mbland / go-script-bash / 314
94%
master: 95%

LAST BUILD BRANCH: multiple-script-dirs
DEFAULT BRANCH: master
Ran
Jobs 1
Files 40
Run time 3s
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
314

push

travis-ci

mbland 
log: Work around bash array `export` bug (again)

While running the tests for `_GO_LOG_CONSOLE_FILTER` in the previous
commit (part of #36), the tests would pass under Bash 4.4.5(1)-release,
but not under Bash 3.2.57(1)-release.

Eventually I traced down the fact that under 3.2.57(1)-release,
`_GO_LOG_LEVELS` was getting erased sometime after the call to
`assert_log_equals` and before the call to `assert_log_file_equals`
causing `_@go.log_level_index` to fail for what should've been valid log
level names. I reasoned that the `.  "$_GO_USE_MODULES" 'log'`
invocation in `assert_log_equals` was causing `_GO_LOG_LEVELS` (and
likely other variables) to go out of scope after it returned, and since
`$_GO_USE_MODULES` ensures a module is imported only once, that variable
wasn't redefined—and there would've been errors with re-importing the
module anyway due to `readonly` variables, as mentioned in commit
9c2ae73f7.

Bash 4.4.5(1)-release didn't have this problem, and it started to ring
familiar. After scanninng the repo history, I realized this is my second
encounter with this bug, the first being documented in commit
b421c7382.

The bug is this: Declaring a global array and initializing it at the
same time triggers a bug in some versions of bash 4.x prior to ~4.25
whereby it won't be visible outside the function in which it's declared
after all. See:

- https://lists.gnu.org/archive/html/bug-bash/2012-06/msg00068.html
- ftp://ftp.gnu.org/gnu/bash/bash-4.2-patches/bash42-025
- http://lists.gnu.org/archive/html/help-bash/2012-03/msg00078.html

Though the above links mention `declare -g`, as did my original commit,
it appears that `export` effectively triggers the same condition.

This is not a bug Apple has fixed in their shipping 3.2.57(1)-release
version of Bash. This is contrary to what I said in the earlier commit
message, as I think at that point I hadn't run the tests locally as
`PATH="/bin:$P... (continued)

1784 of 1892 relevant lines covered (94.29%)

135.22 hits per line

Jobs
ID Job ID Ran Files Coverage
1 314.1 40
94.29
 Travis Job 314.1
Source Files on build 314
Coverage ∆ File Lines Relevant Covered Missed Hits/Line