I was seeing multiple failures in src/engine/test/test-date when running "make check".  I tracked this down to the usage of mktime() in gnc_iso8601_to_timespec_gmt and gnc_timespec_to_iso8601_buff.  When the timespec is (or would be) very close to the epoch, mktime() will fail if the timezone is east of UTC.

For example, consider gnc_iso8601_to_timespec_gmt where the resulting timespec should be zero.  Because mktime actually returns local time, if the timezone is east of UTC, the local time time_t value corresponding to a zero timespec would actually be negative.  mktime cannot handle this and errors out with the value of -1, causing the conversion to fail.

src/engine/test/test-date.c has multiple tests that specify/require timespecs at or very close to zero.  Lines 93-134 explicitly test strings that should result in a zero timespec, which all fail on systems with timezones east of UTC.  In addition, the loops in lines 397-402 and 404-418 can also result in timespecs that will fail.  Fixing the loops is easy - simply skip any iterations in which the timespec falls in the range [0, max(0, -gnc_timezone()) ).  The zero timespec tests cannot be fixed without either reimplementing gnc_iso8601_to_timespec_gmt to not depend on mktime, or changing the test to use a timespec at least twelve hours later than the zero timespec.  They could probably also be safely removed since there are plenty of other tests using non-failing timespecs to make sure the conversions happen properly.  It is unlikely that the zero timespec tests will catch any errors that the other tests will not.