significantly
Modify the DEBUG and PINFO macros to return unless qof_log_check is
true. Replace almost all direct calls to g_debug and g_message with
DEBUG and PINFO respectively.
Track the highest logging level sent to qof_log_set_level to provide a
short-circuit return in qof_log_check. Remove setting GNC_MOD_TESTS to
QOF_LOG_DEBUG so that the short-circuit threshold isn't defeated by
always being DEBUG.
Net result: 33% improvement in xml load times.
Inspired by PR #983. Transaction sorting on num broke down if the
user had a non-numeric string or a number larger than an int in
the effective num field (might be split-action if the option is set).
The comparison first tries to use strtoull on the two strings and
compares the results. If they're both nonzero and different then the
numeric order is returned. If they're both nonzero but the same the
unconverted parts of each are passed to g_utf8_collate; if either is 0
then the whole strings are passed to g_utf8_collate. strtoull will
return 0 for a negative number.
For clarity. In so doing found the backend behavior a bit inconsistent
so it's modified to do what the enum values indicate.
In the course of changing the various calls I found some implementation
errors in the back end and corrected them.
There are more, but these are most common ones.
There are also a number of urls that don't behave well when https, so those are skipped
At some point I have also started marking non-working URLs as [DEAD LINK], though
that's not a full coverage.
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.
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 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.
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.