|
From: Konstantin S. <kon...@gm...> - 2009-10-29 09:19:33
|
This affects other systems calls, not just epoll_wait. I submitted https://bugs.kde.org/show_bug.cgi?id=212267 with an example for read() system call. I'd appreciate a comment. Thanks, --kcc On Wed, Oct 28, 2009 at 7:54 AM, Konstantin Serebryany < kon...@gm...> wrote: > Any suggestion? > > > On Thu, Oct 22, 2009 at 10:36 AM, Konstantin Serebryany < > kon...@gm...> wrote: > >> -valgrind-users >> +valgrind-developers >> >> I observe a situation where the number of invocations >> of PRE(sys_epoll_wait) is greater than the number of invocations >> of POST(sys_epoll_wait). >> Is that expected? >> >> This is causing memcheck to think that memory passed to epoll_wait() as a >> second parameter is left uninitialized... >> >> Thanks, >> >> --kcc >> >> >> >> >> >> On Tue, Oct 20, 2009 at 3:30 PM, Konstantin Serebryany < >> kon...@gm...> wrote: >> >>> Hi, >>> >>> I am investigating a memcheck's report near a call to epoll_wait(). >>> I am running my program (sorry, not small test case) with >>> --trace-syscalls=yes. >>> >>> Usually I get this: >>> SYSCALL[29628,125](232) sys_epoll_wait ( 62, 0x1540ca30, 1024, 1000 ) --> >>> [async] ... >>> SYSCALL[29628,125](232) ... [async] --> Success(0x0:0x0) >>> I assume these two lines come from PRE(sys_epoll_wait) and >>> POST(sys_epoll_wait). >>> >>> But sometimes I get this: >>> SYSCALL[29628,156](232) sys_epoll_wait ( 96, 0x15948a30, 1024, 417 ) --> >>> [async] ... >>> SYSCALL[29628,156]( 15) sys_rt_sigreturn ( ) --> [pre-success] >>> NoWriteResult >>> >>> So, POST(sys_epoll_wait) does not get called and memcheck thinks that the >>> second parameter of epoll_wait is uninitialized. >>> >>> What does this sys_rt_sigreturn mean? Why POST(sys_epoll_wait) is not >>> called? >>> Any idea? >>> >>> Thanks, >>> >>> --kcc >>> >>> >>> >>> >>> >> > |