pkgsrc-Users archive

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

Re: pkg_install & macosx 10.4/ppc failure

gcc 4.0.1 resolves this problem for me. Looks like people need xcode 2.5 on this platform
to have any expectation of success.

Thanks to to those who looked into the problem. I've added a note to bootstrap/README.MacOSX
indicating the requirement of xcode 2.5 when using Tiger.

On 19/04/2008, at 11:22 PM, Tyler Retzlaff wrote:


I don't know if anyone else is experiencing this but pkg_create fails for me with Bus Error
when creating packages on macosx 10.4/ppc always (not intermittently).

I've tracked it down to a strlen() being fed a NULL pointer. The code in question lives on
line 454 of pkgtools/pkg_install/files/lib/plist.c rev 1.17

               if (cmdp->c_argc == 0)
                       len += 1 + strlen(cmdp->c_s) + 1;
               else if (cmdp->c_subst && realprefix)
len += 1 + strlen(cmdp->c_s) + 1 + strlen(realprefix) + 1;
     ^^^^^^^^^^^^^^ <--- specifically this strlen(realprefix)
len += 1 + strlen(cmdp->c_s) + 1 + strlen(p- >name ? p->name : "") + 1;

Of course the keen eye has caught by now that this should not be reached given the
conditional block it lives in.

A few things to note.
- realprefix is definitely NULL; so says code inspection and so says gdb
 (gdb) print &realprefix
 Address requested for identifier "rprefix" which is in register $r25
 (gdb) info register r25
 r25            0x0      0
- cmdp is also totally bogus I cannot see any reason why this should be (the code seems to dtrt)
 also gdb says cmdv is sensible.
 (gdb) print &cmdp
 Address requested for identifier "cmdp" which is in register $r7
 (gdb) info register r7
 r7             0xffffffff       4294967295
- disabling all optimization makes the problem go away
 any level -O causes things to go wrong (for me)
 removing -O makes it all work.
- introducing printf()'s to print cmdp or realprefix makes the problem go away

[morbo:~/pkg_install_sucks] rtr% gcc -v
Reading specs from /usr/lib/gcc/powerpc-apple-darwin8/4.0.0/specs
Configured with: /private/var/tmp/gcc/gcc-4061.obj~8/src/configure -- disable-checking --prefix=/usr --mandir=/share/man --enable- languages=c,objc,c++,obj-c++ --program-transform-name=/^[cg][^+.-]*$/ s/$/-4.0/ --with-gxx-include-dir=/include/gcc/darwin/4.0/c++ -- build=powerpc-apple-darwin8 --host=powerpc-apple-darwin8 -- target=powerpc-apple-darwin8
Thread model: posix
gcc version 4.0.0 20041026 (Apple Computer, Inc. build 4061)

Anyone else having this problem? Should -O just be turned off for this platform when building libinstall?
Does anyone see why it really goes wrong with -O?


Home | Main Index | Thread Index | Old Index