Andrew Clayton

> Seems reasonable, two nits/questions: > > * Do we need to worry about updating a `clean` target anywhere, or including this in the `-uninstall` target? No. This is copied...

Rebased with master ``` $ git range-diff 4a2a43fd...b5fe3eaf -: -------- > 1: d62a5e2c contrib: updated njs to 0.8.5 -: -------- > 2: a9aa9e76 python: Support application factories -: -------- >...

Hi, yes, this is something that has come up before. IIRC we have generally discounted simply trying to remove the socket at startup, for $reasons. However, here are some other...

> Regarding restarting the container/removing the socket on startup to fix the issue - the app's environment is actually rebuilt entirely if the container's restarted so the socket does get...

Well, it's been nearly two years since we attempted this patch. But things have changed, both internally and in the C world and larger programming world in general. I was...

> Both ways work for length calculation. Yes, both ways work (or not) to the same degree for _some_ definition of work. That's not the point. The point _is_ that...

> > Talking of concise, why even use nxt_length()? In its current form it isn't really any better than strlen(3). > > `sizeof()` is evaluated at compile-time, while `strlen()` is...

... and you say that 'you know not use nxt_length() on pointers', but you also know how easily (and unexpectedly) arrays decay to pointers...

How easy would it be to overlook something like ```c char *buf; nxt_str_t str; ... /* Having allocated and set buf */ nxt_set_str(&str, buf); ``` Why _not_ take defensive measures...

It seems like really, nxt_length() should be limited to static initialisations...