Bug #10595: interrupts not handled while finalizers running

• Status: Closed
• Assignee: -
• Target version: 2.2.0
• ruby -v: trunk
• Backport: 2.0.0: UNKNOWN, 2.1: UNKNOWN

[normalperson (Eric Wong)]

Originally noted in [ruby-core:66635]

Trying to Ctrl-C something like the following loop is not always successful:

f = proc { 1000.times {} }
loop do
  ObjectSpace.define_finalizer(Object.new, f)
end

The Interrupt is created and raised, but lost during the postponed
job processing (rb_postponed_job_flush).

Trying to mask out all interrupt flags (instead of just
th->interrupt_mask |= POSTPONED_JOB_INTERRUPT_MASK) allows the
interrupt to be handled, but this is not suitable for expensive
finalizers.

So I haven't figured out how to fix this, yet... Help appreciated, thanks.

[nobu (Nobuyoshi Nakada)]

• Status changed from Open to Closed
• % Done changed from 0 to 100

Applied in changeset r48829.

---

vm_trace.c: defer interrupts while postponed jobs

• vm_trace.c (rb_postponed_job_flush): mask signal trap interrupt
  too to defer handling after finalizers finished.
  [ruby-core:66825] [Bug #10595]

[normalperson (Eric Wong)]

[e-mail] wrote:

vm_trace.c: defer interrupts while postponed jobs

• vm_trace.c (rb_postponed_job_flush): mask signal trap interrupt
  too to defer handling after finalizers finished.
  [ruby-core:66825] [Bug #10595]

Sorry, I tried something like this but it's not effective if the
finalizer has an expensive (or infinite) loop.

I suppose it's better than nothing and finalizers should not
be expensive...

[nobu (Nobuyoshi Nakada)]

Yes, but interruptions of finalizers could cause serious undesirable results.