Last Week on my Mac: does 10.11.6 cut the mustard?

With macOS Sierra well on its way for an autumn/fall delivery, 10.11.6 could be the final version of El Capitan. Although it has only been out for a few days, those have been sufficient to give a good indication of its completeness.

What does it change?

With Apple’s terse and essentially uninformative release notes to go from, you could be forgiven for thinking that this update was small and unimportant. In fact a lot of key system files and apps have been altered, and some areas like audio and Bluetooth support have been largely revamped, hopefully fixing the lingering problems with spontaneous disconnection, and others.

This release is built around a new version of the kernel, dated 23 June 2016, and many of its key kernel extensions (KEXTs) have been updated too. Apple’s more detailed listing of security fixes admits to some of the changes, although it does not suggest how those might affect non-security issues. The XProtect system has also been updated.

The biggest changes which will be apparent to users are in Safari, which is brought up to version 9.1.2, with updates to its WebKit underpinnings too. These have brought some initial oddities in connecting to some sites: here I discovered that I had to re-enter login credentials for my modem-router, although Safari had previously been able to obtain them from my keychain.

Photos remains at version 1.5 but has substantial internal changes despite that. The iLife Media Browser, used by many Apple and third party apps to provide common access to photos and other media, has been updated too.

Other apps which have been updated in unspecified ways in 10.11.6 include:

App Store
Contacts
Facetime
Messages
Notes
Preview
QuickTime Player
Airport Utility
Audio MIDI Setup, presumably to work with audio fixes
Bluetooth File Exchange, presumably as part of the overhaul of Bluetooth
Disk Utility
Keychain Access
iBooks
AirPlay
several widgets
several Automator actions
System Image Utility
Managed Client
ACFS file system support.

My review yesterday of known bugs in El Capitan shows that 10.11.6 has apparently had little impact on that list: most of the more annoying and frustrating bugs in 10.11.5 have been left untouched. (My earlier opinion that the Preview bug had been fixed turns out to be wrong: the bug remains untouched.)

The most serious question over 10.11.6 is whether it fixes the sporadic problems with freezing on some models of Mac. As these occurred every few days here, and only seem significant on a small number of models – including some MacBook Pros and some iMacs – only time will tell whether the new kernel and its extensions have done the trick.

What is still wrong?

This update has done nothing to reduce the torrent of messages being written to logs, nor to frequent crash reports and spindumps (generated automatically when the spinning beachball appears). I see that many days over 200 reports are being sent from my iMac to Apple’s servers. This is in spite of running mainstream apps, most developed by Apple or sold through the App Store, with a bare minimum of third-party extensions or goodies, and long hours spent carving away detritus migrated from previous systems.

My logs now look most peculiar. Much of the time that I am working, there are steady streams of incomprehensible messages and warnings, which do not appear to follow simple patterns. At times there are long bursts of hundreds or even thousands of identical messages, such as:

21/07/2016 21:14:21.001 com.apple.CDScheduler[61]: *** LOG MESSAGE QUOTA EXCEEDED - SOME MESSAGES FROM THIS PROCESS HAVE BEEN DISCARDED ***
kept appearing every minute or so for a day or two, then vanished. You cannot really get less helpful than a message telling you that messages have been discarded, can you?

21/07/2016 20:16:16.618 MarsEdit[24238]: Can't allocate a new block for a pasteboard. Creation of a new Pasteboard will fail.
spewed out in the thousands at one time, then mysteriously vanished.

21/07/2016 22:38:32.000 kernel[0]: Limiting closed port RST response from 355 to 250 packets per second
is another which came, filled the log for a while, and vanished.

21/07/2016 23:11:59.051 PluginProcess[29635]: CGContextScaleCTM: invalid context 0x0. If you want to see the backtrace, please set CG_CONTEXT_SHOW_BACKTRACE environmental variable.
is another completely unhelpful message which dominated entries before disappearing.

This is not the behaviour of a finished product, of a mature and reliable modern operating system. It better resembles a beta release, a work in progress. This failure to address longstanding known and obvious bugs, and profusion of log entries, implies that OS X El Capitan 10.11.6 was rushed to meet a deadline, perhaps being imposed more by Apple’s concerted update schedule involving its other operating systems, notably iOS.

