RFR: 8227493: Return a more useful error message from lookupAllHostAddr if getaddrinfo results in EAI_SYSTEM error [v2]

[Daniel Jeliński]

On Mon, 2 Dec 2024 16:53:32 GMT, Daniel Fuchs <dfuchs at openjdk.org> wrote:

>> You are right about + 3 instead of + 2. Furthermore, now that you mention it, I'm not sure why there's a `strlen(format)` in that size computation. The original change was introduced in https://hg.openjdk.org/jdk7/jdk7/jdk/rev/b5d37597c815 and at that time it was using `sprintf` (and not `snprintf`) and the `sprintf` as per its documentation, considers size to be `INT_MAX + 1`, so `size` wasn't being passed to it and only used for `malloc()`. Even then, I don't understand why `strlen(format)` (or even + 2) was considered for the size.
>> 
>> I have updated this part of the PR, but I'm going to look at this size computation bit more tomorrow with a fresh mind.
>
> oh - yes - the `strlen(format)` is strange - maybe it allowed to accout for the various ": ", and the rest would be > 1 which would allow for the null terminator? But then + 2 wouldn't have been needed.

I'd keep the `strlen(format)` and remove the fixed `+2` part; using `strlen(format)` will allocate a bit more than needed, but will free us from the need to manually calculate the exact number of extra bytes required.

Other than that, LGTM.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/22484#discussion_r1925662409