Subject: Re: [ub] Draft 1 of Stackable, Thread Local, Signal Guards
From: Jens Maurer (Jens.Maurer_at_[hidden])
Date: 2019-05-31 13:55:12
On 31/05/2019 15.35, Niall Douglas wrote:
> On 30/05/2019 19:09, Jens Maurer wrote:
>> What's the interaction with the existing library facility
> Many thanks for the useful question.
> I have added to the paper a FAQ item answering this. It is as follows:
I didn't mean: how do I impplement the new facility, I'm asking about
(possibly existing) programs that currently use <csignal> in at
least some part, and another part now starts to use your proposed
new facility, which seems to offer notification of a similar set of
exceptional circumstances (e.g. SIGFPE).
What does/should the specification say is supposed to happen?
(I don't care about implementation for now.)
> 6.2 What is the interaction with the existing library facility <csignal>?
> On POSIX only, signal guard could be implemented using <csignal>, apart
> from the additional signals described above, which are implemented by
> POSIX in any case. It is highly unlikely, however, that anyone would
> actually do so when POSIXâs sigaction() is far superior to signal().
> On non-POSIX, I would find it extremely unlikely that anybody would use
> <csignal> to implement this facility as, in this authorâs
> experience,<csignal> implementations have a very low quality of
> implementation on non-POSIX platforms. To my knowledge, every non-toy
> non-POSIX system has a proprietary mechanism by which this facility
> could be completely implemented to a high degree of quality.
> ub mailing list
SG12 list run by herb.sutter at gmail.com