Update signal_handler, removing bitflags

[hecatia-elegua]

First I wanted to change that piece of code to use bilge, but I'm not even sure why this wasn't an enum in the first case.
It's just some commented out code, btw.
Also using variant_count.
#963

[kevinaboos]

variant_count is cool, thanks for informing me about that feature! Didn't know that existed; we should definitely be using that.

Regarding the set of exception vectors, it's purposefully not an enum because we want to be able to specify a union of multiple variants at the same time, mostly for the purpose of registering a single handler that covers multiple distinct exception variants. While I know that code isn't used yet, that's why we used bitflags there. In fact, the only time we prefer bitflags is in this kind of scenario where we want to allow multiple variants to be selectable using only the space of a single primitive type, e.g., u32.

In general, I'm excited to experiment with bilge and highly in favor of replacing bitflags (and other related crates). This PR doesn't seem to include bilge; was that planned for an upcoming changeset?

[hecatia-elegua]

Well I can change it to use a struct with bools instead, just didn't seem like the right thing to do

[kevinaboos]

Well I can change it to use a struct with bools instead, just didn't seem like the right thing to do

Do you mean a struct of bools that uses bilge (or another crate) for bitpacking? Because a regular struct of bools would be much larger -- at one byte per bool, that's 32 bytes instead of 32 bits. That's also another reason why we use bitflags, to pack a lot of booleans into a small space.

[hecatia-elegua]

Edit: yes, I mean a bilge struct, which is bitpacked

[kevinaboos]

Edit: yes, I mean a bilge struct, which is bitpacked

Ah i see, ok that does make sense.

For this PR, I think it's best to accept it with just the variant_count addition. I do want to use bilge, but the right place to experiment with that is the pte_flags crate, which is actually used in tons of other places and would be a much better proving ground for bilge. In contrast, if we add usage of bilge here in the signal handler crates, there's no easy way to actually test it thoroughly because this code isn't used and it also doesn't require any complex features like pte_flags do.

[hecatia-elegua]

Please look at #963. It lists all places with bitfiddle/field/flags, some of those are just tiny changes or notes, though.
As discussed shortly on discord, pte_flags is a bit special due to how bitflags works, so we want to look into other usecases too.

[kevinaboos]

Thanks!