Re: [perl #93428] potential race condition when mixing signals and select()

On Sun, Dec 11, 2011 at 11:11 PM, Eric Brine <ikegami@adaelis.com> wrote:
> Perl signal handling uses an approach that suffers from a race condition.
> It's buggy.

That's no news. Are you suggestion a solution?

> As for being a "fundamentally difficulty", just because race conditons can
> be tricky to avoid when using threads or signals doesn't mean we shouldn't
> fix them.

Life is much easier when you know exactly what the user wants, but
core can't make any assumptions, that is the problem. Core doesn't
know which signal should be blocked when, only the user does. There is
no magical solution, best we can do is doing exactly what the user
asks us to do, and assume he knows what he's doing.

> And it's neither tricky nor difficult to solve on systems with
> pselect.

Even if it was portable, it's not going to solve the problem by itself.

Leon

• [perl #93428] potential race condition when mixing signals and select() by Leon Timmermans via RT
• Re: [perl #93428] potential race condition when mixing signals andselect() by Paul LeoNerd Evans
• Re: [perl #93428] potential race condition when mixing signals and select() by Eric Brine
• Re: [perl #93428] potential race condition when mixing signals and select() by Leon Timmermans
• Re: [perl #93428] potential race condition when mixing signals and select() by Leon Timmermans
• Re: [perl #93428] potential race condition when mixing signals and select() by Eric Brine
• Re: [perl #93428] potential race condition when mixing signals and select() by Eric Brine
• [perl #93428] potential race condition when mixing signals and select() by Father Chrysostomos via RT