Live data from Hacker News

Mail Data Loss in macOS 10.15

mjtsai.com

61–70 of 131 posts

Re: Mail Data Loss in macOS 10.15

#61

The position of Apple today reminds me a lot of Microsoft after they peaked (~2000?). Cocky to the point of arrogance because of market position, past innovations and brand loyalty (e.g., it is cool to own Apple). But loyalty and market sentiment can only endure so much amidst plummeting product quality, nonsensical pricing (e.g. dongles) and lack of any meaningful innovation (thinner doesn't count). Apple doesn't li…

I don't know if I'd term it 'cocky'. It's more that they don't have to pay attention to the competition so their mind wanders. When you're behind you can have laser focus on precisely what's important and what's not. MacOS isn't important to Apple any more which is why they have an OS full of bugs.

Re: Mail Data Loss in macOS 10.15

#62

I wish I only lost mail data in the upgrade to Catalina. I lost 400GB of data when the installer repartitioned the hard drive and the files that were in folders in root went bye bye. I know I'm not supposed to put stuff there, but it was a family video project I was putting together for my wife for Christmas, and she would have seen it anywhere else. There should have at least been a notice or a warning or some kind…

Did you backup before you upgraded?

Re: Mail Data Loss in macOS 10.15

#64
A lot of this is a repeat of weird issues that we saw 8(!) years ago with OSX Lion. Perhaps I'm off the mark here, but I strongly suspect there's lack of a proper manual QA team at Apple.

QA at most so-called "modern" companies these days has turned into overly fetishising automated testing. Automation is valuable, but crucially also leads to situations where the least-common-denominator situations get tested since it's easier to automate the conditions for testing them. So for example you might write an automated test that verifies that the mail index is correctly upgraded from (say) El Capitan to Mojave to Catalina; but that doesn't show the whole picture. There could be 20 other complex factors that affect it when it's used as part of an OS upgrade, and none of those get tested.

Project managers, line managers, and other suits also seem to love automated testing, because you can show off some kind of real number to someone in an "Oh we wrote 200 tests this cycle, isn't it great that we can run them everyday" way, even if there are 0 issues found. Manual testing in the same case leaves you with no output that can be run in an automated fashion. Also, no one wants to argue against automation (even if it's taking away manual testing resources) because that can make you seem old-school and out of touch. And once this attitude gains steam, no one on the QA team will want to go against the grain and say "You know what, we should spend more time not writing automated tests".

All of this is to say -- sometimes during testing, you simply have to put yourself in a user's place (for as many different kinds of user you can imagine) and actually use the thing you're supposed to be testing. I strongly suspect that Apple is not investing enough, or at all in this kind of testing. Instead, they are pushing it off to those enthusiasts who install and use beta or newly released versions of their software. So far, this strategy seems to have paid off for them, but perhaps people will wise up and stop testing their stuff for free. I am certainly not upgrading my machines to Catalina any time soon.

Re: Mail Data Loss in macOS 10.15

#65

A lot of this is a repeat of weird issues that we saw 8(!) years ago with OSX Lion. Perhaps I'm off the mark here, but I strongly suspect there's lack of a proper manual QA team at Apple. QA at most so-called "modern" companies these days has turned into overly fetishising automated testing. Automation is valuable, but crucially also leads to situations where the least-common-denominator situations get tested since i…

Not so much automated testing but too much focus on unit and shallow integrations and not enough the big picture testing.

Such as load testing, end-to-end testing, fuzzing. All of which can be automated to a large degree.

Re: Mail Data Loss in macOS 10.15

#66

A lot of this is a repeat of weird issues that we saw 8(!) years ago with OSX Lion. Perhaps I'm off the mark here, but I strongly suspect there's lack of a proper manual QA team at Apple. QA at most so-called "modern" companies these days has turned into overly fetishising automated testing. Automation is valuable, but crucially also leads to situations where the least-common-denominator situations get tested since i…

Not so much automated testing but too much focus on unit and shallow integrations and not enough the big picture testing. Such as load testing, end-to-end testing, fuzzing. All of which can be automated to a large degree.

My hypothesis is that the difficulty of writing automated end-to-end tests where failure can reliably indicate actual brokenness, combined with some of the other factors I mentioned is the reason for all the focus on unit and shallow integrations that you mention.

