bug#61504: 29.0.60; executing byte-code from previous build causes SIGSEGV crash

[Gregory Heytings]

> > > By the way, doesn't the patch switch the restoration order of 
> > > narrowing and restriction, respectively? Maybe it doesn't matter?
> >
> > Hmmm, that's a good question!  The evaluation order of parameters is 
> > unspecified in C, so actually the order could be switched or not, 
> > depending on what the compiler chooses to do.
>
> Yes, the saving order is undefined but the restoring order seems 
> well-defined. It currently restores narrowing locks first, then the 
> restriction, but your patch flips the order.

Indeed, I misunderstood what you said above, now I got it!

> Please at least make the saving order well-defined, preferably in the 
> reverse order of restoration for symmetry.

I did that.

Thanks again for your review/feedback.