savecore: add missing call to cap_openlog when in capabilities mode

[stephane-rochoy-stormshield]

When run in capabilities mode, savecore seems to lack a call to cap_openlog. Notably it break the use of LOG_PERROR.

With the original one:

$ /sbin/savecore -vC
checking for kernel dump on device /dev/gpt/swapfs
No dump exists

With the patched one:

$ ./savecore -vC
checking for kernel dump on device /dev/gpt/swapfs
savecore 1424 - - /dev/gpt/swapfs: Permission denied
No dump exists

[stephane-rochoy-stormshield]

I'm looking to add a test as it should be trivial.

[emaste]

Looks good. CC @oshogbo.

Also @bsdimp do we have advice on rebasing/squashing fixup commits in pull requests?

[stephane-rochoy-stormshield]

Looks good. CC @oshogbo.

Also @bsdimp do we have advice on rebasing/squashing fixup commits in pull requests?

Yes you have ;) I've put this PR in draft mode for now. I'll squash once ready.
I'm discovering atf(7) and kyua(1) to add at least a basic test.

[emaste]

Oh yes I didn't notice the Draft tag. The C change itself looks good :)

[bsdimp]

git rebase -i main
squash all the tests into one commit and all the stuff in the actual binary into one. I think that's just the first one and all the rest squashed.

[stephane-rochoy-stormshield]

Sorry for the initial noise. I cleaned the commits.

But one question remain about the test I propose to add: what is the proper way to test the instance of savecore from the installed build, i.e., not the one on the system? I find it convenient to make world DESTDIR=… but the tests don't seems to be aware of $DESTDIR.

[stephane-rochoy-stormshield]

But one question remain about the test I propose to add: what is the proper way to test the instance of savecore from the installed build, i.e., not the one on the system? I find it convenient to make world DESTDIR=… but the tests don't seems to be aware of $DESTDIR.

@oshogbo Do you have an opinion on this? Do you want me to just ignore $DESTDIR?