Sunrise/Sunset seems to have some kind of very subtle bug

[tfabris]

During the 2pm hour every day, my log repeats this message every time CrowCam.sh runs:

CrowCam Controller - Problem obtaining sunrise/sunset from Google. Falling back to previously saved values: 7:07 AM/7:24 PM

However when I run the script from the Bash shell (regardless of what system I run it upon or what user I run it under) then the retrieval works.

(Note: This is after I already made the parsing fix which works around google changing the space from a regular space to a nonbreaking space).

Get to the bottom of this!

[tfabris]

Fixed in checkin c700694.

WEIRD solution. To fix this, I had to turn on extended regex (-E), and instead of searching for a single weird unicode character ("."), I had to search for 1-3 characters (".{1,3}"). This was only necessary when running under the Synology Task Scheduler; the problem didn't manifest itself when running on the Synology at the SSH shell prompt.

[tfabris]

Reopening this bug. The previous fix was a mere workaround. The actual fix is related to the linux "locale" variable, LC_ALL. Details in this EmpegBBS thread: https://empegbbs.com/ubbthreads.php/topics/374308 - For some reason, the locale is set correctly, to en_US.utf8, when running from the shell, but it's set to blank/null when running from the Synology Task Scheduler. The fix code looks something like this:

forceLocale="en_US.utf8"
# Force the value of the locale variable:
echo ""
echo "Previous Locale:   $LC_ALL"
echo "Forcing locale to: $forceLocale"
export LC_ALL=$forceLocale
echo "Current Locale:    $LC_ALL"
echo ""

Cause of bug:

• System variable "LC_ALL" is null when running under Synology Task Scheduler
• This causes grep commands to fail when grepping for unicode characters.
• GetSunriseSunsetTimeFromGoogle had this grep in it: grep -o -m 1 '[0-9][0-9]*:[0-9][0-9].[AP]M
• The "." in the grep is supposed to search for a single character, between the time digits and the "AM/PM" designator.
• In the google search results, that single character is either a space, or a unicode narrow nonbreaking space, 0x202f.
• When running at the shell, LC_ALL is correctly set to "en_US.utf8" and the grep works and finds the single unicode character in that position.
• When running under Synology Task Scheduler, LC_ALL is blank and the grep fails to find that character.

I had previously implemented a workaround which searched for 1-3 characters in that slot, but that's a bad grep which could potentially result in false positives. So I'm going to fix this properly, by forcing the locale, at the same time as I fix issue #84 and make additional improvements to the grep.

[tfabris]

Fixed in checkin 2f3ce99