3.8.2010 11:51, Andy Green kirjoitti:
I'm uplevelling a Fedora ARM Cross gcc package that worked OK on gcc
4.1. However, there were a lot of changes needed to get it to work with
4.5.0.
The result seems good, it's able to build an arm kernel that works, but
there is one funny. When I build a 'hello world' test app (there's no
cross glibc yet but that shouldn't be the problem), I get this -->
$ armv5tel-redhat-linux-gnueabi-gcc hello.c
/usr/bin/armv5tel-redhat-linux-gnueabi-ld: skipping incompatible
/usr/lib/gcc/armv5tel-redhat-linux-gnueabi/4.5.0/libgcc.a when searching
for -lgcc
/usr/bin/armv5tel-redhat-linux-gnueabi-ld: cannot find -lgcc
/usr/bin/armv5tel-redhat-linux-gnueabi-ld: skipping incompatible
libgcc.a when searching for libgcc.a
/usr/bin/armv5tel-redhat-linux-gnueabi-ld: skipping incompatible
/usr/lib/gcc/armv5tel-redhat-linux-gnueabi/4.5.0/libgcc.a when searching
for libgcc.a
/usr/bin/armv5tel-redhat-linux-gnueabi-ld: cannot find libgcc.a
collect2: ld returned 1 exit status
The MeeGo/ARM crosstoolchain(s) made for MeeGo/i586 host has/have
the same problem/bug :
-------------- clip ----------------------------------
> But I start to build the package, I got a error meesage:
> /opt/cross/bin/armv7l-meego-linux-gnueabi-ld: skipping incompatible
> /opt/cross/libexec/../lib/gcc/armv7l-meego-linux-gnueabi/4.4.2/libgcc.a
> when searching for -lgcc
The 'libgcc.a' in the prebuilt toolchain really seems to be broken,
the first object in the archive shows :
[root@localhost 4.4.2]# file _udivsi3.o
_udivsi3.o: ELF 32-bit LSB relocatable, no machine, version 1 (SYSV),
not stripped
-------------- clip ----------------------------------
There are other really weird bugs which hint that some newbie made
the toolchain(s). The MeeGo 1.0/i586 system is a bleeding edge
glibc-2.11 based host so its binaries will work only on other
bleeding edge distros like the latest Fedora, SuSE, Ubuntu etc. Not
on my stable conservative development system, so I built my own
'armv7l-meego-linux-gnueabi' and 'armv5tel-meego-linux--gnueabi'
targeted crosstoolchains using the original MeeGo/ARM glibc-2.11's,
curses, X11-libs etc. as the target libraries. The binutils were
the 2.20.51.0.8 and GCCs were 4.4.4 and 4.5.0, yesterday I made
also gcc-4.5.1 for the armv7l variation. Nothing like your or the
MeeGo problem(s) were seen...
So my suspect is in some bug(s) in the bleeding edge experimental
distros. What is your cross host or the aimed "least common nominator"
runtime host you produced the binutils and GCC executables for?
