that's a lot of red!

Analysis of PR #17364 at c1ac8a8b:

Test 1560 failed, which has NOT been flaky recently, so there could be a real issue in this PR. Note that this test has failed in 216 different CI jobs (the link just goes to one of them).

Generated by Testclutch

I was testing it by statically linking libcurl and it was working well, but when I dynamically link it the same curl_url_get returns the punycoded url. I've made sure the curl-config in my path is the one built from curl. I've configured curl as follows:

./configure --with-openssl --with-idn2 --with-ldap --enable-test-bundles --enable-debug 

The same fix is also needed here:

diff --git a/lib/urlapi.c b/lib/urlapi.c
index 071ee057eb..0ec3de32ec 100644
--- a/lib/urlapi.c
+++ b/lib/urlapi.c
@@ -1588,11 +1588,11 @@ CURLUcode curl_url_get(const CURLU *u, CURLUPart what,
         *part = allochost;
 #endif
       }
     }
     else if(depunyfy) {
-      if(Curl_is_ASCII_name(u->host)  && !strncmp("xn--", u->host, 4)) {
+      if(Curl_is_ASCII_name(u->host)) {
 #ifndef USE_IDN
         return CURLUE_LACKS_IDN;
 #else
         char *allochost;
         CURLcode result = Curl_idn_encode(*part, &allochost);

Thanks for the hint. I was chasing fake configuration errors. I'm not sure why those two tests are failing, it looks like one is flagged as flaky in testclutch, but the OmniOS one is a bit confusing. I have to get ready for work but I can look at this more this weekend.

Thanks!