Repository navigation
Conversation
|
|
||
| steps: | ||
| - uses: actions/checkout@v6 | ||
| - name: setup |
There was a problem hiding this comment.
Having a MacOS-specific change in a PR that explicitly addresses MSVC is a bit weird.
There was a problem hiding this comment.
Fair comment - I will remove the Mac-OS CI run from building the tests and create a separate PR for it.
There was a problem hiding this comment.
Actually, because building unit-tests is introduced in this PR, cunit needs to be installed, or MAC (APPLE) needs to be explicitly removed from building testdriver in tests/CMakeLists.txt and then a new PR created specifically for MacOS.
| *buf++ = number & 0xFF; | ||
| } | ||
| return klen; | ||
| return (int)klen; |
There was a problem hiding this comment.
Since we are casting size_t (usually an unsigned type) to int I wonder if we should do a bit more hardening here. The compiler warning wanted to warn us about something, I suppose.
There was a problem hiding this comment.
As can be seen in https://github.com/eclipse-tinydtls/tinydtls/actions/runs/37139114165/job/111249679547?pr=296 (expand out Build tinydtls) there are many implied casting issues which need to be resolved.
I just fixed the windows specific file one to align with the dtls_prng() definition as it is very unlikely, even on a 16 bit system, the be filing a buffer of 32768 bytes. Perhaps the better fix there would be to update the return from dtls_prng() to be the same as the passed in len parameter.
There was a problem hiding this comment.
Hm, this is interesting. The original intention was to have boolean result, so len == 0 would have done the trick. Now, this function has two exits: First is a return err when rand_s returns a non-zero result. AFAICT errno_t yields a positive value if an error occurs. So, dtls_prng() will indicate a "good" result even after the PRNG_FAILED message.
There was a problem hiding this comment.
It makes sense to me that this is boolean (as indicated in dtls_prng.h), but all the other dtls_prng() variants return the filled in length to pass testdriver's ecc tests.
I will do a separate PR to make it all boolean for dtls_prng().
There was a problem hiding this comment.
Separate PR created for dtls_prng(), and changes to dtls_prng() removed.
|
Some of this work is duplicated, or done differently as in #263. |
There are a lot of size mismatch warnings that still need to be reviewed and fixed. The tests have been updated to build with a MSVC environment. Signed-off-by: Jon Shallow <supjps-libcoap@jpshallow.com>
2 participants
There are a lot of size mismatch warnings that still need to be reviewed and fixed.
The tests have been updated to build with a MSVC environment.
Fixes #283.