 4. jikes 1.22 – built using the gcc 3.4.4, so it appears that the
> overall build environment is OK.
You may need to apply a patch or two to jikes to have it actually work:
I had to, at least. ;)
> 5. gnu ld 2.16.91
> ./configure --with-staticlib --with-staticbin
what I use is:
../kaffe/configure CC="ccache gcc" CXX="ccache g++" --with-gc=kaffe-gc
--disable-native-awt --with-staticvm --with-staticbin
--with-threads=unix-jthreads --enable-pure-java-math --prefix=/tmp/kaffe
the ccache bits speed compilation on Cygwin considerably if you are
hacking on C & C++ code or have to wait for ages for configure. It still
takes hours on my 600MHz Win98 notebook, but hey ... ;)
--with-gc=kaffe-gc sets the gc to kafe'S gc. it's the default, anyway,
but I've had configure scripts to stupid things on cygwin, so I set it
manually to work around whatever made me set it manually. You may not
need to set it explicitely.
--disable-native-awt disables the AWT build. You're probably not
interested in building graphical applications for Cygwin/XFree86 and
enabling it means dealing with X11 headers, Gtk+, ... which I didnt have
time for. Bug reports and in particular fixes and screenshots
appreciated, if you are interested in that. But then, I'd recommend
trying to get Kaffe built with the Qt4 peers. ;)
--with-staticvm --with-staticbin was the combination that works for me,
without linking problems on CVS head.
--with-threads=unix-jthreads sets the threading model to jthreads, since
the default pthreads seem to spend a few seconds getting stuck somewhere
deep in windows/cygwin libraries, and eventually produce a warning
message from Cygwin that confuses the test scripts.
--enable-pure-java-math If you are not using crypto and don't have gmp
--prefix=/tmp/kaffe That's just me setting the prefix for the resulting
binary to keep different versions apart.
With this setup, and after a round or two of make && make check, I get
63 passes out of 148 regression tests with jit3. There probably are
various issues that the compiler warnings should be able to help you
fix, I'd appreciate patches.
More information about the kaffe