So we avoid lots of memory allocation.
We can work with char* directly since both the pattern and the file
name are in UTF-8 and there is no need to understand unicode for
such pattern.
(In fact, '?' would not match anyore non-ascii characters, but I
don't think that's a problem. I don't think anyone use '?' in its
exclude list. And the two allocations per call to csync_fnmatch are
really worth getting rid of)
This function only checks the full path and the basename and is thus
around 7x faster. It is very useful in a csync_update context where
we know that the leading dirs have already been checked for exclusion.
Before we only had c_utf8_to_locale, but now functionality is needed to
convert a path to UNC before converting it. That does c_utf8_path_to_locale
now, while c_utf8_string_to_locale only converts the plain string, ie.
to generate wide char strings for output.
This enables the calling function to free these accordingly. That is needed
because the makeLongWinPath for efficiency reasons does not always realloc
the original string.
this fixes the error and makes complete oCC compile with GCC 5.
error: ISO C does not support '__FUNCTION__' predefined identifier
[-Wpedantic]
According to the porting guide:
The fix is either to use the standard predefined identifier __func__
(since C99), or to use the __extension__ keyword.
There could be a race condition if the file was updated on the server
between the discovery and the propagate phase. By taking the mtime from
the server, we make sure that we do not have a race.
This is tested by t6.pl with BIG3.file because the script was modifying
the file between the two phases
The fix in commit 86d8079 made the detection of changes inside
a directory more robust.
As a result, changes are detected inside a directory, even if the
files that are changed were moved away on the other remote, the algorithm
won't let the directory be removed and the directory will be re-created
Found while investigating #2296
The problem is that we should not remove a directory locally if it contains
modified files.
But the modification time of the directory is not necessarily chaning (so
the instruction of the directory may still be NONE)
We have to move the child_modified test a bit down to be recursive
