We don't know which of the claims of this patent bound on HTC.
The initial independent claims of patents are deliberately written to be as broad as possible, to insulate the patent from technological changes. The IP lawyers who draft Apple's patents go way out of their way to make sure that first anchoring independent claim is as general as possible.
A reasonable expectation is that it's not simply the initial independent claim --- which, by design, is the claim mostly likely to be challenged as over-broad --- but a combination of independent and dependent claims that actually bound HTC.
Given that, I'm super not interested in discussing whether a court is likely to find that any computer ever built running Linux violates this patent. I'm strongly inclined to believe that no, your infringing product needs to embody a purpose-built device.
I think your logic about "huge amounts of modern software patent language" pertains to the patent description, not the claims.
Does the answer to the above question affect how you feel about the patent (and/or patents in general)?
Generally, I think software patents are as bad as we tend to say they are.
However, caricaturing their badness with statements like "Apple patented regular expressions on URLs" or whatever sucks power away from the argument that they should be abolished, by allowing the "grown-ups" to point and say that opponents of patents think that e.g. this one says you can't regex URLs.
A patent protects the invention, not the text of the patent. The more realistic question is whether HTC's implementation was worked out by looking at how the feature works on iOS.
So far as I understand, iOS doesn't have the same sort of Intent architecture, so it seems unlikely that that Google's implementation drew any inspiration from iOS (especially as the same sort of thing has been done in web browsers for years to the point where a browser might be prior art).
But regardless of that, having read the 21 claims of the '647 patent (I don't fully understand the other one, but this is the one that covers what this thread is discussing), there's nothing about it being tied to purpose built hardware.
The example implementation described in the patent itself uses a Power Macintosh, the only part that refers to purpose built computers is "For example, program 165 in FIG. 1 can be stored in ROM, disk, or in dedicated hardware. In fact, it may be realized as a separate electronic circuit."
The term "a processing unit coupled to the input device, the output device" is only being used as a generic way of describing a computing device (hence the Macintosh being used in the example and the reference to dedicated hardware is only made as an alternative method of implementation)
If this patent is valid, it's likely covers a wide range of computer application including those found on desktop computers (for example Skype's phone number highlighting plugin).
So in other words... a computer?
A huge amount of modern software patent language is simply generic, copy-paste babble to say, in long and drawn-out terms, "on a computer". This doesn't make the patent any less stupid.