The KVP value for the qof_book_get_num_days_autoreadonly was being
called many times so it makes sense to cache it in the book to avoid
the KVP lookup.
test_suite_gncInvoice sets up the test suite. It's not part of the test
runtime, so stack variables in it have gone out of scope by the time the
tests are actually run. Making invoiceData static makes it permanent so
it exists at runtime.
First, save isn't necessary if the book is dirty, so don't... but that
means that the book has to be marked dirty after a session swap. No more
laziness.
Second, regardless of the outcome of inner_main_add_price_quotes the
session must be destroyed to remove the lock.
A couple of cleanups in QofSessionImpl::save as well: Rewrote the
descriptive comment to reflect how it really works when the backend has
gotten disconnected and removed the superfluous qof_book_set_backend
with the backend that we'd *just gotten from the book*.
Change all instances of bugzilla.gnome.org to bugs.gnucash.org, reflecting
our migration to a self-hosted bug tracker.
Inform the Translation Project Coordinator at release that this affects
translatable strings and that all message catalogs have been updated.
I think this crash is triggered because the 'account' variable
defaults to the first available AR account. If there's no AR account
it becomes null, and querying null's default book leads to segfault.
I guess I can fix segfault too by fixing gnc_account_get_book.
Change some plain string literals to std::string constants, which helps
avoiding typos and also saves some string constructors/destructors
in the KVP lookup. Nevertheless the functions in Account.cpp do not
contribute that much to the overall UI speed, but whatever.
The function qof_book_use_split_action_for_num_field gets called quite a
lot in each register display refresh (due to sorting all splits from
Split.x's xaccSplitOrder function), but it always used to use a KVP
lookup, which is rather expensive compared to accessing a gboolean member
variable.
To get rid of this cost, I had to remove the KVP lookup in this
simple-looking function. The pattern is this: A gboolean cache variable is
introduced, along with an isvalid flag. The lookup makes the expensive
KVP lookup once, then caches the value. The GObject property mechanism
offers a callback for when the setter was called, which is used to mark
the cached value as invalid. A parallel setter method (here:
qof_book_set_option) also just marks the cache as invalid. This covers
all setters, and the getters will use the cached value except for their
first invocation.
The NUM_FIELD_SOURCE feature was introduced in 2012 by the very large
commit 7cdd7372 and apparently its costs never were a problem
until the KVP lookup became more costly due to the std::vector
construction and destruction.
Turns out that the on-the-fly conversion from const char* (the KVP_OPTION_PATH
constants) to std::string with their immediate deletion afterwards is
a quite costly operation. Avoiding this is surprisingly easy: Just keep
local std::string objects at hand, and they don't have to be created
and deleted anymore.
The more optimized solution might be to turn the std::vector<std::string>
into a std::vector<GQuark>, but this commit at least improves the picture for now.
First change is to ensure gncEntry rounding is consistent. Internally
calculated values in the entry are never rounded. Consumers of
gncEntry's calculated values can request them either rounded or not.
Next use a pragmatical approach for calculating values on invoices based on
the entry values: do the rounding such that we never
create an unbalanced transaction while posting
That means
- round each entry's net value before summing them in net total
- accumulate all tax totals on invoice level per tax account before rounding
and round before before summing them in a global tax total
Hopefully this will catch a few more rounding issues in this area.
A complete solution can only offered if we allow users to manually correct
tax entries. This requires changes to user interface and data format
so that's not going to happen in gnucash 3.x.
strptime/strftime supports various modifiers to their parameters.
'E' and 'O': alternate locale-specific formats
(used in default format for Persian, Oriya, Azerbaijani)
'-': padding
(used in default format for Czech)
GnuCash passes dates as integer y/m/d without using locale-specific
formats, so we need to strip out 'E' and 'O' from the format when
scanning dates or determining separators in gnc-date.
None of '-', 'E', or 'O' are supported by boost (and '-' causes
errors), so strip them out from formatters in gnc-datetime as well.
See https://bugzilla.gnome.org/show_bug.cgi?id=795247.
This will probably need more refinement because the multiplications
are still missing rounding methods, but the changes in this commit
will already allow guile code to correctly create entries.
This commit contains another round of cleanups in the
timespec to time64 conversion. There were a number of
false assumptions that time64 = 0 would be a bad date
in the xml parser. This commit corrects enough of them to
eliminate the bug. Further cleanup is probably advised but
can be done at a later stage.
Also Bug 791825 - Accounting Period dates off by 1.