Sierra or 10.11.7 next?

Apple has not told us whether it intends providing any more updates to El Capitan. Much will depend on progress with macOS Sierra and the other major system releases due this autumn/fall. If those are to appear together in October, we may still get one last update to El Capitan in late August or September.

As so often happens, Apple’s operating system problems are of its own making. Its determination to deliver major releases annually (since Lion in 2011) provides insufficient time in the cycle to achieve stable, essentially bug-free releases before being overwhelmed by the tsunami of the next major release. This is not just a matter of devoting sufficient engineering resources (developers) to the project: testing betas takes time, and securing fixes in complex interconnected software cannot be accomplished by some instantaneous magic wand.

An annual cycle would be hard enough to accomplish if Apple was developing and releasing a single operating system. But with four different operating systems which have to be coordinated, particularly in their connectivity and cloud features, Apple is demonstrating how the level of difficulty rises exponentially with each additional operating system.

Doing all this while keeping your users in the dark over what has changed has major impact on the user experience. Apple is currently only getting away with it because its competitors, Microsoft and the jungle that is Android, are doing even worse. That is no excuse for Apple, which must deliver more reliable operating systems at a pace which it can achieve, and not let itself be driven by its own marketing hype.

(Amended 1445 on 24 July 2016, to correct issue over Preview bug fix – it is not fixed at all.)

3Comments

Add yours

1

Tony on July 30, 2016 at 1:18 pm

Absolutely! I have never understood how Apple expected to sustain an annual release cycle for OS X. They have painted themselves into a corner by adding lots of coupling to iOS features, with such features tied to the need for new goodies with each iPhone hardware update.

Whilst the need to add features to each iPhone generation is probably unavoidable, I would have expected the Mac either to add its compatibility according to its own timetable (let’s face it, the majority of iPhone users don’t have Macs anyway) or to take a low-coupling approach of a Mac app to interoperate with a new iOS feature followed later by deep integration into OS X.

Rushing these features that were likely designed initially for iOS into OS X is probably the root cause of woes like the wireless stacks that were broken in Yosemite and continue to be fragile to this day (I haven’t yet tested 10.11.6).

LikeLike
2

Tony on July 30, 2016 at 9:13 pm

And while we’re on the subject, he said, warming to his theme, why did Apple launch its Public Beta programme? Beta Testing is an expensive, in engineering terms, exercise. The fault reports must be sifted and understood, those that are known shortcomings of the beta software and those that are just wrong must be removed. The remainder must, in the absence of the originator, be validated, recreated and fixed.

This all takes time and is futile whilst the underlying product is of insufficient quality: when the developers know that the code is not ready, the most efficient approach is to let them complete their systematic process of testing and fixing. A random approach (resulting from such fault reports as have been received) is simply disruptive.

As you have observed, El Capitan (and Yosemite before it) hardly seemed to reach beta quality months after release. This means the public beta exercise is in the disruptive category. In fact, the quality of the released product suggests that the inputs from (all) beta testing have not be actioned. So the logical approach is the traditional one: only do beta testing when you are ready and then do it with a small group of expert beta testers. That way, the number of unhelpful reports is much lower (it’s surprisingly difficult to write a good fault report, even for a valid fault, and that doesn’t consider all the “I don’t like” reports that will result from public exposure) and such effort as is available from the developers is used more effectively .

Dropping the public testing also removes the reputational damage that results from those who don’t understand what should be expected of a beta test product and consequently complain about every rough edge. I am surprised that Apple’s engineering management tolerates this.

LikeLike
- 3
  
  hoakley on July 30, 2016 at 9:23 pm
  
  Thank you for more excellent points. It seems now that the bulk of Apple’s test feedback is automated, via logs sent to its servers (as with ‘release’ products), so it may have reduced some of those costs. I don’t know how many developers participated in beta testing before the programmes became open, and suspect that most would have been more concerned at testing their own products than the OS.
  With good automated analyses, Apple should be able to get a much wider range of systems tested, which might help with some of the bugs, as they seem to be confined to certain models which might not be popular among developers.
  But yes, you should only go to open programmes when what you have is very close to release quality. OS X 10.11.8 or 9 perhaps?!
  Howard.
  
  LikeLike

Share this:

Like this:

Related