1347175 - Request Nagger not sending emails due to DST change and impossible date math

Status: RESOLVED FIXED

(bugzilla.mozilla.org :: Extensions, defect, P1)

Description

[Dylan Hardison [:dylan] (he/him)]

Request nags seems to be having a problem.

Comment 1

[Dylan Hardison [:dylan] (he/him)]

I think this actually an error rather than a warning.

Comment 2

[Dylan Hardison [:dylan] (he/him)]

I think this relates to setting the TZ variable to begin with a colon.

Comment 3

[Dylan Hardison [:dylan] (he/him)]

I'm sorry, this was not obviously a code issue to me, I assumed it was infra and moved on. After running strace() I realized "the signal is coming from inside the house".

Another meta-problem of this is that currently the script is configured to just give raw error messages without context,
so we'll want to fix that.

I'm able to get this to give a stack trace by using Carp::Always (which should be the default for cron scripts):

Invalid local time for date in time zone: America/Los_Angeles
 at local/lib/perl5/DateTime/TimeZone.pm line 239.
	DateTime::TimeZone::_span_for_datetime(DateTime::TimeZone::America::Los_Angeles=HASH(0x6641d78), "local", DateTime=HASH(0x6c23fb8)) called at local/lib/perl5/DateTime/TimeZone.pm line 196
	DateTime::TimeZone::offset_for_local_datetime(DateTime::TimeZone::America::Los_Angeles=HASH(0x6641d78), DateTime=HASH(0x6c23fb8)) called at local/lib/perl5/x86_64-linux-thread-multi/DateTime.pm line 934
	DateTime::_offset_for_local_datetime(DateTime=HASH(0x6c23fb8)) called at local/lib/perl5/x86_64-linux-thread-multi/DateTime.pm line 379
	DateTime::_calc_utc_rd(DateTime=HASH(0x6c23fb8)) called at local/lib/perl5/x86_64-linux-thread-multi/DateTime.pm line 1728
	DateTime::add_duration(undef, DateTime::Duration=HASH(0x6c0e9d0)) called at local/lib/perl5/x86_64-linux-thread-multi/DateTime.pm line 1635
	DateTime::add(DateTime=HASH(0x6c23fb8), "days", 1) called at ./extensions/RequestNagger/bin/send-request-nags.pl line 186
	main::_include_request(HASH(0x6794798), "requestee") called at ./extensions/RequestNagger/bin/send-request-nags.pl line 84
	main::send_nags("reports", ARRAY(0x6641f88), "requestee_sql", "\x{a}    SELECT\x{a}        flagtypes.name AS flag_type,\x{a}        flag"..., "setter_sql", "\x{a}    SELECT\x{a}        flagtypes.name AS flag_type,\x{a}        flag"..., "template", "user", "date", ...) called at ./extensions/RequestNagger/bin/send-request-nags.pl line 55

Comment 4

[Dylan Hardison [:dylan] (he/him)]

Attachment: 1347175_3.patch

This is less than ideal because we're shifting the definition of "weekend" to be UTC-weekend. I think this better than the other immediate alternative of reverting bug 984438.

Comment 5

[:glob ✱]

Comment on attachment 8858069 [details] [diff] [review]
1347175_3.patch

Review of attachment 8858069 [details] [diff] [review]:
-----------------------------------------------------------------

r=glob

::: extensions/RequestNagger/bin/send-request-nags.pl
@@ +156,4 @@
>  
>  sub _include_request {
>      my ($request, $report) = @_;
> +    state $now = datetime_from($db_date, 'UTC')->truncate( to => 'day' );

you probably want
state $now //= datetime_from(..)

Comment 6

[Dylan Hardison [:dylan] (he/him)]

(In reply to Byron Jones ‹:glob› from comment #5)
> Comment on attachment 8858069 [details] [diff] [review]
> 1347175_3.patch
> 
> Review of attachment 8858069 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> r=glob
> 
> ::: extensions/RequestNagger/bin/send-request-nags.pl
> @@ +156,4 @@
> >  
> >  sub _include_request {
> >      my ($request, $report) = @_;
> > +    state $now = datetime_from($db_date, 'UTC')->truncate( to => 'day' );
> 
> you probably want
> state $now //= datetime_from(..)

I don't think so:

perl -E 'while (1) { state $foo = say "once" }'

Comment 7

[Dylan Hardison [:dylan] (he/him)]

To [e-mail]:mozilla-bteam/bmo.git
   cfded4d..f04f94b  master -> master