The DST start/end dates were reversed *and* the DST offset had the wrong
sign in Windows, resulting in the effective timezone always being one to
the west off (i.e. PDT was -9 and PST was -8).
Bills and invoices that are posted and subsequently unposted again still store their
posted account internally as a convenience to the user (upon reposting the old
account will be offered by default) so it's not a reliable test for the posted state.
The posted transaction on the other hand is guaranteed to only exist when the invoice is
posted. This should fix a slew of small and perhaps larger side effects, such as
a posted bill still appearing as editable, critical warnings when creating new bills/invoices
and so on.
An odd corner case: BST apparently came off of DST at 23:00 26 Oct 2014,
so midnight that day was ambiguous about being DST or not; that causes
the local_date_time constructor to throw in spite of the tm.is_dst element
being 0 (meaning pick standard time).
Instead of just failing in that case, try constructing a local_date_time
three hours later then adjust it back three hours. If *that* doesn't work
then throw a std::invalid argument.
After much thrashing this turned out to be caused by a date string
with a 3-digit year and that caused an unexpected boost::bad_cast
exception from boost::posix_time::time_from_string().
To prevent that and anything like it, pre-parse the string with
regular expressions to classify them and split out the timezone
if there is one. If neither (perhaps eventually none) of the
regexes match throw std::invalid_argument. The C function will
catch this and return 0.
91f4b19 changed the test for gncInvoiceDateExists from date != 0
to date != INT64_MAX, which isn't backwards compatible, so test
for both. But the submitted file had a posted date of -1 so
gncInvoiceIsPosted returned true anyway. That's not consistent with
the logic in dialog-invoice.c, which checked gncInvoicePostedAcc != NULL,
a better test. The result was that the "Post" button lit up but
gncInvoicePost returned immediately, doing nothing, so change
gncInvoiceIsPosted to use gncInvoicePostedAcc instead.
There were several problems that broke the Imap Editor that have been
fixed due to kvp changes. The import-map-bayes entries were being added
to the tree view based on the number token entries squared. Retrieving
import-map entries resulted in an empty list and also deleting entries
from the tree view failed.
Found with -Wimplicit-fallthrough, but older versions of GCC don't support it. So it will take some effort to add it to the build config.
I think it did not result in a reproducible bug.
Turns out the stream facet parser is really slow. Since we have a
well-constrained universe of input formats we don't really need that
overhead. This got a 33% improvement in loading a large SQLite database.
This commit fixes business reports caused by 91f4b19039
whereby the posted-date for unposted invoice was changed
from 0 to MAXINT. Now we need to use (gncInvoiceIsPosted)
or (gncInvoiceDateExists) instead.
When building from git it will add targets to generate the swig files.
When building from tarball it will just point at the generated source
files from the tarball.
Several were only there because they used to be generated via autogen.sh
and hence had to be included because autogen.sh was not supposed to be run
in a tarball based build.
A few others aren't clear so I have left them in for now:
- gnucash.1 (man page)
- gnucash-design.info (because it's unclear what we want to do with that one)
These files should clearly be in the tarball:
- gnucash.pot -> for our translators
- ChangeLog -> can't be generated outside of a git repo
- gnc-vcs-info.h -> can't be generated outside of a git repo
- guile/python bindings -> to avoid a swig dependency when building from tarball
- the two dist_add_... macros now both take a list of file names
as argument so more files can be added at once to the dist tarball.
- dist_add_generated now creates the right target by itself. There's
no need to pass one any more
- make the swig generated *.py module files explicit output files
- change a couple of custom_targets into custom_commands. The only
reason they were defined as targets was to ensure they got built
before the dist tarball. This is now properly handled by the
dist_add_... macros.
- correctly handle dependency on swig-runtime.h (using OBJECT_DEPENDS
was not the way to do it according to that property's help page)
cmake with unix makefiles fails to resolve dist dependencies
added from COPY_FROM_BUILD if these dependencies aren't built yet.
This commit replaces the COPY_FROM_BUILD based logic with two new functions
'dist_add_configured' and 'dist_add_generated' to indicate which files should
be included in the dist tarball. The latter also adds a target level dependency
to the dist tarball custom command. Hence the former should
be used for files that get generated during a cmake run while the latter
should be used for files generated as the result of a 'make/ninja-build' run
(like files for which an add_custom_command rule exists).
Note: this commit also temporarily disables the dist target when building
from a tarball (and hence it won't be tested in distcheck either). This
will be handled in a future commit.
