SC2034 misses variables exported with allexport/set -a

[brchrisman]

For bugs

• Rule Id (if any, e.g. SC1000): SC2034
• My shellcheck version (shellcheck --version or "online"): 0.7.2
• [ X ] The rule's wiki page does not already cover this (e.g. https://shellcheck.net/wiki/SC2086)
• [ X ] I tried on https://www.shellcheck.net/ and verified that this is still a problem on the latest commit

For new checks and feature suggestions

• [ ] https://www.shellcheck.net/ (i.e. the latest commit) currently gives no useful warnings about this
• [ ] I searched through https://github.com/koalaman/shellcheck/issues and didn't find anything related

Here's a snippet or screenshot that shows the problem:

[bchrisman (team/business_ps): scale-product] $ cat /tmp/a
#!/bin/bash
set -o allexport
FOO=bar
bash -c 'echo $FOO'
[bchrisman (team/business_ps): scale-product] $ bash /tmp/a
bar
[bchrisman (team/business_ps): scale-product] $ shellcheck /tmp/a

In /tmp/a line 3:
FOO=bar
^-^ SC2034: FOO appears unused. Verify use (or export if used externally).

For more information:
  https://www.shellcheck.net/wiki/SC2034 -- FOO appears unused. Verify use (o...
[bchrisman (team/business_ps): scale-product] $

Here's what shellcheck currently says:

^-^ SC2034: FOO appears unused. Verify use (or export if used externally).

Here's what I wanted or expected to see:

<no output, as though the line was export FOO=bar

I might enjoy diving into the code and seeing if I can provide a fix if there are hints at where to start.

[koalaman]

You're certainly right. It's might be a bit tricky to fix though.

ShellCheck's handling of set -e is naively to look for set -e or equivalent anywhere in the script, and assume it's enabled for the whole script. Obviously it's trivial to trick with false && set -e, but it's worked well enough in practice.

I see from real-world examples that set -a is much more commonly used in blocks, like set -a; source "$file"; set +a;, so disabling all unused variable warnings may be too much. There's currently no robust way of seeing which assignments do and don't fall into such a block either, but pragmatically probably 95%+ of cases can be handled just by looking at commands between a set -a and set +a in any given block of code.

[OJFord]

Duplicate of #1986