Op 26-03-18 om 19:37 schreef Martijn Dekker:
> Op 26-03-18 om 19:00 schreef Stephane Chazelas:
>> Thanks. See also 42446
>> (https://www.zsh.org/mla/workers/2018/msg00252.html)
>> with other suggested options.
> 
> Ah yes, sorry, forgot you'd already brought it up here.

Re-reading that, my thought is: while option 2 might be nice to have,
option 4 is the simple, obvious and immediate fix, so that's the one I'm
capable of providing -- especially with a release imminent.

If you'd like to have a go at implementing option 2, so much the better.

>> There I mentioned the potential need to block signals between
>> the time the umask is changed temporarily and when it's restored
>> (to avoid traps running with the wrong umask (0177 instead of
>> the user's requested one)).
> 
> Good point.
> 
> Please consider my patch withdrawn.

Today was apparently not my best day. Since that function was already
blocking (a.k.a. queueing) signals to do its thing, of course my patch
was trivial to fix. Take 2.

- M.