Note with -fretain-comments-from-system-headers, the .gch of bits/stdc++.h becomes larger by 1%, but that is fine.
And improve comment handling in textDocument/hover
Fix#373
By default, the background indexer doesn't handle names of no linkage.
They are indexed when their files are opened. This saves memory and
makes cache files smaller.
If the workspace folder is a symlink and the client doesn't follow it.
Treat /tmp/symlink/ as canonical and convert every /tmp/real/ path to
/tmp/symlink/.
Combined Daan De Meyer's #227 with other simplification
* USE_SHARED_LLVM is deleted in favor of LLVM_LINK_LLVM_DYLIB
* LLVM_ENABLE_RTTI is deleted as it is provided by LLVMConfig.cmake
* Only direct Clang/LLVM dependencies are required in target_link_libraries
* Restrict -DCLANG_RESOURCE_DIRECTORY= to src/utils.cc
cache.hierarchicalPath: store cache files as $directory/a/b/c.cc.blob to
work around NAME_MAX limitation.
cache.retainInMemory: after this number of loads, keep a copy of file
index in memory. If set to 1, it avoids cache corruption if the index
file is changed after the initial load, which may happen if several
language clients open the same project and share the same cache
directory.
Also rename cacheDirectory cacheFormat to cache.{directory,format}
In clients that support DiagnosticRelatedInformation, display
clang notes as these nested diagnostics rather than appending
them to the parent diagnostic's message. Behaviour for clients
that don't support related information should be unchanged.
indexer.cc: use index.maxInitializerLines instead of kInitializerMaxLines
messages/initialize.cc: some ServerCapabilities are toggable:
documentOnTypeFormattingProvider.firstTriggerCharacter
foldingRangeProvider
workspace.workspaceFolders.supported
Fix#190
If a new header is added, the preamble size changes. Language clients may cache completion results, thus we rebuild preamble to avoid inaccurate results.
Fix#180
index.initialBlacklist: ["."] can inhibit initial indexing (useful for larger code bases).
Opened files are still indexed, though.
This heuristic allows related files (a/foo.c a/b/foo.cc) to be indexed when a header (foo.h) is opened.
This fixes a plethora of "not indexed" errors when the document has not been indexed.
* Message handler throws NotIndexed if not overdue
* The message is put into backlog and tagged with backlog_path
* path2backlog[path] tracks backlog associated with document `path`
* The backlog is cleared when the index is merged
* backlog[0] is forced to run if it becomes overdue
* In the "initialized" callback, send client/registerCapability with DidChangeWatchedFilesRegistrationOptions
* In workspace/didChangeWatchedFiles callback, call pipeline::Index
* In pipeline::Index, add a `deleted` status
* Add %h for C header files (the suffix .h is considered a C header, not a C++ header)
* Add %hpp for C++ header files
* If .ccls exists, it provides full command line for files not specified by compile_commands.json (before, compile_commands.json was ignored)
* If the first line of .ccls is %compile_commands.json, it appends flags to compile_commands.json "arguments", instead of overriding.
Files not specified by compile_commands.json will not be added to folder.entries, but their command line can be inferred from other files.
Also fix `#include <` completion of -I flags for clang < 8
In completion, underscore prefixed builtin macros may be annoying when the first type character is not an underscore.
When `strict` is true, `Match` enforces the first characters should be loosely of the same category.