Conflicts:
- gnucash/gnome-utils/gnc-main-window.c
I have chosen to ignore the changes pulled in from maint. The same change will
be reimplemented in a followup commit.
- po/ru.po
Same here. The translation conflicted too much and there's a
translation update available for unstable I will pull in soon.
Resolved conflicts:
gnucash/gnome-search/dialog-search.c
gnucash/gnome-utils/gnc-splash.c
gnucash/import-export/aqb/gnc-plugin-aqbanking.c
gnucash/import-export/import-account-matcher.c
po/de.po
src/report/standard-reports/budget-barchart.scm
src/report/standard-reports/budget.scm
Note the de.po conflict was resolved by dropping all changes
and merging the file again with gnucash.pot (generated after
the merge but before committing).
GLib's scribbling of freed memory is enabled on Arch so attempting to
read the deleted members of inst and book crash instead of reaturning
invalid results. These weren't really useful tests anyway.
The first attempt to fix this, a17bc85a, doesn't work because the
boost::date_time constructor gets enough information in most cases to
generate a date, just not the one we expect. This change looks for '-' in
the fourth position and if it's there assumes iso-extended format, otherwise
it assumes delimiter-less ISO without the 'T', i.e. %Y%m%d%H%M%S.
Set a feature to prevent versions older that 2.6.20 from loading a
database from which they cannot read the dates. Ideally we would do this
only if the database is written to, but the current persistence design
includes committing back to the database during the load so the net
effect is that the flag would be set anyway.
This allows direct conversion between Scheme numbers and gnc_numeric
without the performance or accuracy penalties arising from using doubles
as an intermediary.
This commit introduces a new feature flag:
GNC_FEATURE_GUID_FLAT_BAYESIAN. It signifies that the bayes import map
data are stored flat and by guid. Any time bayes import map data are
accessed, they are converted if necessary.
We need to allocate enough space for the terminating null character.
Also, I double-checked the documents for std::basic_string::c_str () and
verified that it does guarantee the terminating null, so I put a comment
in the code that depends on that.
Before, it was necessary to remove '/' from tokens so that they won't be
divided up within kvp. Now that kvp doesn't parse tokens, it's okay to
pass '/', and it's better not to translate user-provided tokens if at all
possible.
The conversion assumed there were only three levels to bayes import
map kvp: IMAP token, user-supplied token, GUID/account name. In
actuality, since user-supplied tokens could have the delimiter in them,
there could be several. This fix takes that into account like so:
IMAP token, potentially several user-supplied tokens, GUID/account name.
The import map is undergoing two conversions at the same time: account names
to guids and an hierarchical representation to a flat representation in KVP.
The bayes data are stored in the KVP store. Before this commit, they are
stored under /import-map-bayes/<token>/<account guid>/count (where count
is the datum that "matters" in bayes matching).
The problem with this is that any token including the kvp delimiter
(currently '/') gets divided, and is not found correctly during bayes
kvp searching. The quickest solution to this is to replace all "/"
characters with some other character. That has been done, along with a
re-structuring of the bayes matching code to take advantage of c++
features to make the code more concise and readable.
Also modified some test functions to fix leaks and double-frees: the
same kvp value can't be in the kvp tree twice.
Also, when I added code to clean up after the tests, some things started
breaking due to double-delete. Apparently const_cast was hiding some
programming errors. Really? You don't say? When giving a GUID* to KvpValue,
the latter takes ownership of the former.
The nested representation was very noisy. Now, the string representation
shows one line per value with the full prefix which is also more
expressive than the old version.
The template avoids the need to cast to and from void*, and adds flexibility to
the targeted function's signature.
test-stuff.h defines a macro, "failure" which is used as an identifier
in the standard IO library, so I moved any inclusion of test-stuff.h to
the last include position so that "failure" wouldn't be defined before
the IO library was included.
Since Account.c is now Account.cpp, the function signatures look a bit
different internally. The tests rely on function signatures in error
messages. Instead of trying to figure out what the exact
function signature might be, I use a substring matching strategy to
ensure that the correct error was issued.
Sets paths for finding componenents depending on the state of ENABLE_BINRELOC,
GNC_UNINSTALLED, GNC_BUILDDIR and whether any install paths have been set
outside of CMAKE_INSTALL_PREFIX.
GNUInstallDirs changes the name of CMAKE_INSTALL_LIBDIR depending on the
operating system and distro. When CMAKE_INSTALL_PREFIX is /usr,
/usr/local, or any subdirectory of /opt it also changes
CMAKE_INSTALL_FULL_SYSCONFDIR to /etc. An earlier commit by Aaron Laws
mirrors the name of CMAKE_INSTALL_LIBDIR to the build library directory.
