Re: Opportunistic GC

[Philipp Stephani]

Am Mo., 8. März 2021 um 17:18 Uhr schrieb Andrea Corallo <akrl@sdf.org>:
>
> Philipp Stephani <p.stephani2@gmail.com> writes:
>
> [...]
>
> >> I also expect fork will work, I'm more wondering about interaction with
> >> sockets file descriptors or anything else in our codebase and/or in
> >> libraries we have loaded.
> >>
> >> I've read in the past this is an issue in practice in non trivial code
> >> and that, because the typical use of fork () nowadays is just to run
> >> afterward exec (), operating in the grey between the two was not really
> >> a good idea.  But I don't have the reference now and if you say it works
> >> I've no reasons not to trust you.
> >
> > The restriction is that you can only call async-signal-safe functions
> > in subprocesses before exec. See e.g. the man page for fork on
> > GNU/Linux.
>
> Can we enforce that for our case?
>

It would be useful to have compiler attributes to mark
async-signal-safe functions and other attributes that enforce
async-signal-safely within a region, but I'm not aware of the
existence of such attributes.