On Wed, Nov 06, 2013 at 10:15:14AM +0100, "Jörg F. Wittenberger" wrote:
When the issue was introduced into chicken, I forked my own environment
and just tried following the text books. Safe for some dead code I never
had to deal with these problems again. (Let me repeat: the limited queue
of pending interrupts was a problem in practice for me. All I have to do
is pull the network plug when running certain code to make it visible.)
As far as I know, this issue was never "introduced" to CHICKEN, but was
always there. In earlier versions, only one signal could be handled at
a time. Felix added the backlog of pending signals in order to avoid
dropping (too many) signals while handling other signals.
I think it's important to fix, and I agree that your solution is
probably the correct one, but I'm afraid it will be even more code, not
less. Anyway, I'm up for fixing this, but I want to get 4.9.0 released
first before making any more large code overhauls like you describe.
Our main problem with this interrupt mess is that it's very hard to test
correctly, especially in the presence of threading, so if you have a
clean stand-alone test case which exposes this problem and can be
integrated in the test-suite, that would be *very* helpful already!
I'd love to help bringing these fixes into the mainline chicken. (After
all it's a horrible burden to always maintain the diff). @Peter: I'm
still waiting for your reply on that topic.
Last I remember I sent you a reply explaining exactly what I explained
above: we need to get 4.9.0 released before making any more large
changes. I just checked, that mail dates from Tue, 15 Oct 2013 and has
this literal quote in response to your question about sending a patch:
``If it's a clean and well-explained patch which doesn't break existing
code I'm sure that will be accepted. (again, only AFTER 4.9.0 is out)''
This is still 100% accurate.