Jay Satiro

Of course I'm aware of the caching of results. In my opinion caching a negative result is different because I would think it's more likely to turn positive in the...

Sure, I'm not against DNS caching of positive results I think it is very helpful. Negative results in my own experience in the browser have just been so temporary that...

We probably don't support that compiler any longer. /cc @vszakats

> I am not sure that this macro MP_HAVE_INT_EXTENSIONS that limits the format string parser abilities (when not defined) is necessary. IMO it is a point of failure. The format...

All data should be passed to the client immediately if the transfer is not paused. There are clients that process HTTP responses line by line (usually sent by chunk) and...

> It was triggered in pytest `test_20_06`, but it depended on how data arrived from the server. Would [CURL_DBG_SOCK_RMAX](https://curl.se/libcurl/c/libcurl-env-dbg.html) work for that

I don't agree that it is dangerous. If you don't select a secure protocol then that's what happens. I like Dan's suggestion to use [--proto-default](https://curl.se/docs/manpage.html#--proto-default) in curlrc to force it...

Please review [libcurl thread safety guidelines](https://curl.se/libcurl/c/threadsafe.html), specifically that connection caches cannot be shared between threads. Also try the latest curl, 8.6.0. #10936 and #12356 may be related and both fixes...

This is most likely a problem with your resolver configuration and not with curl. curl uses getaddrinfo for IPv6 builds. Try a program that also uses getaddrinfo like wget. Try...

Please try [getaddrinfo.c](https://gist.github.com/jay/0fe0db29305184d63b7914acb3eae820) which retrieves addresses the same way libcurl does.