Testing notes: Based on the averages of 3 runs, the net
user CPU to save the XML file I use is:
10.2 seconds without this change
7.6 seconds with this change
In my environment the first call to the format routine
in question, the call that sets the cache value, is at
the end of the XML load.
The refactoring provides roughly 10% reduction in user CPU
use for XML file load by moving an expensive function
to within an if-clause where the result is used. The diff looks
like a full re-write but only the if statements, indenting,
and commentary changed.
This value is queried on each comparison of split or txn sort function,
which means it is called quite a lot. Avoiding the KVP lookup of this
property gains a lot in terms of CPU cycles.
when moving entry between accounts
When using the cut transaction option the 'associated file' value was
not being pasted to the new transaction. Added scheme code to get this
value and save it to new transaction when using 'cut/copy' and then
'paste' operations. When using the duplicate option, a dialogue allows
you to keep the copied association or not. It does not get copied for
autocomplete.
This involves renaming 3 functions:
gnc_uri_get_protocol -> gnc_uri_get_scheme
gnc_uri_is_known_protocol -> gnc_uri_is_known_scheme
gnc_uri_is_file_protocol -> gnc_uri_is_file_scheme
The *_protocol variants are marked as deprecated.
Additionally a number of local variables have been renamed from
protocol to scheme to support this change.
- gnc_uri_get_components will now return NULL as protocol if the input is a normal
file system path instead of a uri (it used to return 'file')
- gnc_uri_get_protocol will now return NULL if the input is a normal
file system path instead of a uri (it used to return 'file')
- gnc_uri_is_file_protocol now returns FALSE if protocol is NULL (it used to return TRUE)
- gnc_uri_is_file_uri now returns FALSE if input is a normal file
system path instead of a uri (it used to return TRUE)
- a new function gnc_uri_targets_local_fs will return TRUE only if its input
is either a file uri or a normal file system path. This function is now mostly
used instead of gnc_uri_is_file_uri in the current code base
- a new function gnc_uri_is_uri is added to check whether its input
is a valid uri (has protocol, path and hostname for non-file uris)
Previously the account color slot has been populated with "Not Set"
when any field for the account has been edited and saved. This routine
should run once and remove all such entries.
There are a very few left that need deeper study, but this gets
rid of most of the noise. For the most part it's just getting rid of
extra variables or removing an assignment that is always
replaced later but before any reads of the variable. A few are
discarded result variables.
It can't, because if b is 0 the function would have
returned already; since b.m_hi is 0 b.m_lo can't be. The assert
reassures clang that this is the case.
We need to compare the magnitudes of the remainder and the denominator
in order to round negative numbers correctly. Note that while gnc_numeric
is constrained to a positive denominator the C++ rounding functions cannot
assume that constraint in all cases.
Combined with the previous commit, this fixes
Bug 796949 - Incorrect conversion of 0,01 USD to EUR
So that the returned price tuple has the two commodities of interest
converted to a common currency. Before the first pair that that shared
any random currency would be returned, perhaps creating an absurd result.
This function complements gnc_time64_get_day_begin/end. There was
time64CanonicalDayTime but this returned noon of the given day, where we
want 10:59am in most cases. I haven't changed time64CanonicalDayTime
directly as that may break assumptions in other parts of the code.
Instead I have created a new function that can be gradually introduced.
date-posted to not be saved.
Check the stored GDate for being in the GncDateTime range as well
as the GDate range before returning it and check trans->date-posted
against INT64_MAX instead of 0 before changing it.
The first fix for this bug handled structs tm with ambiguous times.
This one fixes the GncDate constructor when the time is ambiguous
because it's in the DST-change hour, using the same add 3 hours,
construct the LDT, and subtract the 3 hours from the result.
The string constructor handles only simple-offset HH:MM timezones and so
is immune to the bug.
And don't ask to save a not-dirty or empty book, fixing
Bug 794870 - If no book is opened, gnucash still asks if the user wants
to save changes when opening a file
These are queried continuously by the owner tree view (on Customer/Vendor/Employee
Overview pages) and recalculating them is an expensive operation.
The cache will be invalidated each time a lot reated to the owner
changes (modify or delete). The net effect is a huge responsiveness
improvement of said overviews in case of a large book.
Any operation that can overflow will throw an underflow if it's a
negative number. The C interface needs to catch both to prevent
unhandled exception crashes,
The core issue was that the delete visitor was never called because its parameter
type (char *) didn't match the boost::variant type (const char *).
Fixing the visitor's parameter type also require a const_cast
back to char * because that's what g_free takes as argument.
The rest of this commit is merely fixing KvpValue instantiations that
tried to create a char* KvpValue from a stack based const string instead
of a heap allocated one. That would bomb out on calling the
delete visitor.
Splits were not marked for deletion if the transaction is read-only
and the account is not marked for deletion yet. The net result is
that split will not be freed later on.
However xaccSplitDestroy is also called from a Transaction's do_destroy.
At that point accounts are not necessarily marked for deletion yet (like
is the case when a datafile is closed). This turned out to be a problem
for invoice post transactions (which are also read only) and hence
would cause memory to leak.
Casting a char* to a struct containing a uint32_t is not universally safe
due to alignment constraints on reads on some platforms. Copy our possibly
unaligned source data into an aligned area of memory to avoid SIGBUS on
armhf.
Reported by vorlonofportland in PR#403. This commit the John's optimized
version of Vorlon's proposed fix.
- add -Wno-deprecated-declarations to CXX_FLAGS as well. This was
reported by vorlonofportland in PR#401 to become necessary for glib 2.58
as that has deprecated g_type_class_add_private which appears in our
c++ code.
- change -Wno-deprecated-register into -Wregister. The former appeared to
be a clang dialect and alias for the latter (see
https://github.com/Barro/compiler-warnings for an overview of clang
and gcc warnings). It was moved to global CXX_FLAGS as it can only be
added for g++.
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.
John Ralls
Christopher D. Carson
Christian Stimming
Christopher Lam
Robert Fewell
Geert Janssens
YOSHINO Yoshihito
Christoph Holtermann
Ruben Cheng
Di Mang