[Bug time/33971] New: mktime()/localtime() sets tm.tm_{gmtoff,zone} according to TZDEFRULES when falling back to it

[me at ziyao dot cc]

https://sourceware.org/bugzilla/show_bug.cgi?id=33971

            Bug ID: 33971
           Summary: mktime()/localtime() sets tm.tm_{gmtoff,zone}
                    according to TZDEFRULES when falling back to it
           Product: glibc
           Version: unspecified
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: time
          Assignee: unassigned at sourceware dot org
          Reporter: me at ziyao dot cc
  Target Milestone: ---

Created attachment 16648
  --> https://sourceware.org/bugzilla/attachment.cgi?id=16648&action=edit
testing program

On a system where /usr/share/zoneinfo/posixrules is present and identical to
EST5EDT provided by tzcode, the attached test program results in weird output,

Offset from UTC time in minutes: -300
Timezone name: EST
Offset from UTC time in minutes: -300
Timezone name: EST
Offset from UTC time in minutes: -300
Timezone name: EST

while TZ is setting to "IST-2IDT". With posixrules file removed, the same
program gives out expected output,

Offset from UTC time in minutes: 120
Timezone name: IST
Offset from UTC time in minutes: 120
Timezone name: IST
Offset from UTC time in minutes: 120
Timezone name: IST

which suggests glibc to wrongly take the zone name and time offset from
TZDEFRULES.

The calculation of localtime() is done in __tzfile_compute(), and when the
epoch is after the latest known transition, __tzset_parse_tz() is called to
fill the tm structure, including tm_gmtoff and tm_name, based on tzspec,
causing the issue since tzspec is derived from TZDEFFILES.

mktime() internally invokes localtime_r(), thus is affected by the same
problem.

-- 
You are receiving this mail because:
You are on the CC list for the bug.