|
From: Julian S. <js...@ac...> - 2011-01-18 12:37:51
|
IIRC there is some special-case kludgery with handling sys_readlink, to catch the case where the arg is /proc/self/exe. Maybe worth checking for that and disabling it? J On Monday, January 17, 2011, Nikolaus Rath wrote: > Julian Seward <js...@ac...> writes: > >> or are there any issues I should look out for? > > > > None that come obviously to mind. > > > > Send us the patch when you are done. > > I tried the attached patch against 3.6.0~svn20100724, but I seem to have > introduced a new problem: > > SYSCALL[3677,1]( 85) sys_readlink ( 0x82e3ca0(/usr/bin/python-dbg), > 0xbeb01cfa, 4096 ) --> [pre-success] Success(0x0:0xd) valgrind: > m_syswrap/syswrap-main.c:1473 (vgPlain_client_syscall): Assertion '0 == > (sci->flags & ~(SfPollAfter | SfYieldAfter | SfNoWriteResult))' failed. > ==3677== at 0x380269BF: report_and_quit (m_libcassert.c:191) > ==3677== by 0x38026C74: vgPlain_assert_fail (m_libcassert.c:265) > ==3677== by 0x380676FB: vgPlain_client_syscall (syswrap-main.c:1473) > ==3677== by 0x38063927: handle_syscall (scheduler.c:885) > ==3677== by 0x38065402: vgPlain_scheduler (scheduler.c:1081) > ==3677== by 0x38096ACA: run_a_thread_NORETURN (syswrap-linux.c:94) > > sched status: > running_tid=1 > > > Any idea what I did wrong? > > > > -Nikolaus |