The documentation of how to build cross-compilers is inadequate, and
is split into two parts (current installation manual doc/install.texi,
and parts in doc/install-old.texi that have not been merged into the
main installation manual) plus an unofficial FAQ.
I have many times requested that people deal with this; no-one has
done so. List of requests in:
3.1 20020107 (experimental)
System: Linux digraph 2.2.20 #2 Sat Nov 10 16:44:22 UTC 2001 i686 unknown
configured with: ../gcc-cvs/configure --prefix=/opt/gcc/mainline --disable-shared --enable-threads=posix --with-system-zlib
Attempt to build cross-compilers using just the current documentation,
without the old install manual or CrossGCC FAQ, especially in more
complicated cases such as:
* Canadian crosses.
* Crossing into a glibc-based target, without pre-existing GCC or
glibc libraries or headers for that target (where glibc and libgcc
* Not using a unified tree.
Document cross-compilers properly.
State-Changed-Why: Confirmed as a bug, marking priority "high" at request of
On the theory that a working build script is a good adjunct to
the documentation, I am assembling a script that can build and
test gcc/glibc/linux cross toolchains from scratch for all architectures
supported by glibc. I am initially focusing on building from
release tarballs in a nonunified tree. I am also documenting
how to set up dejagnu to do remote testing of gcc/glibc.
Eventually I may add gdb building/testing to the script.
The script and documentation is online at http://kegel.com/crosstool
while I work on it.
http://kegel.com/crosstool has come a ways since I last posted.
It's now quite easy to build and test crosscompilers.
My script still doesn't run the glibc regression test, but at
least it runs the gcc/g++/stdlibc++ tests. Anyone who's having
trouble building or testing crosscompilers should have a look
at that page. In fact, it'd be nice if http://gcc.gnu.org/install/
or a related page linked to http://kegel.com/crosstool.