Fix warnings about implicit cast

[TedLyngmo]

Added a few explicit casts:
unsigned char flags = (unsigned char)s[-1];

Signed-off-by: Ted Lyngmo ted@lyncon.se

[TedLyngmo]

I see it's failing in Ubuntu and CentOS, presumably because of the skipped tests? Any idea what could cause it to skip the tests?

#86 Does not return a reply when the command times out: SKIPPED
#118 Does not return a reply when the command times out: SKIPPED
#147 Does not return a reply when the command times out: SKIPPED
*** 0 TESTS FAILED ***
*** 3 TESTS SKIPPED ***
Error: Process completed with exit code 1.

When running make -j hiredis-test && ./test.sh on my Ubuntu (in WSL2) it doesn't skip any tests and it passes all tests.

[bjosv]

Any idea what could cause it to skip the tests?

You might already have found the reason, but the command DEBUG SLEEP which is needed by these testcases was introduced in Redis version 2.2.12. Some testcases requires an already running redis and there might be an old pre-2.2.12 version running on those distros by default.

[TedLyngmo]

Any idea what could cause it to skip the tests?

You might already have found the reason, but the command DEBUG SLEEP which is needed by these testcases was introduced in Redis version 2.2.12. Some testcases requires an already running redis and there might be an old pre-2.2.12 version running on those distros by default.

Oh, ok, so there is currently no way to get patches through this gate - or what can I do?
I have some more ideas for patches to be able to enable more warnings. Will it be possible to get them in somehow?

[michael-grunder]

I'm hesitant to merge changes to the sds library because this code is also in Redis itself so we try to minimise divergence.

How are you invoking the tests? Treating skip as failures is an option that you don't have to set.

Edit: Ahh I see, they are even being skipped in GitHub CI. We recently switched to using the official Redis package so perhaps something has changed. I need to investigate why they're being skipped

[TedLyngmo]

I'm hesitant to merge changes to the sds library because this code is also in Redis itself so we try to minimise divergence.

I totally understand that. If I manage to get a PR through in Redis itself, will that eventually end up in hiredis? If so, I may start there instead.

We recently switched to using the official Redis package so perhaps something has changed. I need to investigate why they're being skipped

Great! Cheers! Oups, now the macOS test failed too. :-)

[michael-grunder]

If I manage to get a PR through in Redis itself, will that eventually end up in hiredis

Yes for sure

Great! Cheers! Oups, now the macOS test failed too

The problem is due to Redis 7.0.0 disabling the DEBUG command by default. The fix is pretty simple so should be up shortly.