On Dec 14, 2012, at 4:31 PM, "Eric P. Mangold" firstname.lastname@example.org wrote:
On Fri, Dec 14, 2012 at 02:04:40PM -0800, Glyph wrote:
On Dec 14, 2012, at 9:11 AM, Eric P. Mangold email@example.com wrote:
Awesome work, Thomas.
I have an issue with this pre-release on OpenBSD. setup.py aborts trying to compile sendmsg.c
This is a known bug that has a patch, and is presently awaiting a review. http://twistedmatrix.com/trac/ticket/5907
I would like to request a new pre-release with this fix once it has been reviewed and merged. (someone? pretty please?)
We generally don't issue new pre-releases for issues that are not regressions. Thomas is the release manager in this case so it is his call, but it would be nice to get the 12.* releases all out within 2012 :). I think it'd be a lot better to focus on this for 13.0, and try to push for 13.0 to be released sooner rather than 12.3 to be released later.
Not only is this ticket not a regression, and is OpenBSD not a supported platform for Twisted, but it looks like we don't even have an un-supported builder for that platform, so other developers would have a hard time reviewing the ticket without setting up their own personal virtualization infrastructure. In the meanwhile you might want to consider provisioning and contributing an OpenBSD buildbot, so we can start the process of transitioning it from unsupported to supported.
Technically, it's only NOT a regression BECAUSE OpenBSD is not a supported platform (yet). It's clearly a regression of functionality, because releases of Twisted prior to sendmsg.c did not have any issue building on OpenBSD (afaik). But I agree with the definition of regression being constrained to behaviors on supported platforms only...
OTOH - the bug is trivial, and the fix is also trivial, and has been verified by two people (the original patch author, and myself). Yes it still needs to pass review, but *possibly* this will not delay the release much... as long as it's out by Christmas, that is soon enough, right? =)
Well, okay, but you'll have to convince Thomas, not me ;).
In my mind, it's fairly embarrassing that our setup.py is borked on the 2nd most popular BSD, and so I would still like to ask Thomas if he wouldn't mind trying to fit this in. But as you say - it is his call :)
Personally I'm not embarrassed. In fact I think it's great! If people rely on Twisted to work on their platform, they need to be motivated to contribute the resources to keep that platform working - and that's both in terms of buildbots and in terms of review time. I would prefer, in fact, that "your platform is not necessarily going to work right" failures be more loud and immediate.
Also - the time to get stuff into a release is at the start of its cycle, not the end. Even if this particular change does get in, I'm glad we've had this discussion to remind everyone of that fact.
As far as getting your ticket reviewed, The best way to get your tickets reviewed is to do code reviews for other people, so that when a reviewer with some free time comes around, you'll be first in the queue. Unfortunately it looks like your ticket is dead last at the moment, so clearing out any other reviews would be extra helpful in its case!
Since clearing the _whole_ queue is probably an impossible amount of effort, personally I'd be happy to offer you the same deal that Martin van L?wis pioneered on the Python core mailing list: if you do five code reviews, I'll review a ticket of your choice, virtualization infrastructure hassles notwithstanding :).
It's a fair deal - but with things as they are, I barely have time to do even the small bits of work that I do. When it comes to reviews, I'm afraid I'll probably always be a beggar :/
Doing one review a week is a pretty small commitment (speaking here as the pot).
Oh, and uh technically, since I have not changed the original author's patch (hence, I am not the original author of the code) I can perform the review on this patch myself? I believe I have done this before, and unless there is some objection, I guess I can go ahead and do that...
Yes, you can do the review. Don't half-ass it though :-).
You should work on this soon regardless of whether it gets into 12.3 or 13.0 though. It seems like the current release is the most important when stuff like this happens, but time marches inexorably forward, and someday soon 12.3 will be old news, and you'll wish you got it in time for 13.0 instead.
Oh and yes - I AM going to see about getting some OpenBSD buildslaves, whether via Snakebite farm, or some other route.