* Add deprecation warnings to recently deprecated procs
* Fix bad usage of the times module
* Introduce sub second resolution
* Fix usage of C's time()
* Switch to nanosecond resolution
* Make Time & Duration opaque again and fix some errors
* Change back to TimeInterval for shorthands
* Fix JS test
* Fix build error for windows
* Undeprecate epochTime
* Documentation and minor changes
* Lots of bugfixes and doc comments
* Attempt to make travis & appveyor green
* Fix edge cases for dealing with the local timezone
* Workaround JS backend overflow/underflow bug
* Use better workaround for not knowing the size of time_t
* Use all available timezones for tests
* Fix indentation
* Add procs for accessing the fractional part of a duration
* Order time units from smallest to largest since it makes more sense
* Include months and years in `TimeUnit`
* Review fix
We accept a new environment variable, NIMTEST_COLOR,
which override the effect of NIMTEST_NO_COLOR.
The environment variable, NIMTEST_COLOR, can be 'never'
or 'always', which set the color output to false or true,
respectively.
Under Linux (probably POSIX), the current code tests for timeout, but
does not test for connection failure. connectAsync() returns succesfully
upon an EINPROGRESS result; but at this point, the connection state is
still unknown. After selectWrite() is done, we need to test the socket
for errors again.
Following a quesiton on Gitter about loading libraries dynamically based upon some runtime parameter, I noticed there's no documentation on the `dynlib` module for something that's quite common.
* Add ability for users to elide ':' or '=' when CLI authors pass a
non-empty partial symbol table. Behavior should be identical to the
old behavior if empty partial symbol tables are passed. "Partialness"
of the symbol table refers to the fact that one need only specify
option keys that are toggles/booleans/do not take arguments, hence
the "NoArg" suffixes in shortNoArg and longNoArg.
commandLineParams() returns seq[TaintedString], so use that consistently
in getopt() and initOptParser(seq[TaintedString]) dropping the taint at
the quoting stage just as with the paramStr() logic.
Fix capitalization inconsistency of cmdLongOption.
Export OptParser.cmd and OptParser.pos so that, at least *in principle*,
users of this API can handle "--" option processing termination or some
"git-like" sub-command stop word with a separate option sub-syntax.
{ Eg., ``case p.key of "": echo "trailing non-option args: ", p.cmd[p.pos..^1]``
or ``case p.kind of cmdArgument: if p.key == "mysubcmd": ...``. } Really,
searching for the last delimiter before p.pos is probably needed to frame
the trailing text..Not the nicest API, but still possible with effort.
* Make requested changes from string to seq[char]
(see https://github.com/nim-lang/Nim/pull/7297)
* Document new behavior and elaborate on some special cases.
* NoArg => NoVal to be less ambiguous/more clear.
* Add more documentation and an example snippet.
* Tweak language. Clarify still using ':'/'=' is ok.
* Add a test case for new NoVal behavior.
