memcheck doesn't recognize the fact that a ptrace call with PRACE_GETREGSET has written into the given target memory buffer. Thus valgrind reports false positives for gdb, e.g. when using gdb's inferior function calls. Reproducible: Always
Created attachment 74759 [details] Proposed regset support patch Proposed patch against mainline svn (r13080). The patch only fixes s390x (verified).
Looks sane. Just in case, do you have a simple testcase other than gdb?
Created attachment 74793 [details] Test case for PTRACE_GETREGSET Here you go. Notes: - The program should be platform-independent, although only tested on s390x. - The constant PTRACE_GETREGSET is hard-coded, to be independent of any system header files. - With the fix, memcheck shouldn't yield any complaints. - Only 'get' (not 'set') is tested.
Created attachment 75094 [details] Updated test case without "random" output
thanks. added the testcase (+minor fix that initializes c)and the s390 specific code
also added x86 (tested) amd64 (untested) and arm(untested) handlers to ptrace. ppc and mips dont seem to have ptrace support. Julian, I would consider this fixed but I cant change the state of bugzillas
Tested it on amd64. Works.
Reopening because the test fails on s390 z10-ec. florian@l005036:~/valgrind-patched/memcheck/tests/linux> ./getregset ptrace getregset: Input/output error florian@l005036:~/valgrind-patched/memcheck/tests/linux> ../../../vg-in-place ./getregset ==9584== Memcheck, a memory error detector ==9584== Copyright (C) 2002-2012, and GNU GPL'd, by Julian Seward et al. ==9584== Using Valgrind-3.9.0.SVN and LibVEX; rerun with -h for copyright info ==9584== Command: ./getregset ==9584== ptrace getregset: Input/output error uname -a Linux l005036.zseriespenguins.ihost.com 2.6.9-42.EL #1 SMP Wed Jul 12 23:21:43 EDT 2006 s390x s390x s390x GNU/Linux
On 09/11/12 05:22, Florian Krohm wrote: > https://bugs.kde.org/show_bug.cgi?id=308886 > > Florian Krohm <britzel@acm.org> changed: > > What |Removed |Added > ---------------------------------------------------------------------------- > Status|RESOLVED |REOPENED > Resolution|FIXED |--- > Ever confirmed|0 |1 > > --- Comment #8 from Florian Krohm <britzel@acm.org> --- > Reopening because the test fails on s390 z10-ec. Grrr, the old RHEL4 is really a PITA....but in fact, its actually a very good sanity check. Fixed.
All right then. Fixed for good. Thanks.
Guessing that this was fixed with r13112, r13115.
FYI. The set of revisions that I backported to the fedora package for this was: r13110 r13111 r13112 r13113 r13115
yes fixed by the given revisions . (I dont have the access rights to set bugzilla state).