Re: [PATCH 5/6] t1500: avoid setting environment variables outside of tests

[Eric Sunshine <sunshine@sunshineco.com>]

On Tue, May 10, 2016 at 2:39 PM, Jeff King <peff@peff.net> wrote:
> On Tue, May 10, 2016 at 01:20:54AM -0400, Eric Sunshine wrote:
>> diff --git a/t/t1500-rev-parse.sh b/t/t1500-rev-parse.sh
>> @@ -7,11 +7,13 @@ test_description='test git rev-parse'
>>       while :
>>       do
>>               case "$1" in
>>               -C) dir="-C $2"; shift; shift ;;
>>               -b) bare="$2"; shift; shift ;;
>> +             -g) env="GIT_DIR=$2; export GIT_DIR"; shift; shift ;;
>
> This will expand $2 inside $env, which is later eval'd. So funny things
> happen if there are spaces or metacharacters. It looks like you only use
> it with short relative paths ("../repo.git", etc), which is OK, but this
> would probably break badly if we ever used absolute paths.
>
> I don't know if it's worth worrying about or not. The usual solution is
> something like:
>
>   env_git_dir=$2
>   env='GIT_DIR=$env_git_dir; export GIT_DIR'
>   ...
>   eval "$env"

Makes sense; I wasn't quite happy with having $2 interpolated
unquoted. Like you, though, I don't know if it's worth worrying
about...

>> @@ -36,6 +38,8 @@ test_rev_parse () {
>>       do
>>               expect="$1"
>>               test_expect_success "$name: $o" '
>> +                     test_when_finished "sane_unset GIT_DIR" &&
>> +                     eval $env &&
>
> I was surprised not to see quoting around $env here, but it probably
> doesn't matter (I think it may affect how some whitespace is treated,
> but the contents of $env are pretty tame).

I flip-flopped on this one several times, quoting, and not quoting.
Documentation for 'eval' says:

    The args are read and concatenated together into a single
    command.

so, I ultimately left it unquoted, but don't feel strongly about it.

> This will set up the sane_unset regardless of whether $env does
> anything. Would it make more sense to stick the test_when_finished
> inside $env? You could use regular unset then, too, since you know the
> variable would be set.

I didn't worry about it too much because the end result is effectively
the same and, with all the 'case' arms being short one-liners, I think
the code is a bit easier to read as-is; bundling 'test_when_finished'
into the 'env' assignment line would probably require wrapping the
line. I do like the improved encapsulation of your suggestion but
don't otherwise feel strongly about it.

Nevertheless, I can re-roll with these changes if you feel more
strongly than I about it.