It's possible for builders to set any of the install directories
anywhere they please.
Setting any directory outside of CMAKE_INSTALL_PREFIX breaks Binreloc so
the toplevel CMakeLists.txt now detects that and disables Binreloc.
If Binreloc is enabled then all path queries use it to find paths. This
works in the build directory because the gnucash executable and all of
the test programs are in build_directory/bin and LIBDIR, DATADIR, and
SYSCONFDIR can be found in the same root path.
If Binreloc is disabled then in order to build or run programs from the
build directory one must set GNC_UNINSTALLED and set GNC_BUILDDIR to the
absolute path of the build directory. When those are set GNC_BUILDDIR
replaces CMAKE_INSTALL_PREFIX in all paths that are subdirectories of
CMAKE_INSTALL_PREFIX; paths that are not in CMAKE_INSTALL_PREFIX are
appended whole to GNC_BUILDDIR. This process is constent between CMake
and gnc_path_get_foo. GnuCash is unlikely to run from a DESTDIR without
Binreloc.
I.e., remove the shell invocation and with it the need to set the shebang.
Surprisingly this required some build-system modifications particularly
for cmake in order to correctly set the environment.
There seems to be more than one problem that causes the exception handler ("catch") to
get lost on Windows:
* Throwing from a constructor called from a member function of another object of the same
class. That's fixed here for the GncNumeric string constructor, but there's at least one other
instance I'm still working on in GncNumeric::to_decimal.
* Hidden memory allocation in a stack-allocated object like std::string, std::istringstream,
or boost::smatch: The throw causes the object to go out of scope which calls its destructor
and in that case the catch reference is either lost or never compiled in.
This change ifdefs out the creation of detailed exception messages on Windows to avoid
the destruction of the std::istringstream and its attached std::string, creates a series of
helper functions to ensure that the boost::smatch is in a non-throwing scope, and puts the
computed values directly into the member variables instead of delegating the construction
to a temporary and then copying out the values. The last item is more correct anyway, as
C++ constructor delegation is supposed to happen in the member initialization part rather
than the function body.
With these changes the exceptions from the GncNumeric string constructor are handled
correctly.
This will avoid a ninja-build from picking up a config.h generated by the autotools build
(in the root build directory). Picking up the wrong config.h may lead to all kinds of
subtle issues if the autotools run was done with different options than the cmake run.
The reconcile account dialog always displays a value of 0.00 as the
Ending Balance, regardless of account and statement date.
This is caused by function xaccAccountGetReconcilePostponeBalance
returning the wrong value, returning TRUE when it should return FALSE,
and setting balance to the default {0,1}.
The code uses bal.denom!=0 as an indicator that a valid balance was
received in variable v. However, bal is initialized to {0,1}
making the test always true even when we didn't receive a valid
value in variable v.
Thus, this function returns TRUE with *balance={0,1} when no valid
balance was found in "reconcile-info/postpone/balance".
This patch fixes the function to return FALSE if v doesn't hold
a valid value or if bal.denom is set to 0.
gtk-mac-bundler can't access the executable's rpath list so it can't
follow dependencies if they're not in $install_dir/lib from @rpath.
Autotools always sets absolute path install names so this should have no
adverse affects on other Mac builds.
Setting a feature writes to KVP which in the SQL backend causes the book
to be written out, which is rather expensive. Doing this in a loop (as
is the case in convert_imap_account, is quite slow, and this change
afforded a 4x speedup.
There are now options to specify which commodities to remove prices
from as well as options to leave none or the last one for the week,
month, Quarter, year and a scaled option.
cmake:
- add test-app-utils
- rename test-link-module to test-link-module-app-utils
- add gtest-import-map
autotools:
- move gtest-import-map from TEST_PROGS to TESTS (autotools) so it shows up in the colored results list
It is split into
- /libgnucash (for the non-gui bits)
- /gnucash (for the gui)
- /common (misc source files used by both)
- /bindings (currently only holds python bindings)
This is the first step in restructuring the code. It will need much
more fine tuning later on.
Geert Janssens
John Ralls
Robert Fewell
Christopher Lam
Christian Stimming
Paul Wassi
Bill Nottingham
John Ralls
fell
Niels
Andreas Müller
Rob Gowin
lmat
Carwyn Nelson
Jose Marino
Mike Alexander