Luca Bruno

Thanks for the report. Generally speaking errors can come from 1) a mix of several kernel syscalls, or 2) improper API usage. The latter is more or less already documented...

On top of raw performance timing, after profiling some isolates in production we suspect that this is also causing a lot of memory pressure. One possible culprit is the `StringOrInit`...

General doubt: a `coreos-installer` somewhere which is not the same thing as `coreos-installer` somewhere else seems problematic. Could we consider adding some kind of highly visibile prefix/suffix to those aliases?

Thanks for the report. Have you perhaps tried the features-set described in https://github.com/tikv/rust-prometheus/issues/342#issuecomment-783738547 already?

Mmh, it looks like the `libc` crate is missing some constants on Android and uclibc. I'll try to find some time to look into that.

Ok, I landed several fixes in `libc` and now the CI here is green. This is ready for review.

@rtzoeller @asomers may I ask for a review on this one?

@SteveLauC thanks for the review! (And no need to apologise :blue_heart:) I've rebased against latest `master` and addressed the comments.

This is a bug in Satellite, as `application/json` is a binary format and thus doesn't allow a textual charset. This was already discussed, covered and fixed in https://github.com/opencontainers/distribution-spec/pull/41. While I...

> Do I understand correctly now that the content-type is actually contradicting in itself? Yes. `application/json` takes no additional parameters, as per RFC. That said, I'm fine with either landing...