sys/net: fix output of ipv6_addrs_print()

[benpicco]

Contribution description

Output of ipv6_addrs_print() is broken when the fmt module is used.
There is also a duplication, so the output without fmt will be

{"IPv6 addresses": ["fe80::7837:fcff:fe7d:1aaf", ""]}

Just get rid of the fmt fallback and simplify the code.

Testing procedure

Run e.g. examples/telnet_server:

master

main(): This is RIOT! (Version: 2022.07-devel-726-g8e1f9)
RIOT telnet example application
╔═══════════════════════════════════════════════════╗
║telnet is entirely unencrypted and unauthenticated.║
║Do not use this on public networks.                ║
╚═══════════════════════════════════════════════════╝
{"IPv6 addresses": [""]}
All up, awaiting connection on port 23

this patch

main(): This is RIOT! (Version: 2022.07-devel-726-g8e1f9)
RIOT telnet example application
╔═══════════════════════════════════════════════════╗
║telnet is entirely unencrypted and unauthenticated.║
║Do not use this on public networks.                ║
╚═══════════════════════════════════════════════════╝
{"IPv6 addresses": ["fe80::7837:fcff:fe7d:1aaf"]}
All up, awaiting connection on port 23

Issues/PRs references

[kaspar030]

Just get rid of the fmt fallback and simplify the code.

Hm, there's an underlying issue that causes this, which should be fixed, no?

There was also a chain that made the ipv6_addr_print() a function (vs. shell commant), now fmt support is removed.
I did use this on our smallest micros, and intentionally used fmt to drop the libc stdio dependency. It was first half-added by fmt using "fwrite()" in all cases, now ipv6_addr_print() drops fmt support altogether.
I guess I can live with that for the sake of simplicity, but honestly I'm not happy how these optimizations got "cleaned up".

[benpicco]

Feel free to open an alternative PR :)

[maribu]

I hoped #18162 would fix the issue with broken output from fmt. The unit test in tests/fmt_print was extended to actually catch the issue. What board are you using? Does it pass the test?

[benpicco]

This is on native

[maribu]

fmt still works fine on real hardware as well in the CI. I guess this is the general native being broken issue.

I indeed fucked up the printing when reverting the drop of fmt. But how about fixing it like this instead:

diff --git a/sys/net/network_layer/ipv6/addr/ipv6_addr.c b/sys/net/network_layer/ipv6/addr/ipv6_addr.c
index a6a7962b08..6efafd0ac1 100644
--- a/sys/net/network_layer/ipv6/addr/ipv6_addr.c
+++ b/sys/net/network_layer/ipv6/addr/ipv6_addr.c
@@ -176,10 +176,9 @@ void ipv6_addrs_print(const ipv6_addr_t *addrs, size_t num,
     ipv6_addr_to_str(buf, &addrs[num], sizeof(buf));
     if (IS_USED(MODULE_FMT)) {
         print_str(buf);
-        print_str(separator);
     }
     else {
-        printf("%s%s", buf, separator);
+        printf("%s", buf);
     }
 }

[benpicco]

closing in favor of #18186