[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
progressions in the testsuite (all for -m32): -FAIL: gcc.c-torture/execute/nestfunc-3.c execution, -Os -FAIL: gfortran.dg/char_cshift_1.f90 -Os execution test -FAIL: gfortran.dg/char_cshift_2.f90 -Os execution test -FAIL: gfortran.dg/char_eoshift_1.f90 -Os execution test -FAIL: gfortran.dg/char_eoshift_2.f90 -Os execution test -FAIL: gfortran.dg/char_eoshift_3.f90 -Os execution test -FAIL: gfortran.dg/char_eoshift_4.f90 -Os execution test -FAIL: gfortran.dg/default_initialization_3.f90 -Os execution test -FAIL: gfortran.dg/namelist_28.f90 -Os execution test -FAIL: gfortran.dg/parent_result_ref_2.f90 -Os execution test -FAIL: gfortran.dg/ret_pointer_2.f90 -Os execution test regressions (for -m32): +FAIL: gcc.dg/torture/calleesave-sse.c -Os execution test No changes for -m64. -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
This bug was fixed in the package gcc-4.4 - 4.4.1-4ubuntu4 --- gcc-4.4 (4.4.1-4ubuntu4) karmic; urgency=low * Disable the build of neon optimized runtime libs on armel. * libjava: Use atomic builtins For Linux ARM/EABI, backported from the trunk. * Proposed patch to fix wrong-code on powerpc (Alan Modra). LP: #43. * Update to SVN 20090925 from the gcc-4_4-branch (r152174). - Fixes PR target/40473. -- Matthias Klose d...@ubuntu.com Thu, 24 Sep 2009 14:16:41 +0200 ** Changed in: gcc-4.4 (Ubuntu) Status: In Progress = Fix Released -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
Found the problem. It's nothing to do with _FORTIFY_SOURCE, and those intructions indexing off r2 are to do with -fstack-check so no problem there either. Fixed as follows. Incidentally this bug was triggered by fixing the obvious bug in no_global_regs_above Index: gcc/config/rs6000/rs6000.c === --- gcc/config/rs6000/rs6000.c (revision 152105) +++ gcc/config/rs6000/rs6000.c (working copy) @@ -16148,7 +16148,7 @@ rs6000_emit_prologue (void) int using_store_multiple; int using_static_chain_p = (cfun-static_chain_decl != NULL_TREE df_regs_ever_live_p (STATIC_CHAIN_REGNUM) - !call_used_regs[STATIC_CHAIN_REGNUM]); + call_used_regs[STATIC_CHAIN_REGNUM]); HOST_WIDE_INT sp_offset = 0; if (TARGET_FIX_AND_CONTINUE) -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
running a test build ** Changed in: gcc-4.4 (Ubuntu) Status: New = In Progress -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
please reopen with a testcase if this turns out to be a problem in GCC ** Changed in: gcc-4.4 (Ubuntu) Status: New = Invalid -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
Is the test case attached in comment #8 insufficient? According to the gcc documentation ( in particular, http://gcc.gnu.org/onlinedocs/gcc/Nested-Functions.html ), the code should be valid as long as the call to the nested function is executed before the outer function exits. I take it this functionality isn't actually used much, but to me anyway it seems like a bug.. (I should also confirm that mountall 0.1.7 fixes the issue for me, thanks.) -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
Re: [Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
On Thu, 2009-09-24 at 00:13 +, davidh wrote: Is the test case attached in comment #8 insufficient? According to the gcc documentation ( in particular, http://gcc.gnu.org/onlinedocs/gcc/Nested-Functions.html ), the code should be valid as long as the call to the nested function is executed before the outer function exits. I take it this functionality isn't actually used much, but to me anyway it seems like a bug.. (I should also confirm that mountall 0.1.7 fixes the issue for me, thanks.) Well, it did look like a gcc bug to me indeed. However, the test case here doesn't seem to break on fedora gcc 4.4.0 or jaunty 4.3.3 I'm still putting back that karmic machine into shape and will then be able to check what's up with the karmic gcc version. Cheers, Ben. -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
I also couldn't reproduce the problem here with any of the compilers I have lying around here. davidh, can you attach the .o for your testcase to this bug report? -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
.o as requested, produced by gcc -c -fPIE -Os test.c Linking and running: gcc test.o -o gcc-test ./gcc-test a in extern_func: 123456789 a in intern_func: -1074204592 ** Attachment added: gcc -c -fPIE -Os test.c http://launchpadlibrarian.net/32340419/test.o -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
A typical run of a slightly modified test case (extended printf) gives: a (@0xbfc22160) in extern_func: 123456789 a (@0xbfc22150) in intern_func: -1077796448 where the address is off by the same nicely even amount every run. -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
Huh. You have what looks like ppc64 code in there. 4c: 80 02 8f f8 lwz r0,-28680(r2) 50: 90 01 00 3c stw r0,60(r1) Where did that come from? It seems you have _FORTIFY_SOURCE defined for you too, somehow. Maybe that is pulling in a bad printf define? Please attach the .i generated by gcc -E -fPIE -Os test.c -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
Attached file generated by gcc -E -fPIE -Os test.c test.i Whatever the problem is, it should at the very least be shared by the system building the powerpc packages. Either that, or my test case is broken. :) ** Attachment added: test.i http://launchpadlibrarian.net/32342890/test.i -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
