summaryrefslogtreecommitdiff
path: root/libproxy
Commit message (Collapse)AuthorAgeFilesLines
...
* | get-pac-test: Fix build with clang/libc++Khem Raj2019-01-301-1/+1
|/ | | | | | | get-pac-test.cpp:55:10: error: assigning to 'int' from incompatible type '__bind<int &, sockaddr *, unsigned int>' ret = bind(m_sock, (sockaddr*)&addr, sizeof (struct sockaddr_in)); ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Signed-off-by: Khem Raj <raj.khem@gmail.com>
* Revert "Do not invoke extra shell to run pxgsettings"Dominique Leuenberger2018-10-031-1/+1
| | | | | | | | | | | | | | """ sigh OK I now see why it was using /bin/sh to start the helper. It's because the code is concatenate the gsettings keys to watch (in all_keys) as parameters So this is not enough to just remove /bin/bash """ This reverts commit 8012bed4405efa31fffb57d6282de3e1513e4bd9.
* Add call to JS::InitSelfHostedCode()Laurent Bigonville2018-10-021-0/+2
| | | | This is needed otherwise mozjs crashes
* Build with mozjs-52Pierre Lejeune2018-07-022-13/+8
| | | | Fixes #71
* Do not invoke extra shell to run pxgsettingsDominique Leuenberger2018-01-081-1/+1
| | | | Fixes issue #57
* Add version number to the version scriptFabian Vogt2017-09-201-1/+1
|
* Add public px_proxy_factory_free_proxies functionFabian Vogt2017-09-203-1/+20
| | | | Fixes issue #43.
* Add symbol versions - be ready to introduce new APIs as neededDominique Leuenberger2017-09-072-0/+12
| | | | | APPLE uses a linker derived from llvm, which does not support --version-script, so ignore it there (sorry guys).
* get-pac-test: Fix typoDavid King2017-07-241-1/+1
| | | | Closes: #64
* Use closesocket() instead of close() on WindowsJérôme Laheurte2017-05-231-3/+5
|
* Version 0.4.150.4.15Dominique Leuenberger2017-05-111-1/+1
|
* Merge branch 'master' into masterJeremy Linton2017-04-2611-19/+94
|\
| * Merge pull request #61 from bigon/bug-59Dominique Leuenberger2017-02-071-0/+1
| |\ | | | | | | Set lasturl to NULL after free/delete
| | * Set lasturl to NULL after free/deleteLaurent Bigonville2017-02-071-0/+1
| | | | | | | | | | | | | | | | | | | | | It avoids the lasturl to be freed a 2nd time when the rewind() method is called. Closes: #59
| * | Use StateChanged signal instead of StateChangeLaurent Bigonville2017-02-071-1/+1
| |/ | | | | | | | | | | | | | | StateChange signal has been deprecated since NM 0.7 and has been completely removed in 0.9, see: https://developer.gnome.org/NetworkManager/0.9/ref-migrating.html Closes: #58
| * fix spelling mistakeEdward Betts2017-02-021-1/+1
| |
| * Install libs/cmake/pkgconfig thingy into right place for MinGWĽubomír Carik2017-02-021-2/+2
| | | | | | | | | | | | | | | | MinGW folder structure is closer to Linux. This allow use other MinGW package to utilize CMake or pkgconfig to setup the project. Signed-off-by: Ľubomír Carik <Lubomir.Carik@gmail.com>
| * Use different name when checking for 2nd set NetworkManager packagesDominique Leuenberger2017-01-182-2/+5
| | | | | | | | | | | | | | | | | | | | | | | | Seems different to Autoconf, cmake does keep some cache around and does not transparently reuse the name fully to make a complete test. This for one resulted in old versions of NetworkManager being queried, yet the final build was disabled. If cmake was ran a 2nd time the cache was refreshed and the NetworkManager module being built. By using a 2nd name we aboid the internal confusion for cmake. Fixes issue #53
| * Prepare for 0.4.14Dominique Leuenberger2017-01-091-1/+1
| |
| * Merge pull request #48 from amigadave/get-pac-testDominique Leuenberger2017-01-091-11/+16
| |\ | | | | | | Fix intermittent test failures in get-pac-test
| | * get-pac-test: Fix overflow testDavid King2017-01-091-6/+16
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Rather than checking for EBADF when the socket is closed due to an overly-large request, check that the sent size differs from the received. Use a large buffer to ensure that the number of bytes sent are greater than the kernel TCP receive buffer size, as otherwise the send() call succeeds even though the bytes were never retrieved from the socket. Pass the MSG_NOSIGNAL flag (or the SO_NOSIGPIPE socket option) so that SIGPIPE is not generated when the socket is closed once the content length has been read. Close #47
| | * get-pac-test: Call close() on an fdDavid King2017-01-011-5/+0
| | | | | | | | | | | | | | | | | | | | | Do not call close() on the return value from send(), which is a number of bytes. Fixes #47
| * | pxgsettings: Avoid key listing deprecationDavid King2017-01-031-0/+7
| | | | | | | | | | | | | | | g_settings_list_keys() is deprecated, and g_settings_schema_list_keys() should be used instead as of GLib 2.46.
