Daniel Stenberg
Daniel Stenberg
We see a similar problem (seemingly SONAME related) on Linux: https://github.com/curl/curl/pull/14030
I don't think this is the 11129 bug, because @roobre says it reproduces in 8.0.1 and it does not seem related to the URL...
I don't think adding text about another option to the error message is a good idea. It complicates the message and error messages are already hard. As @dfandrich says, the...
The string for `--socks5` is parsed as a URL. When the requested socket type is socks, the hostname is `localhost` and there is a path component, that path is used...
> Another solution may be adding a --socks5-socket option. It could work in all envs and filename formats without the localhost trick. That would also require a new option for...
How about for example we simply skip the leading slash from the path component when used on Windows? ~~~diff diff --git a/lib/url.c b/lib/url.c index f646b74ed4..d2ea659dfb 100644 --- a/lib/url.c +++ b/lib/url.c...
Using drive-letters to a path is already not portable...
> Also for relative paths which remains a valid use-case It might be, but it can be noted that we don't support relative paths for other platforms.
> (Windows also allows using backslashes in filenames, which would be nice to support too.) Aren't they already implicitly? I don't think the URL parser objects to them, does it?
> > Using drive-letters to a path is already not portable... > ? Sorry, I don't understand. You argued that special-casing Windows would make the option non-portable. I argue that...