Re: [PATCH v2 00/16] Interrupting programs that block/ignore SIGINT

[Eldar Abusalimov via Gdb-patches <gdb-patches@sourceware.org>]

Hi Pedro,

Let me chime in to say that the issue addressed is indeed relevant, in
case you need any reassurance. :-)  This being addressed is great news!

A related problem[1] was reported in the CLion IDE issue tracker, and
the discussion also mentions the Kernel PR linked by Konstantin.  Since
CLion signals the inferior process directly, we ended up changing the
signal sent by the IDE from SIGINT to SIGSTOP so that there's no way the
inferior program can intercept the signal.  Apparently, this is also how
LLDB handles that as well.

However, that didn't cover[2] the target-async case, that is, when the
signal is still being sent by GDB (using the "interrupt" command), and
remote targets, where it's the gdbserver who ultimately supervises the
inferior process, not GDB.

From the patchset, I gather that the target-async case has now been
fixed as well, but what about remote targets?  Do you think it makes
sense to also change gdbserver to use SIGSTOP instead of SIGINT?  As far
as I understand, unlike for a locally spawned process, this shouldn't
require waltzing with pseudo-terminals for the inferior on the remote
gdbserver side.  At the same time consistency w.r.t. signal handling on
suspending the target would definitely be a nice thing to have
eventually.

  [1]: https://youtrack.jetbrains.com/issue/CPP-14435#focus=Comments-27-3105589.0-0
  [2]: https://youtrack.jetbrains.com/issue/CPP-21975#focus=Comments-27-4635249.0-0


--
Regards,
Eldar