John's fix for bug #713 changes the way basis cost are calculated.
The patch also fixes#712, which caused ledger to create automatic
Equity:Capital Gains that were not correct. Update the test cases
accordingly after verifying the new output.
When a cost was specified without a whitespace after the @ symbol,
as in @$5.01, this was incorrectly parsed as 5.01 (losing the
commodity) rather than $5.01.
Bug fix for #1050
Using the Y 2014 syntax works fine, but using
apply year 2014 resulted in the following error:
Error: Year is out of valid range: 1400..10000
since part of the given year string was chopped off.
These aren't actually failures, just issues with the way that Jenkins
works on my system (mainly because ~/.jenkins is a symlink pointing at
another volume).
This test appears to test the garbage input handling on ledger.
garbage-input.dat has been added to the test directory for use by this
test, allowing development with amount.h to proceed without tripping
this particular test inappropriately
Some commits were on the master branch but not yet on the next branch due to a
pull request to master after those two branches diverged. This merge
ensures that next ⊇ master.
For `-f /dev/stdin`, the `pathname` of the parsing context will be empty
as for any other streamed input.
`instance_t::include_directive` did not work as expected for `-f /dev/stdin`
and relative file names. One would expect them to be relative to the current
directory rather than `/dev`.
This will lead to `While parsing file ""` messages. This could be
adjusted to read `While parsing standard input`, but maybe it's not
worth the special cases.
This commit also fixes bug 788: behaviour of source_context for '-f -'
`session_t::read_data` did not set context.pathname to `/dev/stdin`
for the special case `-f -`. I chose to adjust `source_context` too
as there is no sensible context if no file name is provided.
ledger would abort with the following error message:
TypeError: No Python class registered for C++ class boost::optional<std::string>
The changes pass a CallPolicy to make_getter when adding the transaction
code property for python, so that the correct to_python conversion is
made. For details see:
http://www.boost.org/doc/libs/1_52_0/libs/python/doc/v2/faq.html#topythonconversionfailed
This was caused by both 'A' and 'bucket' using
default_account_directive. This function was still stripping the 'A'
directive, so the first character of the account name used with 'bucket'
was cut off. Maybe the code for the other directives should be changed
accordingly for consistency (put line + 1 in call instead of function).
test/regress/CAE63F5C-b.test and test/regress/CAE63F5C-c.test should both
pass, but test/regress/CAE63F5C-c.test does not, because the total line of
$6.46 is rounded wrong; it should be $6.45.
There seems to be different rounding occurring for totals vs. postings.
This seems to be related to Bug #492.
