@LeeTibbert I don't have a Linux env so I am not sure I can complete this. Any chance you could look into this?

@ekrich

Be glad to, but it would have to be the beginning of next week, and not Tuesday, April 1.
I hope a few day's delay does not break your momentum.

IMO, you are entirely right, any change or code not exercised or tested is, by definition, broken.

I'm trying to get a large & complicated 'Arrays` JDK9 PR off the shipping dock.

@LeeTibbert I was thinking that the smaller Linux sizes are big enough to handle 32 and 64 bit flags but wondering then what happens with struct alignment on macOS?

@ekrich

Sorry this fell off the bottom of my work list.

In the interest of not holding you up, I'll give the TL;DR and then
fall back to my usual careful development of a chain of reasoning & evidence.

TL;DR

Jumping to solution, this looks like a case whereOps extension methods could
be used to hide os differences. The hard problem there is letting people know
that such exist and that they need to import not onlytermios but alsotermiosOps.

We, as in you and I, dealt with such an situation with, IIRC, BSD sin6_len & sin_len.

The critical & discriminating pieces of information are:

1. What are the actual os sizes on (64 bit) macOS and Linux?

2. If the Linux os structure is smaller, what happens when it is passed more bytes
   than it expects? Does it ignore those bytes? complain? halt & catch fire?

If the linux systems we have at hand ignore those bytes, then we can do a
design sketch usingOps wizardry. This is a 'reduction in strength' fix
not 'drop-in & run'.

Discussion

Punting, for now, the "how do we inform people about Ops methods & get them to use them?" concern.

• I can, I think, write a small C program to determine the actual total size used by one each of a macOS and
  Linux system. That gives us some clues about padding.

• I can extend the study above to execute a termios function with the sizes determined above. A 'get'
  should be easy. I need to figure out a set of attributes which do not totally hose my terminal (or run
  that program in a 'disposable' window, not my main devo window.)

• If the probes above go the way I think/hope they do, that is Linux ignoring the space, we can start
  the tedious process of writing the Ops to deal with the various different offsets.

• I hate to say it, but the above assumes that the sizes & offsets will stay unchanged across time,
  operating system versions, operating systems, & architectures. At a rapid run, it looks like
  some pre-condition checking "guard" code should be added to the existingtermios.c

More as it develops.

FromTermiosTest
| assertTrue(true)

Neat trick to get a test to always pass, baring Exceptions. I'll have to remember that.

TIL: Today I learned.

How to bookmark an Issue or PR so I can easily find items on my currentguilt TODO list.
Obvious after the fact, but a discovery for me and, I hope, solution to a long standing desire.