Chris Down

icon indicating a website link https://chrisdown.name icon indicating an email [email protected]

icon company @Facebook icon location London, England icon location I keep @facebook running as part of Meta's Linux Kernel team, and also help maintain @systemd. I mostly work on memory management and cgroups.

Results 178 comments of Chris Down

Hash collisions regularly occur between strings containing Chinese characters

The decoding is fine, we just read them as uint. The choice to use a simple hash function is intentional: we need to handle collisions anyway, and this one is...

Hash collisions regularly occur between strings containing Chinese characters

3ddf819cc4951dfc6ae8df16bd6cb79498cdd694 will help with collisions. We should still do collision detection, but at least this avoids the immediate problem.

Hash collisions regularly occur between strings containing Chinese characters

Thanks for reporting! Now both real world collisions are highly unlikely, and they will be rejected.

core/service: Add ReloadSignalRequireHandler to require signal handler

Thanks for the review! > apps should never issue the initial READY=1 before they installed the SIGHUP handler. It's a bug if they do that. I guess we could make...

core/service: Add ReloadSignalRequireHandler to require signal handler

> Hmm, I'd suggest the other way around and fail the reload job but never the unit itself. Right. > I don't follow the outage part btw. We should not...

core/service: Add ReloadSignalRequireHandler to require signal handler

superceded by #39692

suggestion: optionally print selection instead of setting CLIPBOARD

It's reasonable to have that be configurable, thanks.

suggestion: optionally print selection instead of setting CLIPBOARD

oh, sorry, wrong issue :-)

core: Prevent corruption from stale alias state on daemon-reload

@DaanDeMeyer I have an integration test that doesn't require fixed ordering patch coming up, once that's done I will move this out of draft :-)

core: Prevent corruption from stale alias state on daemon-reload

@ikruglov That was my first thought as well (to avoid duplicating the parsing logic), but I backed away from it because of side effects. Unit objects are pretty tightly coupled...