On Tuesday, 2 June 2020 21:17:32 CEST Carlos Cabanero wrote: > We need it, so let’s get to work then.
Awesome! We use gitlab for development in the meantime and have a complete CI running: https://gitlab.com/libssh/libssh-mirror/ All new stuff needs to come with tests! Untested code is broken code :-) > Is an interface like the one you suggested in that thread still how you > would like it to look? I'm adding Tilo as he is probably also interested in that API. He also started to look into it iirc. How it should look like I'm not 100% sure, but maybe we should model it after liburing, which is the new hip async IO interface of the Kernel. https://kernel.dk/io_uring.pdf https://github.com/axboe/liburing/blob/master/src/include/liburing.h > My only constraint is that I need some type of callback to receive the data > or bytes written, and that I think it would be better if we could control > the offset from outside instead of done by libssh (as currently on the > read). I’m getting great numbers by scheduling multiple concurrent reads at > 32KB size each, so I was thinking scheduling multiple writes would be the > way to go too. Yeah, the current API is not nice at all. I'm open for suggestions for the current API. Andreas -- Andreas Schneider a...@cryptomilk.org GPG-ID: 8DFF53E18F2ABC8D8F3C92237EE0FC4DCC014E3D