| * | pxgsettings: Fix memory leak of key namesDavid King2017-01-031-0/+1
| | | | | | | | | | | | | | | The key names returned by g_settings_list_keys() are annotated as being transfer full, so must be freed after use.
| * | pxgsettings: Avoid g_type_init deprecation warningDavid King2017-01-031-0/+2
| |/ | | | | | | g_type_init() was deprecated in GLib 2.36.
| * Add a test that ensures that query parameters are parsed and sentPaul van Tilburg2016-09-191-0/+9
| |
| * Use the query string (parameters) when requesting PAC filesPaul van Tilburg2016-09-192-4/+20
| |
| * Add test for a PAC get without content lengthPaul van Tilburg2016-09-191-0/+21
| |
| * Handle PAC files without specified HTTP content lengthPaul van Tilburg2016-09-191-3/+8
| |
* | Update build to mozjs38Jeremy Linton2017-04-262-39/+54
| | | | | | | | | | | | | | | | This pull libproxy forward a little more to mozjs38, which is where gnome is currently at. This means most distro's will already be shipping this version of mozjs. Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
* | Why isn't SO_NOSIGPIPE defined when OSX manpage says its in sys/types.h, ↵Jeremy Linton2016-09-191-1/+3
| | | | | | | | sys/socket.h
* | Try slightly diffrent osx buildJeremy Linton2016-09-181-4/+4
| |
* | Fix intermittent unit test failures.Jeremy Linton2016-09-171-6/+12
| | | | | | | | | | | | | | | | | | | | | | There are two things wrong with the unit test. First depending on the state of the socket the overflow send may terminate with either EBADF or EPIPE. Second, the length of the send is being used as the fd in the call to close. Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
* | Update build to mozjs24Jeremy Linton2016-09-172-24/+27
|/ | | | | | | This pulls libproxy forward a little, and brings it in line with gnome which is also on mozjs24. Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
* GNOME3: Wait for pxgsettings to vanish while destroying the pluginDominique Leuenberger2016-07-221-0/+2
| | | | | | | | This helps avoiding zombie processes in case a caller creates/destroys ProxyFactories for each URL (even though we recommend to have a long-living PF for caching reasons). Originally reported at https://bugzilla.opensuse.org/show_bug.cgi?id=967601
* config_kde: Don't pollute stderrFabian Vogt2016-05-111-1/+3
| | | | - https://bugzilla.opensuse.org/show_bug.cgi?id=979232
* config_kde: Remove use of C++11Fabian Vogt2016-05-111-12/+14
| | | | | | - Use std::map::find instead of map::at (issue #29) - Add additional safety check - Return as const ref
* NetworkManager: detect libnm before looking for NetworkManagerDominique Leuenberger2016-05-021-2/+5
| | | | | | | Prefer the newer libnm symbol and only fall back to the legacy NetworkManager if not found Fixes https://github.com/libproxy/libproxy/issues/26 - 2nd part
* NetworkManager: gain compatibility to be built against libnm unstead ofDominique Leuenberger2016-05-021-0/+6
| | | | the legacy libnm-glib (we only use the dbus interface)
* NetworkManager: the location of NetworkManager.h is specified by pkg-configDominique Leuenberger2016-05-011-1/+1
| | | | | | | The directory structure of NetworkManager can change over time. pkg-config informs us of the correct location to find headers at. NM does not add subdirectories on its own. https://github.com/libproxy/libproxy/issues/26
* Version 0.4.130.4.13Dominique Leuenberger2016-04-281-1/+1
|
* Fix #24 NULL pointer dereference in ignore_ip.cpp line 151Adam Wolk2016-04-181-1/+1
| | | | | | | | | | | | | | | | 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.cmkFabian Vogt2016-03-231-1/+1
| | | | | - Broke proper parsing of WITH_KDE option - Fix for issue #20
* Fix mismatched new[]/delete in pacrunner_natus and _webkitFabian Vogt2016-02-172-5/+5
|
* config_kde: Add a basic cache and invalidationFabian Vogt2016-02-171-10/+88
| | | | | | | 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.
* Merge pull request #12 from rakuco/kde-runtime-dependency-onlyDominique Leuenberger2016-01-182-9/+3
|\ | | | | config_kde: Decide whether to build the module based only on WITH_KDE.
| * config_kde: Decide whether to build the module based only on WITH_KDE.Raphael Kubo da Costa2016-01-182-9/+3
| | | | | | | | | | | | | | 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-4.0Dominique Leuenberger2016-01-181-3/+6
| |
* | Merge pull request #3 from andyetitmoves/mac-warnDominique Leuenberger2016-01-181-1/+3
|\ \ | |/ |/| Fix compilation errors with CLang on MacOSX
| * Fix compilation errors with clangRamkumar Aiyengar2015-12-301-1/+3
| |
View Lorry Controller Status