Re: [Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
On Thu, Sep 24, 2009 at 03:32:43AM -, Alan Modra wrote: Huh. You have what looks like ppc64 code in there. 4c: 80 02 8f f8 lwz r0,-28680(r2) 50: 90 01 00 3c stw r0,60(r1) Where did that come from? It seems you have _FORTIFY_SOURCE defined for you too, somehow. Maybe that is pulling in a bad printf define? Please attach the .i generated by gcc -E -fPIE -Os test.c -D_FORTIFY_SOURCE=2 is a default in Ubuntu's compiler. It can be turned off with -U_FORTIFY_SOURCE in the build if you want to see if that could be causing the glitch. https://wiki.ubuntu.com/CompilerFlags -- Kees Cook Ubuntu Security Team -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
** Changed in: gcc-4.4 (Ubuntu) Status: Invalid = New -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
Right, it's a gcc bug. As I'm sure you're away, the PowerPC port is not a first-class port at this point ** Changed in: mountall (Ubuntu) Status: New = Won't Fix -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
Scott, maybe if you use the default optimization settings, but if you do choose to use anything else than -g -O2, please care about the reports. ** Changed in: mountall (Ubuntu) Status: Won't Fix = Confirmed -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
Why? It's still clearly a gcc bug, no? ** Changed in: mountall (Ubuntu) Status: Confirmed = Won't Fix -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
This bug was fixed in the package mountall - 0.1.7 --- mountall (0.1.7) karmic; urgency=low * Build with -O2 on powerpc to work around wrong-code generation with -Os. LP: #43. -- Matthias Klose d...@ubuntu.com Tue, 22 Sep 2009 00:31:52 +0200 ** Changed in: mountall (Ubuntu) Status: Fix Committed = Fix Released -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
** Changed in: mountall (Ubuntu) Status: Won't Fix = Fix Committed -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
** Branch linked: lp:ubuntu/karmic/mountall -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
** Branch linked: lp:~ubuntu-core-dev/ubuntu/karmic/mountall/ubuntu -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
Hrm, somebody is taking chances here. Nested functions are evil and forbidden by ISO C standard. I don't know what kind of black magic gcc is supposed to use to be able to find the local variables of the declaration scope when the nested function is called via a function pointer from outside. I suppose it will have to create some kind of trampoline on the stack, which means you lose non-exec stack capability for your program. It may still be a bug that it doesn't work on ppc with some optimisations, but it's really not proper C in the first place. -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
As this really seems to be a problem with gcc, I'm reassigning this. ** Package changed: mountall (Ubuntu) = gcc-4.4 (Ubuntu) -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
** Description changed: Binary package hint: mountall My system (ibook running karmic) is not booting since the general breakage in the last couple of days. Installing the latest upgrades does not help. + + Edit: This seems to be due to a gcc bug, see comment #8 The boot seems to stop at mountall, whose output contains (perhaps unrelated) some garbled paths that I haven't been able to track down the source of. The garbled pattern is different between the boots, so it's obviously reading from some uninitialized memory somewhere. I attach a log of the output of mountall (obtained by inserting bash into /etc/init/mountall.conf and running mountall --debug .. as described in another bug), and my fstab. I just refuse to believe that I have to actually reinstall.. ;-) Installed versions of upstart and mountall: ii mountall 0.1.6 filesystem mounting tool ii upstart 0.6.3-3 event-based init daemon kernel is: Linux ibook 2.6.31-10-powerpc #34-Ubuntu Tue Sep 15 23:53:36 UTC 2009 ppc GNU/Linux fstab and mountall log attached. -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
As this really seems to be a problem with gcc, I'm reassigning this. and keep the broken mountall binary? If standard optimisation options (-O2), these should be used. ** Also affects: mountall (Ubuntu) Importance: Undecided Status: New -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
Don't use variable names starting with __ - those are reserved for the language implementation, per the C standard. -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 432222] Re: mountall fails, broken (powerpc?) gcc?
Thanks, I failed to figure out how to assign the bug to several packages. :-) Yes, I just confirmed that compiling mountall with -O2 fixes the issue here, or I should say, produces a working binary. So that would be a solution I guess, if it could be incorporated into the official package. The underlying issue still seems to be gcc though. Regarding the patch, to my defense I wasn't really expecting it to be applied officially anyway as it doesn't really seem like a proper fix to me. -- mountall fails, broken (powerpc?) gcc? https://bugs.launchpad.net/bugs/43 You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs