| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| | |
|
| |\
| |
| | |
Fix #24 NULL pointer dereference in ignore_ip.cpp line 151
|
| |/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Qt5 started to use libproxy and this exposed a crashing bug on OpenBSD
when using Qt5 applications on Gnome 3.
Here's a sample backtrace:
https://gist.github.com/mulander/c03997ae7e3fe4e30af0d4f2ce2965f0
The issue is caused by an unguarded dereference of ign_ip.
Other parts of the code properly expect that sockaddr_from_string
might return null and guard against it.
Adding a proper guard prevents a segfault both in OpenBSD
www/otter-browser and devel/zeal.
|
| |\
| |
| | |
Remove space in modules.cmk
|
| |/
|
|
|
| |
- Broke proper parsing of WITH_KDE option
- Fix for issue #20
|
| |\
| |
| | |
python: Avoid a crash on 64-bit systems
|
| |/
|
|
|
|
|
|
|
|
|
| |
Annotate the return type of px_proxy_factory_new() to be void *, as otherwise
int is assumed. This works fine on 32-bit systems, where void * and int are the
same width, but is invalid on 64-bit (Linux) systems.
Additionally, annotate the argument type of free() and px_proxy_factory_free()
to be void * to match.
https://code.google.com/archive/p/libproxy/issues/146
|
| |\
| |
| | |
config_kde improvements
|
| | | |
|
| | |
| |
| |
| |
| |
| |
| | |
After finding out whether to use kreadconfig5 or kreadconfig,
it uses either qtpaths or kde4-config to determine the locations
of the kioslaverc config file, to be able to notice modifications
that require a cache flush.
|
| |\ \
| |/
|/| |
travis: fix config
|
| |/ |
|
| |\
| |
| | |
bindings: perl: Add an option to explicitly link against libperl.so.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The Perl binding uses symbols from libperl.so, but since it is always
run after libperl.so itself is loaded, not linking against it does not
cause the problems that would normally occur.
Doing so was mandatory since 5077b75 ("bindings: perl: Do not link
against libperl"), which was catering for distros which install the
binding into vendor_perl and try not to rebuild all their Perl packages
on every Perl update. The flip-side is that it breaks things for other
groups: distros that install the binding into a versioned Perl
directory and allow different Perl versions to be installed
concurrently (for example, FreeBSD), or distros that pass
"--no-undefined" to the linker.
Provide an option, PERL_LINK_LIBPERL (whose default depends on the value
of PERL_VENDORINSTALL), to make the behavior configurable.
|
| |\ \
| |/
|/| |
INSTALL: update description of the WITH_KDE option.
|
| |/
|
|
|
| |
WITH_KDE4 no longer exists, and the KDE plugin now works with both KDE4
and KF5. Update the text in INSTALL to reflect those changes.
|
| |\
| |
| | |
travis: enable building of modules
|
| |/ |
|
| |\
| |
| | |
config_kde: Decide whether to build the module based only on WITH_KDE.
|
| | |
| |
| |
| |
| |
| |
| | |
kreadconfig/kreadconfig5 are runtime dependencies only. As such, there
should not be a build-time dependency on those in CMake at all. In other
words, building without those binaries works perfectly fine, and they
are only required when running libproxy.
|
| |\ \
| | |
| | | |
pacrunner_webkit: allow linking against javascriptcore-4.0
|
| | | | |
|
| |\ \ \
| |/ /
|/| | |
travis.yml: add OS matrix
|
| |/ / |
|
| |\ \
| | |
| | | |
Fix compilation errors with CLang on MacOSX
|
| | | | |
|
| |\ \ \
| |_|/
|/| | |
travis-ci: add config file
|
| |/ / |
|
| |\ \
| | |
| | | |
Make the KDE config module optional again.
|
| |/ /
| |
| |
| |
| |
| | |
This is a follow-up to bd9bf72 ("Add generic KDE config module"): make
it possible for distributions to not build the KDE config module at all,
even if it does not have any dependencies on Qt/KDE/KF5.
|
| | | |
|
| | | |
|
| |\ \
| |/
|/| |
Replace KDE4 config module by more generic one
|
| | |
| |
| |
| | |
- Does not depend on Qt, uses kreadconfig/kreadconfig5
|
| |/ |
|
| |\
| |
| | |
pxgsettings: g_signal_connect does not emmit changed signals
|
| |/
|
|
|
|
|
| |
In recent versions of GLIB, it is mandatory to FIRST connect to the changed:: signal and then read values.
Without reading the values, it seems there is an initialization missing to start the change monitoring.
So: we now first connect the signal handler, directly followed by reading all the settings we care about.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
As the binding is always running in context of perl, the
perl libraries are guaranteed to be in memory and we do not
have to take care of those.
This allows the perl bindings to be built against a different
version of perl as it will be finally run with (perl keeps on
installing the library in a versioned directory, hence on each
perl upgrade, a rebuild is needed in order to link against the
new location of the library).
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fix the way we detect if get_pac correctly aborted on our server trying
to send a > 102400 byte long stream (which we do to avoid malicious pac
files/wpad server to buffer overflow us).
Replace assert calls with proper error handling, which avoids the server
from hanging (the write command to send the quit was wrapped in
assert()).
When not building for a Debug release with CMAKE, the -DNDEBUG is being
specified, which in turn disables the assert macro.
Fixes bug 189.
|
| | |
|
| |
|
|
|
| |
g_object_unref() does a good enough job for us.
Fixes bug 192.
|
| | |
|
| |
|
|
|
|
|
| |
We have to assume that Mozilla knows what they are doing and thus, if we
find mozjs185.pc, we accept it as being present.
Fixes bug 188.
|
| | |
|
| |
|
|
| |
guidelines and forget GConf
|
| | |
|
| | |
|
| | |
|
| | |
|
