Re: [PATCH][PR guile/17247] Block SIGCHLD while initializing Guile

[Doug Evans <xdje42@gmail.com>]

On Mon, Sep 1, 2014 at 5:48 AM, Gary Benson <gbenson@redhat.com> wrote:
> Doug Evans wrote:
>> On Sun, Aug 31, 2014 at 12:36 PM, Eli Zaretskii <eliz@gnu.org> wrote:
>> > > From: Doug Evans <xdje42@gmail.com>
>> > > Date: Sun, 31 Aug 2014 12:07:58 -0700
>> > >
>> > > Basically, current Guile (git) starts an internal thread (the
>> > > "finalizer" thread), and libgc as of 7.4 now starts several
>> > > marker threads by default (before 7.4.0 one needed to configure
>> > > libgc with --enable-parallel-mark).
>> > >
>> > > When other threads are running, and they haven't blocked
>> > > SIGCHLD, then the kernel may send SIGCHLD to these threads,
>> > > leaving gdb hung in the sigsuspend calls in linux-nat.c.
>> >
>> > A heretic thought: is it at all a good idea to have Guile (and GC)
>> > start threads when they run under GDB?  GDB is a single-threaded
>> > program, so having it linked against libraries that start threads
>> > whenever they like is IME a source of subtle problems (like this
>> > one) and a lot of pain down the road.  Anything GDB does that
>> > affects the global environment of the whole program (e.g., I/O
>> > redirection) will also affect those threads, with who knows what
>> > consequences.
>> >
>> > So maybe The Right Way of fixing these problems is configure Guile
>> > and GC so that they never start any additional threads?
>>
>> Users are going to want to start threads.
>> I've seen that already.
>> I think we should not shy away from them.
>
> This patch ensures the internal threads are created with SIGCHLD
> blocked.  Does something do this for other (user started?) threads?

What to do for user-started threads is an open question, both for
Python and Guile.

For reference sake, see PR 17314.
[Still working on a testcase.  What I added to the PR is just a
snippet to get thoughts down.]