91616abdb0
random values of state->error_buffer_read might otherwise cause memory corruption as we write beyond the stderr receive buffer, or (if state->error_buffer_read > ERROR_BUFFER_SIZE), the size argument to recv might overflow to very large unsigned values, causing the recv call to fail (and hence causing spsupr to reattempt reading from the stderr fd indefinitely) |
||
|---|---|---|
| admin | ||
| build-aux | ||
| doc | ||
| etc | ||
| leim | ||
| lib | ||
| lib-src | ||
| lisp | ||
| lwlib | ||
| m4 | ||
| modules | ||
| msdos | ||
| nextstep | ||
| nt | ||
| oldXMenu | ||
| src | ||
| test | ||
| .clang-format | ||
| .dir-locals.el | ||
| .gitattributes | ||
| .gitignore | ||
| .gitlab-ci.yml | ||
| .mailmap | ||
| autogen.sh | ||
| BUGS | ||
| ChangeLog.1 | ||
| ChangeLog.2 | ||
| ChangeLog.3 | ||
| ChangeLog.4 | ||
| config.bat | ||
| configure.ac | ||
| CONTRIBUTE | ||
| COPYING | ||
| GNUmakefile | ||
| INSTALL | ||
| INSTALL.REPO | ||
| make-dist | ||
| Makefile.in | ||
| README | ||
| readme.md | ||
Emacs for lsp-mode(POC)
A Emacs fork implementing non-blocking and async JSONRPC support
Motivation
The fork aims to fix the strugle of lsp-mode with sync Emacs core and json handling
How it works?
The fork uses separate emacs threads to run the processing and then runs release_global_lock() from the C code when the processing does not involve lisp objects. There are a lot of benefits from this approach:
-
UI does not block the server. Imagine that currently you have font lock running. In stock
emacsthat will prevent reading the process intput even if the server has alredy sent the response to the client. Note that this will improve not only emacs performance but it will improve the performance of single threaded server because the server won't be blocked to wait for IO to be read by the client. -
Server does not block the UI. Similarly, the processing of
lsp-modeserialization of requests/deserialization of responses does block UI processing. Only small portion of wholeJSONRPC -
Server being slow reading requests does not block the UI . In stock
emacssending requests to the server will forceemacsto wait for server to read the request. -
Less garbage. Since a lot of the processing does not involve lisp objects we generate less garbage and as a result the GC runs less often.
Current state
The code runs fine with most of the servers I have tested with. Only Linux/Unix is supported for now.
How to use?
Compile emacs just like normal emacs and then use the latest version of lsp-mode.
Acknowledgments
Thanks to 606u for helping me out with low level process communication code