blob: 844641ecaf836315f67131ac92997462ea2a8a10 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
|
use crate::Syslog;
#[test]
fn call_init_twice() {
// 0. First test that a first call to initialize works.
if Syslog::builder().build().init().is_err() {
panic!("First request to initialize Syslog failed.");
}
// 1. Then assure that another attempt fails, as there can only be one logger.
//
// Reminder to self: Since this init() already returns a Result, there is no real need to
// deal with implementing a separate check with `std::sync::Once`. The log crate practically
// covers our ass here.
if Syslog::builder().build().init().is_ok() {
panic!("Second request to initialize Syslog unexpectedly succeeded.");
}
}
// NOTE It would be neat to have higher test coverage, including some of the actual logging, but
// the question is how to obtain that in a meaningful way. Stubbing out `libsyslog-sys` would
// make the tests mostly pointless, right? Targetting the host's `/dev/log` through `syslog()`
// practically leads to nothing but unverifiable side-effects. Running with a custom syslogd
// look-a-like, in a chroot or by other conceivable means, would require priviledges not
// suitable for unit testing.
//
// On a more practical level, multiple tests can not run in threads of the same test runner
// due to the singleton nature of Syslog, plus the fact that the foreign functions openlog()
// and closelog() should really only be called once per process.
//
// There is https://github.com/rust-lang/rust/issues/47506 which points out this problem, but
// the closest to a solution mentioned is a recommendation to use the `rusty-fork` crate. A
// crate that seems to need a bit more love than anyone appears to be able to give it. By the
// look of things, https://github.com/AltSysrq/rusty-fork/pull/10 is approaching three years
// of waiting in a believed-to-be-mergable state.
|
