Saturday, August 3, 2013

17.0.8 available (plus: goodbye, OS/2)

Current CD: REO Speedwagon, You can Tune a piano, but you can't Tuna fish. "Roll With The Changes" seems like a particularly apropos track these days.

17.0.8 is available for testing from the Downloads tab (release notes). There are no TenFourFox-specific fixes in this release and the changesets are the same as 17.0.7's and apply without modification, so use those for building. Assuming no major issues, it will convert to release on Monday evening Pacific time as usual.

There are only two more scheduled releases for ESR17 left to occur: 17.0.9, to coincide with 24.0, and 17.0.10, to coincide with 25.0 (and ESR24's 24.0.1). At that point, only a chemspill would force a 17.0.11, and then only up until 26.0/24.0.2, when ESR24 becomes the only maintained ESR branch and Mozilla upstream support for ESR17 is terminated. 17 has been a strong, solid stable branch and I'll be sad to see it go. I don't yet know what this will mean for Tenfourbird.

I'm still not able to get simple scripts to run yet in BaselineCompiler under (what is now) 24-beta; I think I'm accidentally pushing wrong or bogus values on the stack, and the JavaScript VM routines are barfing on them. (When I said everything is on the stack, I mean everything.) Some inquiries of whomever was around on #jsapi were sympathetic but ultimately unhelpful, so I'm just going to have to keep plugging at it. I'll be uploading an initial set of changesets soon so those of you with some ideas can have a look, and there is also some activity from someone at IBM about getting PowerPC better supported by Firefox, so I'm hoping they'll have some thoughts on it even though OS X PowerOpen ABI is a bit different from Linux SysV ABI. There's even a Firefox build waterfall for Linux ppc64, which should help to flag any big-endian regressions from future code earlier.

I'm still hoping I can get a 24 beta ready to coincide with 24.0, which is six weeks away. If I can do that, and it's not a train wreck, then we switch people to 24 with 24.0.1 and 17.0.10. If I can't, we issue a 22 beta and use it as a transitional release for stable branch users until I get 24 working, or if I can't do so in a timely manner (two or three version cycles), we give up and drop source parity. We will not make a release without a JIT; we rely heavily on it and the interpreter has been deoptimized in 24 and up to simplify the code, making it a very poor alternative option. (I shudder to think what will happen to PPC Linux Firefox users.) If there is no JavaScript JIT, source parity ends, period, full stop. Whatever happens, 17.0.10 is our final 17.x release.

With respect to 22, there is no draft of the security updates for 23 yet for me to read through, so I will probably not be able to backport fixes until later this coming week. At that point a 22.0.1 will be released (not 22.1); this will be the basis for the "stable branch 22" 22.0.2, if required, and will contain the aforementioned fixes for issues 231 (for 10.4 and 10.5), 232, 233 and 234. None of these issues really affect 17, so they will not be backported. Expect this build to be available probably Friday or Saturday. Remember, 23 has no more methodjit, so there is no point in trying to port it separately of 24.

In other news, Mozilla is discussing removing OS/2 support from the build system due to incompatible changes (note that these changes are, mercifully, unlikely to affect us directly). This is sad but not altogether unexpected, since not even ESR17 works right on OS/2; a porting effort is underway but the chance of it becoming fully operational before ESR17 support ends seems very remote. Another venerable tier-3 build bites the dust.

Also, I finally managed to get a pre-order in for a FirefoxOS phone. Geeksphone, realizing they have a lot of pent-up demand which was not met by the five or six devices they built for FirefoxOS developers (he says with irritation), has now started taking pre-orders for the Peak Plus, a GSM dual-core Snapdragon device with decent specs that hits most of what I need in a handset. In this era of crony capitalism surveillance, we need a third way besides Google and Apple (don't say Microsoft; they were even deeper in it than that lot) for the devices that are intertwined with most everything we do, and Mozilla is large enough and trustworthy enough to be able to pose a plausible alternative. As a side benefit, its open architecture means it's likely to work a lot better with our older Macs. It lists for €149, and with tax and shipping to the USA it ended up being around US$240, which is staggeringly reasonable for an unlocked device even considering the penalty of moving to a new mobile operating system. Look for a review in this blog when I get it in September and a comparison with my Galaxy Nexus handset (bought new and unlocked for $650), which just got the "nothing" 130MB upgrade to Android 4.3 in which nothing changed but the camera app. Well, that's modern life for you.


  1. Lol, if you think Microsoft's worse than Google you should check this out:

    Google's the worst of the lot in my opinion, I don't think Microsoft's that bad

    Many people still think they are that "evil" company from the '90s (Seems like you do too, from that post). Times have changed, move on.

    1. Sorry, forgot to add that this post isn't just directed at you, I think it's really amazing what you're doing for our platform. :)


Due to an increased frequency of spam, comments are now subject to moderation.