Proper error reporting for connect failures #578
Conversation
|
Note that another PR (#461) also exists. Unlike the linked PR, this PR doesn't add a new API, and it doesn't break the abstraction layer between socket and context. |
|
I'd also like to abstract getaddrinfo to be non-blocking; right now it's blocking and in async mode we never really rotate. I'd also like to abstract the sockaddr member as well. |
|
Cc: @yossigo can you review this? |
net.c
Outdated
| if (c->saddr) { | ||
| free(c->saddr); | ||
| } | ||
| c->saddr = malloc(sizeof(*p->ai_addr)); |
There was a problem hiding this comment.
The allocation should be done based on ai_addrlen, as ai_addr is a vanilla sockaddr which may be smaller than what you really need (e.g. for AF_UNIX or AF_INET6).
There was a problem hiding this comment.
oops - done. There's also struct sockaddr_storage, but ideally I'd like to preserve the original addrinfo structure anyway.
|
Still working on getting the tests to pass.. |
This retrieves the actual error which occurred, as getsockopt is not always reliable in this regard.
Not sizeof saddr.
Some ISPs like to inject their own "Suggestions" page whenever you hit NXDOMAIN. This confuses Redis as well as addrinfo (black-holing the route).
No description provided.