mjhjmtu

The bash manual states:

eval [arg ...]

 The args are read and concatenated together into a single com-
 mand. This command is then read and executed by the shell, and
 its exit status is returned as the value of eval. If there are
 no args, or only null arguments, eval returns 0.

I try

eval `nonsense`
echo $?

The result is 0.

Whereas when I execute the back-quoted command separately:

`nonsense`
echo $?

The result is 127.

From what is written in the bash manual I would expect eval to return 127 when taking the back-quoted nonsense as argument.

How to obtain the exit status of the argument of eval?

When you do the following -

`nonsense`
echo $?

You basically are asking "Tell me the exit status when I try to get the output of the command nonsense"
the answer to that is "command not found" or 127

But when you do the following

eval `nonsense`
echo $?

You are asking "tell me the exit status of eval when I evaluate an empty string" (the output of command nonsense) which is equal to running eval without arguments.

eval has no problems in running without arguments and its exit status becomes 0

Actually, it's more the:

$ `nonsense`
bash: nonsense: command not found
$ echo "$?"
127

That is surprising here.

We're asking bash to run the command that results of the split+glob operator on the stdout of nonsense. Since nonsense produces no output, it runs no command, so you may think the exit status should be 0.

But actually, when a simple command line has no argument, only assignment or redirection, the exit status is that of the last command substitution in assignment and normal words (not in redirection targets) that was run (though failure in redirections will also affect the exit status).

That's specially useful with assignments.

In:

output=$(grep pattern file)
status=$?

You can get both the output and exit status of grep, which you couldn't do if $? was otherwise the exit status of that non-command.

In:

output=$(cmd1) cmd2

That is where there are both assignment words and argument words, the exit status of cmd1 is ignored. $? will contain the exit status of cmd2.

And, also $output will only be set for cmd2 only. Exception to that is when cmd2 is a special builtin.

eval is such a special builtin.

$ a=0; a=1 eval; echo "$a"
1

In bash and most modern POSIX shells.

a=`exit 5` eval; echo "$?"

or

eval `exit 5`; echo "$?"

Would output 0, as it's the result of running eval with no argument. But that was not the case in the Bourne shell or ksh88, where for special builtins you'd get the exit status of exit 5 there.

In those shells, you'll also find:

$ a=`exit 3` set x; echo "$?"
3

As set is another special builtin.

. is another special builtin. In the Bourne shell and ksh88:

$ . /some/file `exit 4`; echo "$?"
4

(as long as /some/file doesn't run any command)