PHP 7.2 Release

[tpunt]

Ping @sirsnyder

I believe Joe is assuming we will make the next release (judging by his comment here).

Before we make the next release, what needs to be done on Sockets to stabilise it a bit more? Should we either revert the recent changes and place them into their own branch, or do they just require a bit more work? I think this is the primary blocking issue for a stable release.

Hopefully, if we can sort out the sockets stuff over the coming days, then we can make a new release over the weekend.

[tpunt]

Also ping @dktapps, as someone who has been working on the sockets stuff recently.

[sirsnyder]

Good question. We should keep them, cause I think Sockets are a high anticipated feature.

In case of the primary blocking issue, I could today and tomorrow compare all functions of our current socket implementation with the php socket behavior and adapt it. Of course if possible with your support and/or @dktapps. I would not add #744 yet, but #804 and also I would be very much in favor to add #778. Volatile objects are error prone in complex object structures. PR #778 fixes the well-known Exception "pthreads detected an attempt to connect to an object which has already been destroyed".

[tpunt]

Ok, if you could work on the sockets stuff today/tomorrow, then that would be great. I should be able to find some time to review #778, but I won't be able to help with the sockets stuff as well (hopefully @dktapps can help you with that instead).

[dktapps]

I've been trying to get my guinea pigs to test #778 recently due to a wide range of bug reports I keep getting relating to multidimensional arrays segfault issues and such. I'll see what I can do.

I can assist with socket-related stuff if needed. The primary showstopper related specifically to sockets at the moment is (in my opinion) issues with throwing exceptions in non-blocking mode (such as #809 ). I don't think there are any stability issues wrt Sockets right now apart from odd bugs - the rest is just missing features.

[dktapps]

There are also a bunch of cases of misbehaving Socket tests that may need investigation.

[sirsnyder]

@ which socket tests do you mean? Feel free to correct them ;-)

[dktapps]

@sirsnyder I keep seeing the socket-bind test failing for no apparent reason. It works fine when testing locally so maybe some Travis/AppVeyor resources issue, I don't know.

[sirsnyder]

@dktapps very strange the behavior of this one test on x64 arch.
@tpunt the sockets stuff has been merged into master.

[tpunt]

@sirsnyder Thanks. Assuming nothing else comes up, I'll make a new release over this weekend :)

[dktapps]

@sirsnyder I wouldn't swear to this but it may be some issue with AppVeyor, because I've seen it randomly fail on x86 and x64 both. I can't reproduce it when running tests locally at all, on either arch.

[dktapps]

FWIW I'm not sure what versioning system pthreads follows, but I think sockets (as addition of features) warrant a minor version bump to 3.2.0.

(also, 3.1.7 is ancient now, running for 2 years)

[tpunt]

@dktapps Yep, that's exactly my thinking too. It will be a 3.2.0 release.

[sirsnyder]

@dktapps will have a look next days or week, it's low prio for me.
@tpunt nice to read :) At least it should be a 3.2.0 release. Maybe even a 4.0.0 release, because there is a BC break with 465d3bc and 1fcafd7

[dktapps]

wrt. socket-bind.phpt, there is a random element in the test which I didn't initially notice, which might explain the inconsistency of the test failures. I'll take a look over the weekend and see if I can get it to reproduce locally.