Remove unsafe mem::transmute calls

[joshlf]

No description provided.

[epage]

Is there a concern with the unsafe calls besides it being unsafe? I feel like this doesn't seem to justify the cost of a dependency.

[joshlf]

The only concern would be future-proofing against changes that affect the layout; with this change, those would be detected and compile failures. But it's a minor concern; definitely a judgment call as to whether it's worth taking a dependency for.

[coveralls]

Pull Request Test Coverage Report for Build 16151580562

Details

• 2 of 8 (25.0%) changed or added relevant lines in 2 files are covered.
• 2 unchanged lines in 1 file lost coverage.
• Overall coverage decreased (-0.001%) to 46.632%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
src/stream/bstr.rs	1	4	25.0%
src/stream/bytes.rs	1	4	25.0%

Files with Coverage Reduction	New Missed Lines	%
src/error.rs	2	27.82%

Totals
Change from base Build 16149430809:	-0.001%
Covered Lines:	1592
Relevant Lines:	3414

---

💛 - Coveralls

[epage]

I copied these from bstr and just looked to confirm what they do and apparently the transmute can be replaced with a cast, see BurntSushi/bstr@ef95d88. I'd be open to updating to match bstr.

[joshlf]

I copied these from bstr and just looked to confirm what they do and apparently the transmute can be replaced with a cast, see BurntSushi/bstr@ef95d88. I'd be open to updating to match bstr.

That retains the same fundamental concern about sensitivity to layout, but I agree that a cast is marginally better than a transmute. If you're going to go that route, I'd suggest the following variant, which avoids an as cast:

let slice: *const [u8] = slice;
unsafe { &*(slice as *const BStr) }

[epage]

I'd prefer to copy bstr verbatim rather than us paving our own path.