5 ms·

That would be quite a fundamental change. At the moment _int_ and _uint_ have certain semantics that, if changed, would surely break many applications and libra

That would be quite a fundamental change. At the moment _int_ and _uint_ have certain semantics that, if changed, would surely break many applications and libraries that rely on the current semantics. I am finding it hard to think of a more sweeping and drastic change to the core of a language. Having said that, I'm by no means a Golang expert – this is the comment of an outsider looking in. I get that the proposer is Rob Pike and obviously Mr. Pike is god-like so what is it I am missing? Is Go 2 meant to break everything? That's like, wow. Any low-level hardware bit-banging code is going to have to be audited or break in subtle ways, will it not? Is Golang different from C and C++ and languages of that ilk that this sort of change is not that much of a problem? Help me out here folks, I'm genuinely perplexed.

pstuart 8y ago
If these are new datatypes then they would break nothing (ignoring the "int" default part). > so what is it I am missing? The Go team has been adamant about avoiding any changes that break existing code. The whole point of Go 2 is that at some point that might have to happen and they're trying to minimize that pain.
HALtheWise 8y ago
If the change as originally proposed went through, it probably wouldn't affect most uses of bitbanging low level code because they are already using int32/int64 rather than int (which has different size on different platforms). You are correct that it could definitely break some programs though, and so with recent sensibilities veering away from making any non- backwards-compatible changes, an amended proposal would probably just introduce a new "vint" type (maybe removing the old int type at the same time if it is really ambitious)
copperx 8y ago
I've been thinking about this, and even though it's indeed a breaking change, it's one of those changes that can be easily toggled through a compiler flag. I cannot picture many libraries relying on the edge cases that Pike enumerated.

pstuart 8y ago

If these are new datatypes then they would break nothing (ignoring the "int" default part). > so what is it I am missing? The Go team has been adamant about avoiding any changes that break existing code. The whole point of Go 2 is that at some point that might have to happen and they're trying to minimize that pain.

HALtheWise 8y ago

If the change as originally proposed went through, it probably wouldn't affect most uses of bitbanging low level code because they are already using int32/int64 rather than int (which has different size on different platforms). You are correct that it could definitely break some programs though, and so with recent sensibilities veering away from making any non- backwards-compatible changes, an amended proposal would probably just introduce a new "vint" type (maybe removing the old int type at the same time if it is really ambitious)

copperx 8y ago

I've been thinking about this, and even though it's indeed a breaking change, it's one of those changes that can be easily toggled through a compiler flag. I cannot picture many libraries relying on the edge cases that Pike enumerated.