summaryrefslogtreecommitdiffstats
path: root/src/SocketReader.cpp
AgeCommit message (Collapse)AuthorFilesLines
2024-06-04Clean up includesDavid Robillard1-0/+1
2023-02-03Suppress/fix new warnings in clang-tidy 15David Robillard1-4/+4
2022-08-18Use consistent spacing for line commentsDavid Robillard1-4/+4
2022-08-18Avoid "else" after "return", "break", and "continue"David Robillard1-2/+6
2022-08-18Use default member initializationDavid Robillard1-5/+0
2022-08-18Clean up includesDavid Robillard1-0/+2
2022-08-18Clean up includesDavid Robillard1-1/+0
2021-01-02Update for latest raulDavid Robillard1-1/+1
2020-08-03Clean up includesDavid Robillard1-1/+0
As directed by include-what-you-use. Unfortunately, it is still confused by quite a few things, but this gets a lot closer to a clean report, and at least makes the output more tolerable to manually check.
2020-08-02Remove std::shared_ptr aliasDavid Robillard1-12/+12
2020-08-02Clean up URI map interfaceDavid Robillard1-1/+1
2020-08-02Remove wrappers for standard memory functionsDavid Robillard1-2/+2
2020-08-02Use auto with casts and allocations to remove redundancyDavid Robillard1-2/+2
2020-08-01Use modern castsDavid Robillard1-8/+14
2020-02-26Poll for POLLPRI as wellDavid Robillard1-1/+1
I don't think this actually matters, but I'm now paranoid about poll() not waking up on input for any reason.
2020-02-26Set up serd to read directly from socket without a FILE intermediaryDavid Robillard1-19/+27
This was problematic because using fdopen() conflicts with poll(). If the FILE ends up reading more than is actually processed by serd, then poll will not fire because the socket file descriptor has been fully read, even though there is pending input in the FILE object (which is buffered). Avoid this by using a custom read function and calling recv() directly. In retrospect, this was stupid, but it seemed convenient to be able to use the handy built-in support for reading from a FILE in serd. Now, the client and server are using send() and recv() directly on the socket, as things should be. I am not sure if MSG_WAITALL is the best idea here, or if it's really important at all, but it seems like a good idea.
2019-12-08Cleanup: Add missing explicit initialisationsDavid Robillard1-1/+2
2019-03-09Clean up AtomForgeDavid Robillard1-14/+7
2019-03-09Clean up includes and forward declarationsDavid Robillard1-6/+12
2019-03-09Localise dependency on boost::format and improve logging APIDavid Robillard1-4/+3
2019-03-08Use smart pointers to handle FILE streamsDavid Robillard1-4/+6
2018-09-23Use lowercase namespace namesDavid Robillard1-3/+3
2018-01-21Use C++ style includes for standard language headersDavid Robillard1-1/+2
2017-12-25Use std::move to potentially avoid copyingDavid Robillard1-1/+1
2017-12-25Use nullptrDavid Robillard1-8/+8
2017-12-24Clean up includes in public headersDavid Robillard1-0/+1
2017-02-12Fix various cast alignment warningsDavid Robillard1-7/+6
2016-07-29Remove Forge dependency from AtomReaderDavid Robillard1-5/+1
2015-10-02Unregister socket clients on hangup.David Robillard1-0/+1
git-svn-id: http://svn.drobilla.net/lad/trunk/ingen@5735 a436a847-0d15-0410-975c-d299462d15a1
2015-05-25Use ingen:/ as base URI on the wire.David Robillard1-3/+2
This allows referring to non-graph items, which are converted to bundle-relative URIs on save, resolving issue #1049. Change root graph URI to ingen:/graph. git-svn-id: http://svn.drobilla.net/lad/trunk/ingen@5687 a436a847-0d15-0410-975c-d299462d15a1
2015-05-23Prevent concurrent Sord access.David Robillard1-8/+23
git-svn-id: http://svn.drobilla.net/lad/trunk/ingen@5683 a436a847-0d15-0410-975c-d299462d15a1
2015-02-07Demodularize socket stuff.David Robillard1-0/+187
git-svn-id: http://svn.drobilla.net/lad/trunk/ingen@5535 a436a847-0d15-0410-975c-d299462d15a1