Re: [Help-bash] errexit and attempting to change readonly functions

[Eric Blake]

On 9/27/19 3:43 PM, Greg Wooledge wrote:
> On Fri, Sep 27, 2019 at 09:28:24PM +0100, Stephane Chazelas wrote:
> > It seems as if bash tried to avoid triggering errexit when it
> > can get away (wrt posix conformance) with not to.
>
> The man page says,
>
>      Exit  immediately  if a pipeline (which may consist of a
>      single simple command), a list, or  a  compound  command
>      (see SHELL GRAMMAR above), exits with a non-zero status.

https://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html#set

"When this option is on, when any command fails... with the following 
exceptions:"

and does not list an exception for functions, and the grammar definition 
of command is:

https://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html#tag_18_10_02

command          : simple_command
                 | compound_command
                 | compound_command redirect_list
                 | function_definition
                 ;

Since function_definition _is_ a command (albeit neither a simple 
command nor compound command), failure to define a function SHOULD cause 
errexit behavior, when none of the other exceptions apply.

> I don't see function declarations included under "compound command"
> in that section.  So the question, for me, is whether a function
> declaration counts as a pipeline, a list, or a compound command (or
> none of the above).

A pipeline is a sequence of commands.

> I'm about 90% sure it doesn't count as a compound command, since
> the grammar for a function declaration includes "compound-command"
> right after the "()" part.  Not sure about pipeline/list.

Correct about it not being a compound command.  Incorrect about it not 
being [part of] a list (and thus part of a pipeline), as I can do:

condition && func() { :; }


--
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org