>> + - Function calls, anonymous functions, and the `eval`, `.`, and
>> + `source` commands no longer propagate ERR_EXIT suppression.
>
> This kind of suggests that these constructs always propagated the
> suppression, which isn't the case, but the exact circumstances look too
> complex to explain. Maybe replace "no longer" with "now never".
What were the circumstances under which they previously propagated
suppression?
This is now explained in my new patch proposal (see new thread).
I admit that I don't fully understand what this commit did, so I
more or less copied the entry Bart added to ChangeLog:
* Philippe Altherr: 51071: Src/exec.c, Test/C03traps.ztst: fix
ERR_RETURN when a function using && / || is called within another
statement using && / ||
That's not accurate either, then?
That's just a bit less definitive/precise than your formulation. Enough so that it can be qualified as "not wrong" ;-)
My new patch proposal describes the exact circumstances, which unfortunately are rather convoluted.
Shouldn't this define the term "suppresses"? It's not used in
the documentation of ERR_EXIT.
Is the option suppressed for the function call or the result of the
anonymous function, or /within/ the called function or anonymous
function?
[...]
These are all valid questions. I think that they are all addressed by my two new patch proposals.
The documentation addresses some but not all of the special cases
and weirdness of ERR_EXIT. I figured that fixing that was out of
scope here, but this might actually be a good time to remedy the
omissions. That way NEWS and README don't have to waste a bunch
of space explaining aspects of ERR_EXIT that should be in the
documentation anyway.
Better documentation of ERR_EXIT makes it indeed easier to explain what changed. I took that approach in my two new patch proposals.
This is only incidentally true. The documentation omits many details
about ERR_EXIT's special cases, so your changes largely fall between
the lines (so to speak).
That was my point. The current documentation is so shallow that it fits both the state before and after my patches. So it could remain as is without being more or less wrong than before. But of course that doesn't avoid the need of documenting what changed because some users may have started to depend on the previous behavior. My updated patch for NEWS and README now describes in detail all the changes. And my patch for the ERR_EXIT and ERR_RETURN documentation adds enough details that it would no longer be compatible with the state before my patches.
Philippe