[Bug gdb/20485] New: sigprocmask in tracee affects tracer as well

["jengelh at inai dot de" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=20485

            Bug ID: 20485
           Summary: sigprocmask in tracee affects tracer as well
           Product: gdb
           Version: 7.11.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: gdb
          Assignee: unassigned at sourceware dot org
          Reporter: jengelh at inai dot de
  Target Milestone: ---

Given this program:

#include <stdio.h>
#include <signal.h>
int main(void)
{
        sigset_t m;
        sigemptyset(&m);
        sigaddset(&m, SIGINT);
        sigprocmask(SIG_BLOCK, &m, NULL);
        while (1)
                ;
        return 0;
}

Platform: openSUSE Leap 42.1 amd64, gdb-7.11.1; also CentOS 7 amd64, gdb-7.6.1.

When started from and run under gdb, gdb itself is unresponsive to SIGINT via
Ctrl-C. Furthermore, when the program is started from a shell and separately
attached to with gdb, gdb itself, after issuing "continue", is also
unresponsive to SIGINT delivered via tty/Ctrl-C and via kill(1).

What I believe should have happened: gdb should have stopped the program and
return me to a (gdb) prompt.

An strace on gdb (4446) reveals that it forwards SIGINT to the tracee (3428)
for some reason:

[pid  4446] 23:18:21 poll([{fd=3, events=POLLIN}, {fd=7, events=POLLIN},
{fd=10, events=POLLIN}], 3, 4294967295 [wait]
<SIGINT sent to 4446>
) = ? ERESTART_RESTARTBLOCK (Interrupted by signal)
[pid  4446] 23:18:28 --- SIGINT {si_signo=SIGINT, si_code=SI_USER, si_pid=4780,
si_uid=0} ---
[pid  4446] 23:18:28 kill(3428, SIGINT) = 0
[pid  4446] 23:18:28 rt_sigreturn({mask=[]}) = -1 EINTR (Interrupted system
call)
[pid  4446] 23:18:28 rt_sigprocmask(SIG_BLOCK, NULL, [], 8) = 0
[pid  4446] 23:18:28 poll([{fd=3, events=POLLIN}, {fd=7, events=POLLIN},
{fd=10, events=POLLIN}], 3, 0) = 0 (Timeout)
[pid  4446] 23:18:28 poll([{fd=3, events=POLLIN}, {fd=7, events=POLLIN},
{fd=10, events=POLLIN}], 3, 4294967295<wait>

But it should not forward it! When asking gdb for the signal configuration:

(gdb) handle SIGINT
SIGINT is used by the debugger.
Are you sure you want to change it? (y or n) y
Signal        Stop      Print   Pass to program Description
SIGINT        Yes       Yes     No              Interrupt

It says nopass right there.

-- 
You are receiving this mail because:
You are on the CC list for the bug.