Are limitations of select() hurting us?

[krlmlr]

@zozlak: In https://man7.org/linux/man-pages/man2/select.2.html I read:

WARNING: select() can monitor only file descriptors numbers that
are less than FD_SETSIZE (1024)—an unreasonably low limit for
many modern applications—and this limitation will not change.
All modern applications should instead use poll(2) or epoll(7),
which do not suffer this limitation.

Are we affected here?

[zozlak]

Long story short:

• in general we are affected
• but in practice we probably don't need to care

Why we have used select() in the first place?

I would guess the original reason for using select() was portability. epoll() is linux-exclusive and poll() is not available on Windows.

Does it affect us - details

The actual question is "is it safe to assume that the socket descriptor number will be always lower than 1024?".

Socket descriptors are allocated by the socket() and its man page says:

The file descriptor returned by a successful call will be the lowest-numbered file descriptor not currently open for the process.

Not bad as it means until our process has more than 1024 files and network sockets opened at the same time, we are on the safe side and the limit of 1024 seems pretty safe.

Moreover, the default limit of opened files for normal users in a few linux systems I checked (my laptop with ubuntu 22.04, server at work with centos 7) is also 1024 so the process would anyway fail trying to open more of them (which would cause database connection error as the socket() call made by the PQconnectdbParams() internals would fail).

That being said if someone has higher max opened files ulimit and opens more than 1024 files and sockets in one process, (s)he will be affected by the limit you described.