I'm not sure why Debian's GCC didn't catch this, but NetBSD's did.
Another reason to add CI checks for NetBSD, then, though I am not sure
how to do it safely through VMs(and I doubt Docker/Podman would help me
for weird architectures). Hm.
I need to make it so that NetBSD targets get their own
correctness/build checks in CI. I also had to do some hacks with
Cytoplasm, will make a PR for those soon.
(You may notice that the dates for these commits are off. I'm
currently writing them without an Internet connection, and the
device I'm writing this through doesn't have an RTC. Oops!)
Negociating stanza sizes seems to be real trouble. Also this code is now
`-fanalyzer'-safe!
Also kills the last GNUMakefile present. It wasn't even used...
If parsee were downloaded from sources other than git-clone(1) (e.g.: a
tarball), perror(3) fails and/or git(1) is not available on the system,
"git remote get-url origin" would have unexpected results or even return
a NULL pointer, which would cause undefined behaviour on
strchr(repo, ...);
In such conditions, a placeholder string, namely "N/A", is strdup(3)ed
to `repo`.
So far, collect_sources was called recursively when the entry name did
not match a pre-defined list. However, this assumption broke with files
such as `etc/media/README`, that were not in the list.
Therefore, a deterministic way to distinguish files from directories is
to rely on the S_ISDIR macro. This avoids the need for a pre-defined
list.
Never. Do. This. Never. It should never cross your mind. Doing so will
punish you. Basic computer science concepts will tell you it's O(n^2).
And it WILL actually matter. Never. Never. Never. Never. Never. Never.
NEVER FUCKING EVER DO THAT EVER AGAIN.
NEVER FUCKING EVER DO THAT EVER AGAIN.
NEVER FUCKING EVER DO THAT EVER AGAIN.
NEVER FUCKING EVER DO THAT EVER AGAIN.
NEVER FUCKING EVER DO THAT EVER AGAIN.
Anyways, also fixes licensing year. On est en octobre et il comprend
toujours pas qu'on est plus en 2023.
