test: The test were leaving '2' files behind

[hfiguiere]

The syntax to output to stderr was incorrect

[GeorgesStavracas]

Hm. My bash redirection knowledge is quite limited, but according to this documentation page, redirecting stdout to stderr is documented as 1>&2. This branch changes it to >&2, without the 1.

Is this bashism? If yes, is it a harmless bashism, or something to avoid? Is 1 always the default?

[hfiguiere]

It was doing &>2 which redirect into 2 (a file name). >&2 mean stdout to stderr (the stdout is implicit).

[smcv]

FYI:

>&2 or equivalently 1>&2 is standardized POSIX sh syntax for "make fd 1 (stdout) into a duplicate of fd 2 (stderr)". The 1 does not need to be given explicitly, because > or >> defaults to acting on fd 1 unless otherwise stated.

&>2 is a bashism (i.e. documented by bash, but with no specified meaning in arbitrary POSIX shells). In bash, it means send both stdout and stderr to a file named 2. The POSIX equivalent of &>foo is >foo 2>&1, or equivalently and more explicitly 1>foo 2>&1.

Confusingly, bash also allows writing >foo 2>&1 as >&foo, but only if the filename foo is neither a number nor -, each of which means something different.