| RE: [cdt-dev] Toolchain Builtins and Indexing |
|
Yes, the indexer doesn’t really know
what’s a built in versus a user include path or macro. And, yes, the
indexer only indexes header files that are not part of the project when they
are included by files in the project. Our “Full” indexer operates
just like a compiler and will parse header files every time they included by a
file that has changed. The new “Fast” indexer (which is introduced
as part of the PDOM feature) will only parse each header file once. Doug Schaefer, QNX Software Systems Eclipse CDT Project Lead, Tools PMC member From:
cdt-dev-bounces@xxxxxxxxxxx [mailto:cdt-dev-bounces@xxxxxxxxxxx] On Behalf Of Sennikovsky, Mikhail Hi Stan, From the MBS point of
view, the option “built-ins” information is taken into account only
in case the tool-chain does not have the scanner discovery profile associated with
it, otherwise the scanner discovery profile is used for builtin calculation.
Include paths and defined symbols are presented to the core via the PathEntry
framework, which, AFAIK does not contain any information of whether the given
include path/symbol is built-in or not. As for indexing, Doug
might correct me if I’m wrong, includes are parsed only as they are
referenced in the source code, and with the new PDOM functionality the index
information is stored in the database and includes that are already parsed and
indexed are not reparsed each time the include is referenced or the source code
is modified. Regards, Mikhail From:
cdt-dev-bounces@xxxxxxxxxxx [mailto:cdt-dev-bounces@xxxxxxxxxxx] On Behalf Of Stan Mitchell Include paths and defined symbols can be tagged as
“built-ins” for a toolchain, to indicate that these are not changed
by the user. How does the Indexer treat these include paths and
symbols differently – or does it? I would expect the indexer to parse and index builtin
includes just once; only the user portion of a project would be reparsed
when source changes are made. Is that a fair assumption? Regards, Stan Mitchell |