Make sure the internal split function get_corr_account_split
behaves consistently on multi-split transactions. The transaction
report depends on this.
Add test case to catch potential regressions
Simplify filter test function in transaction report.
The backends were using qof_backend_get_error() to test for sync errors.
This function clears the error, so the tests resulted in the error being
cleared before the session could see it and so it thought that the sync
had succeeded.
Replace those uses of qof_backend_get_error() with a new function
qof_backend_check_error() that doesn't clear the error.
Rewrite get_quote_time() to use the modern OO interface to Date::Manip.
This requires perl-5.12 or later, so require that. schemify_date() isn't
used anywhere, so remove that.
Turns out that there are changes as well as compression, particularly
if the timezone is different from the one from the last commit of the
gml2 files. Since there's no real benefit to saving the files most of
the time, don't, but leave the save line in, commented out, to make it
easy to write new files if necessary.
Gnc-fq-helper needs it to parse date and time strings from Finance::Quote.
The comment about using the iso-date field is removed because F::Q no
longer provides it.
New behaviour is that each time a bill is (re)posted
all the needed exchange rates will be asked.
If an exchange rate was already known, it is suggested
to the user, so s/he can simply confirm it if still ok.
The payment dialog was inconsistent in interpreting amounts and values.
By clicking on documents the document's amount is added to the Payment field
suggesting it's in the owner's currency. However the post-processing code
took that amount as being the amount in the transfer accounts currency
and calculated the rest backwards.
This will now become: the amount fields are interpreted in the owner's currency
and the exchange rate asked will be from owner's currenty to transfer
account's currency.
Work around a bug in libsecret. Under certain conditions
libsecret will silently fail to find a password even though
it is stored. The workaround uses another interface call
to force unlocking the secret store which works around the bug.
This workaround should be removed once a fix for
https://bugzilla.gnome.org/show_bug.cgi?id=748625
will be commonly available.
If libsecret is available use it to search for both
libsecret based passwords as gnome-keyring based ones.
This catches the situation where a password was stored
earlier using gnome-keyring, while the user is now trying
to retrieve it on a system that only has libsecret available.
This used to fail because gnucash depended on gnome-keyring
to be present in that situation.
Since some LDFLAGS settings depend on corresponding CFLAGS ones,
e.g. -fPIC, unsetting only CFLAGS will cause the test-compile to
fail because of the mismatched flags.
Only in some countries the IBAN is really restricted to numeric-only
(most notably in Germany, "DE"). In some other countries parts of the
IBAN may be alphas. These checks could be extended for more countries,
but since aqbanking with the SEPA transfers is used mostly in Germany,
we just leave it with the exception rule for Germany.
Fix configure.ac so that ${prefix}/lib/dbd is searched for modules.
Use gnc_path_get_libdir() to check for modules in the event that
$GNC_DBD_DIR is either empty or wrong and the path compiled into libdbi
itself is invalid. This applies mostly to Mac as other OSes tend to
leave their binaries where configure installs them.