Re: Mail Data Loss in macOS 10.15

#67

I always find it strange when OS updates affect apps in this way. It just shouldn't happen. It reminds me of when Linus preaches that "the kernel should never break userspace". Why would a OS update suddenly cause the Mail app -- developed by the same company as the OS, for what it's worth -- to have these strange (and catastrophic) data syncing issues? And, just to add, I am tired of OS updates updating and/or relea…

> I always find it strange when OS updates affect apps in this way. It just shouldn't happen. Right? Are there not automated tests for this stuff? I would expect that if you commit something that causes this, the next day you get an email that says, "your name is on commit $FOO, which caused regression tests to fail in tests/firstparty/apps/mail; please fix it (sincerely, Jenkins)" or so. Heck, it's more egregious he…

I wouldn't place much emphasis on testing, and Apple most likely has tests, because the exact conditions that trigger such bugs are often so narrow that an artificial testing environment for them to "make the stars line up" would be nearly impossible to create. Anything involving multithreading/asynchronous code tends to fall in that category, but even single-threaded code can be like that; to make an artificial relevant example, it could be something like "if two consecutively received emails are exactly 8192 bytes long[1], and the first one is the 256th message in the folder when sorted by the sender's email, then the second one will be missing." That's a condition which you're highly unlikely to hit, even more so if you're generating test data randomly (as some tools like to do, and fuzzing is one prominent example). Even if you specifically test power-of-2 sizes, the two-in-a-row subcondition would make that improbable. If you let everyone in Apple "dogfood" the software, they might not hit. Yet everyone who has subscribed to a set of mailing lists that happened to issue such an exact sequence of messages would see the bug.

That's also why I think other automated tools and "code quality" metrics don't help so much; tests are more of a sanity/last-chance check and shouldn't be relied upon to catch everything. Correctness should be assured through design and code reviews, and for that nothing beats the power of intense mental deliberation. To paraphrase a popular saying, "there is no replacement for intelligence."

[1] https://news.ycombinator.com/item?id=8797640

Re: Mail Data Loss in macOS 10.15

#68

A lot of this is a repeat of weird issues that we saw 8(!) years ago with OSX Lion. Perhaps I'm off the mark here, but I strongly suspect there's lack of a proper manual QA team at Apple. QA at most so-called "modern" companies these days has turned into overly fetishising automated testing. Automation is valuable, but crucially also leads to situations where the least-common-denominator situations get tested since i…

Also, no one wants to argue against automation (even if it's taking away manual testing resources) because that can make you seem old-school and out of touch.

As someone who is "old-school" and has argued against automation countless times, it's more that your arguments against it will likely be dismissed because the management just wants to see the numbers on their pretty "quality metrics" dashboards going up; regardless of what it means for the actual quality of the product. They'll even ask you to "fix" automated tests by forcing them to pass, if they are failing for some reason you don't know yet. Truly a case of "you get what you measure". It's only when you start showing them the bugs that customers come back with and why they couldn't possibly be discovered in automated testing (at least without an impossibly large and nearly-infinite set of test cases on a similarly large set of hardware and software) that they may start to listen to you. Sometimes.

Re: Mail Data Loss in macOS 10.15

#69

A lot of this is a repeat of weird issues that we saw 8(!) years ago with OSX Lion. Perhaps I'm off the mark here, but I strongly suspect there's lack of a proper manual QA team at Apple. QA at most so-called "modern" companies these days has turned into overly fetishising automated testing. Automation is valuable, but crucially also leads to situations where the least-common-denominator situations get tested since i…

The reality is actually the opposite. They rely pretty much solely on manual QA and only in the past few years started writing unit tests, sometimes. Apple is actually quite an old school company some ways.

Re: Mail Data Loss in macOS 10.15

#70
I’ve never used Mail because I’ve never felt comfortable with Apple’s apps when it comes to long term data storage. Here Mail is said to delete data. This has happened before, IIRC. Someone in this discussion has written that their old downloaded podcast episodes were deleted on upgrade. In the distant past, Apple has also broken compatibility in data formats (for its iWork apps) and left those who upgraded with almost nothing.

Apple seems to focus too much on reinventing things, as if every app team gets a new (and inexperienced) set of developers each year and they see whatever is there as bad, incomprehensible code...so they write things from scratch, introducing tons of bugs and removing backward compatibility. Look at its app history and you’ll see a lot of work just thrown away every few years, only to be replaced with crippled applications that would take long to mature but aren’t allowed to.

Mozilla Thunderbird is “boring software”, and hasn’t made huge strides in UX in the last several years, but I trust it more than I would ever trust Apple Mail.

If I want “office software”, I have LibreOffice, which has its own issues in UX, but I can trust that it won’t drop backward compatibility at the drop of a hat.

If I want a browser, there’s Firefox. Safari 13 has become just ok now (from decent) because of the new extensions system that has junked all old extensions. It’s fine for firing up an ephemeral window, and it still provides the best battery life.

I feel the days of power users on the platform are numbered. I doubt if AppleScript and Automator will survive and be around in the next two years.

Is there any Apple app on Mac that’s been great and really improved in a stable manner?

Post reply on HN