From Owen:
I tried mktime on Darwin an Linux; looks like they both default to UTC if the TZ
environment variable is unset but on Solaris it defaults to the local timezone.
I compiled and ran the tests and checked the behaviour by running:
date
TZ="" date
TZ="UTC" date
On Solaris the first two are in local time and the third is in UTC.
On Darwin and Linux the first one is in local time and the other two are in UTC.
Found this as a reference, too, which mentions the suggested implementation in
'man 3 timegm':
http://stackoverflow.com/questions/6467844/is-c-mktime-different-on-windows-and-gnu-linux
Signed-off-by: Paul Beckingham <paul@beckingham.net>
- Duration were parsed as <int><unit> in A3, whereas <number><unit> is
correct.
- Added Nibbler::str method for internal access.
- Nibbler::save, ::restore now return positions. Useful for extracting
substrings.
- Modified countdown.t to use the (now) correct units for durations.
- When compiling the 2.0.0 branch of Taskwarrior with GCC 4.6.1, the compiler
generates a number of warnings. This patch eliminates these warnings.
Signed-off-by: Paul Beckingham <paul@beckingham.net>
- Eliminated the . prefix for DOM references to the current task - it is
unnatural to state ".due < today".
- Prioritized parsing dates ahead of integers, which were masking all
dates.
- Implemented Nibbler::getWord.
- Re-implemented Nibbler::getDOM.
- Modified DOM addressed for context-based attributes from "due" to ".due",
the help disambiguate DOM references in expressions. There is now a
consistency:
<id>.due Task-specific
.due Contextual
<uuid>.due General
- Implemented associated unit tests.
- Implemented RegX class to maintain a separate compile, and match
method, thereby allowing efficient re-use of the regex. This is
critical to Expression::eval, where an identical regex might be
applied to every task.
- Obsoleted rx.{h,cpp}, which combined the compile and match steps
into a single call, and is therefore not efficient when used in
the context of filtering.
- Fixed some unit tests that weren't building. Now they do. They
don't work of course (don't be silly) but that's a problem for
another day.
- Modified all code that relies on rx.h to use RegX.h.
- Reordered operator table so that longer operators match first, thus
disambiguating between ! and !=.
- Eliminated Expression::expand_expression.
- Modified Nibbler to know what a DOM reference looks like.
- Removed alpha equivalent operators (lt, le, gt, ge, not, eq, ne) because
these are common in descriptions (French: le, ne).
- Modified Arguments and Nibbler unit tests.
- Disallowed @ and / from an attribute name or modifier. This causes
input like 'user@host:path' to now be parsed as description, and not
as an unrecognized attribute.
- Migrated taskd JSON parser into task, to provide encode/decode
capability to Task::composeJSON.
- Migrated taskd utf8 code, replacing old unused code.
- Added unit tests or JSON.
- Migrated Tree updates from taskd.
- Applied patch to fix bug #581, in which backslashes in annotations and
descriptions caused problems (thanks to Itay Perl).
- Corrected unit tests that were shown to be incorrect, after the patch
corrected other behavior.
- Fixed bug #542, which sorted the countdown columns incorrectly (thanks to
Michelle Crane).
- Duration could not parse all varieties of it's own output, which is why the
sorting was broken, in particular negative durations and real numbers in
durations (such as (1.2y').
- Sorting in Table::sort_compare for Table::periodAscending and
Table::periodDescending was broken.
- Missing unit tests for countdown_compact.
- Imported Nibbler::getNumber from the Attitash project.
- Additional Duration unit tests.
- Additional Nibbler unit tests.
- Additional countdown.t unit tests.
- Nibbler: Now locally stores input length rather than repeatedly calling
std::string::length.
- Nibbler: Makes greater use of std::string::find_first_not_of, instead of
looping.
- TDB: No longer applies empty filters to lists - just copies the lists.
- TDB: Now caches data when reading completed.data.
- TDB: During loadPending and loadCompleted, makes fewer copies of the data.
- TDB: In commit, breaks out of search loops after finding the right data.
- TDB: In gc, only writes out minimal pending or completed data, instead of
all data, all the time.
- TDB: No longer reads completed.data in gc, and simply appends completed
and deleted tasks to it.
