Neven Sajko

In case it helps, there are three different key file location cases here that mkinitcpio supports: 1. key file is on a block device, possibly with an offset 2. key...

Judging by your last comment it wont be merged, but this is what I meant with "the concurrency issues could be fixed by hiding the global variables behind atomic set/store...

The diag_l0_debug, diag_l1_debug, diag_l2_debug, diag_l3_debug, diag_cli_debug fall in the same "extern and in global scope basket".

> However, backward searches are still not as efficient as they could be because they are implemented in terms of repeated forward searches. > except for a reversal of all...

> should probably write down desirable properties and possible alternative designs. I'm looking into the same stuff myself currently, so here's what I found so far, hope it's not too...

A quick thought, maybe useful: I think that part of the reason for the current inefficiency is the bool-float multiplication method from base/bool.jl `Base.:*(x::Bool, y::T)::promote_type(Bool,T) where T

Do you mean just indiscriminately replacing all the openlibm implementations with musl implementations, or would somebody first check which implementation is better?

A quick look at this commit's commit message seems to indicate it would fix the issue: https://chromium.googlesource.com/breakpad/breakpad/+/605c51ed96ad44b34c457bbca320e74e194c317e

According to the linked commit message, this is triggered by glibc > 2.33. So GCC version doesn't matter, I guess.

I think this problem may not be as easy as it seems at first hand, because any type may be callable in Julia and the proposed suggestion only makes sense...