Archive for July 22, 2016

Friday, July 22, 2016 [Tweets] [Favorites]

Xcode 8 Illegal Hard Links Prevent Cloning

After I installed Xcode 8 Beta 3, I could no longer back up my hard drive. SuperDuper reported errors like:

Error creating hard link /Volumes/HD Clone 14A/Applications/ to /Volumes/HD Clone 14A/Applications/ for inode (null)

Cloning HD, error creating hard link for file in

At first I thought that the drive was damaged, but the error still occurred after reformatting it. SuperDuper’s Dave Nanian:

Structure became illegal (hard link into app bundle) in 10.10.3 - break link by duplicating file...

I assume they just did this in the latest beta. (It was previously an issue with Retrospect & Anaconda.)

Indeed, does contain a hard-linked file:

ls -l /Applications/\
total 48
-rw-r--r--@ 2 mjt  staff   6141 Jul 14 18:08 AuditIssues.strings
-rw-r--r--@ 1 mjt  staff     42 Jul 14 18:08 Localizable.strings
-rw-r--r--@ 1 mjt  staff  10726 Jul 14 18:08 LocalizableOSX.strings

Note the “2” for “AuditIssues.strings”. Here are the two files with the same inode:

find /Applications/ -name AuditIssues.strings -print -exec stat -f "%i" {} \;

The fix is to make the two files independent copies:

cd "/Applications/"
mv AuditIssues.strings AuditIssues.strings.old
cp AuditIssues.strings.old AuditIssues.strings
rm AuditIssues.strings.old

Now the files have different inodes:

find /Applications/ -name AuditIssues.strings -print -exec stat -f "%i" {} \;

Update (2016-07-22): Mike Bombich (developer of Carbon Copy Cloner) on the similar issue with Anaconda:

This turns out to be an issue specific to *.app/Contents/PkgInfo and *.app/Contents/Resources/*.lproj files. OS X does not want to permit the creation of a hard link between one of these items in an application bundle to another file in a non-application-bundle folder. I was unable to find an explanation for this behavior in Apple’s documentation, nor in the source code for HFS or the OS X kernel.

Pixellating or Blurring Text Creates Identifiable Patterns

Kashmir Hill (via Nick Heer):

If you’ve ever pixelated an email address or blurred a phone number before putting an image onto the internet in order to protect someone’s privacy, I’ve got bad news for you: Researchers at the University of California-San Diego have found that the popular Photoshop redaction techniques are decodable such that the underlying text can be read.

The researchers were able to recover text from a variety of redacted screenshots that they found online, said computer science professor Hovav Shacham by email. They were, for example, able to figure out the blurred email address in this screenshot of a conversation between a corrupt DEA agent and the then-CEO of Bitcoin exchange Mt. Gox.

What Exactly Is “Compressed Memory”?

John Siracusa:

In Mavericks, the OS has one more option before it has to resort to swapping: compressed memory. Mavericks will find the least-recently-used data in memory and compress it, usually to about half its original size. Et voilà, more free memory.


Memory compression is a triple play for Mavericks. It’s a performance win; compressing and decompressing data in RAM is much faster than reading from and writing to disk, even an SSD. It’s an energy win; the less time spent moving data between RAM and disk, the more time the system can spend in its idle state. And finally, it’s a capability win; Mavericks can handle much more demanding workloads than previous versions of OS X before crying uncle.

This seems like a great feature, but I’ve never fully understood how it’s reported in Activity Monitor. What do the “Compressed Memory” numbers for each process and the “Compressed” total actually mean? This AskDifferent post lists the columns in Activity Monitor, but commenter James K Polk has the same questions as me:

If the activity monitor says a given process uses 621.4 MB of memory and 615.4 MB of compressed memory, does that mean that the process is really only using up 6 MB of memory? Or that 615.4 out of 621.4 MB was compressed down to some unspecified size? Or something else?

In other words, does a high number for Compressed mean that the compression has been effective or that the system is nearing capacity because most of what can be compressed already is?

It’s also not clear how compressed memory interacts with the other reported numbers. For example, why did Siracusa’s App Memory go down when the Compressed memory went up? I would have expected that App Memory would include the part that’s compressed.

And, presumably the memory that is paged out to disk is compressed. Does Swap Used take that into account?

Update (2016-10-03): Activity Monitor in macOS 10.12 has a tooltip for the “Compressed Memory” column, which says “Memory sent to the VM compression.”

Sandboxing Wisdom

Daniel Jalkut (tweet):

[I’ve] managed to produce two versions of my app, one of which causes the sandbox container to be apparently unwritable to the other after running! Specifically, preferences are not saved and console messages indicate an attempt to write preferences outside the host app’s sandbox.


These kinds of issues scare the bejeezus out of me because I really fret my users running into data migration problems after I ship an update, and because the relative opacity of the sandboxing system makes a lot of issues very hard to debug.

Daniel Jalkut:

In a nutshell: for the past 4 years or so, sandboxing has been a massive, amorphous “bug” that I have to wrap my head around. So, so tired.

Rarely a day goes by when I don’t worry that I made a huge mistake betting on sandboxing and MAS for the long run.

Peter Maurer:

Answer from a very tired sandbox wrangler: Don’t do it. Don’t waste your time on custom settings, etc. Instead, import once…

…via open dialog automatically, then make additional imports (for whatever reason) available via menu or preferences.

Ilja A. Iwas:

We released GarageSale 7 this week, still sandboxed, but won’t be submitted to the MAS. Ah, that sweet feeling of relief.

Daniel Jalkut:

This tool [asctl] appears to offer extensive insight into the sandbox’s understanding of containers. I wish I had discovered it earlier!

Apple doesn’t seem to have posted the asctl man page, but this online version has the same date stamp as the man page on my 10.11 system.