I rewrote QFileSystemWatcher some time ago for it to be able to watch
non-existing directories, did rewrite KDirWatch too to not use FAM
and inotify making the created() and deleted() signals non-operational
(with the plan to maybe implement them in QFileSystemWatcher but that
will be hack-ish for directories with stat()-based implementation)
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
reducing the global statics used by KDebug to two - one for the null device
and one for the configuration and internal cache
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
most programs use 5-15 areas and if kdebugrc contains preferences for many
areas the initial caching process would take a lot of time
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
sub-optimal but recommended in glibc/manual/syslog.texi for shared
libraries, there is also the case of multiple KDebugSyslogDevice opening
the system log more than once per-program
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
for program that outputs something to stdout and other program reads it,
don't want programs to read debug messages
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
if I rewrite kBacktrace() there is nothing left not coded by me there, I
wrote backtrace printer for Katie with symbols demangling btw
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
it has the same effect as -ffile-prefix-map="${CMAKE_SOURCE_DIR}=." which
would affect not only KDebug if reproducible builds are a must (Debian) or
references to the build directory are not desired (Arch Linux). Debian
sets -ffile-prefix-map automatically by default as for Arch there may be
more warnings from `makepkg` after this change about references to source
directory but that's not something I have to look into
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
note that KLockFile is used only by KConfig (its INI backend), anything
not using KConfig to access files is not affected by the lock method
KLockFile uses
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
it has to solve a few problems like:
1. blocking the current thread while listing, extracting or adding to
archive
2. lack of support for some archive formats such as 7-zip in the
previously available classes for dealing with archive
3. progress and error reporting
a KArchiveJob class may have to be implemented to solve the above
issues. the API is just as frustrating to use as KTar and KZip right now
tho and I am not happy with that but the class itself is a 3-rd rewrite
of what was once Python module, plugin for another project and now used
as base in library.
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
