Fix issue where read_to_end() into provided rx_buf does not work with… #115
Conversation
| // Also, if requests on embedded platforms fail with Error::Dns, then try to | ||
| // enable the "alloc" feature on embedded-tls to enable RSA ciphers. | ||
| let mut request = client | ||
| .request(Method::GET, "https://api.dictionaryapi.dev/api/v2/entries/en/orange") |
There was a problem hiding this comment.
Is there a way to test this without actually hitting this external API? We probably don't want the test to start failing if this website ever goes dark.
|
@lulf do you have any good idea on how we could setup a test harness that can recreate this issue? |
|
I wonder if there is something similar to Verify for rust, where we could load desired responses raw from disk, and then compare them to a desired and expected response. |
|
Hmm, I'm not familiar with any appropriate mocking framework, but maybe one could use something like https://github.com/drogue-iot/embedded-tls/blob/main/tests/tlsserver.rs to reproduce it. It requires a bit more work to setup though. |
|
Are there more potential uses for something like https://github.com/drogue-iot/embedded-tls/blob/main/tests/tlsserver.rs in this repo, or is this going into the territory of testing |
|
I also ran into this issue. Any updates on this? |
|
Unless anyone has a good idea on how to proper test this, i guess we have to merge this as is. @lulf what do you think? |
|
Maybe we can just switch to calling a different API in the tests (something more stable, with less chance of being shut down)? Or delete the test entirely (as much as it pains me to say that). |
… tls
This is a fix for #114. The issue is that
embedded-tlsmay provide a different buffer when callingbuf_read()than we expect, so this PR allows us to force to not use its buffer.