Community
Participate
Working Groups
Build Identifier: I20110613-1736 I have set up a remote C/C++ project containing the Linux kernel source. Even though I've carefully set up the remote include paths, the indexer is not finding the include files. For example, I get an Unresolved Inclusion error for the line: #include <linux/interrupt.h> which resides in one of the C source files. I have tried adding both a remote absolute path to where linux/interrupt.h resides, and also a workspace-relative path and neither one worked. Both resulted in the same unresolved inclusion errors. A related problem is that after quite a lengthy indexing time, the indexer spends at least five minutes spilling these errors out to the error log, and there appears to be no way of canceling this job. Reproducible: Always Steps to Reproduce: 1. get the latest Linux kernel source from https://github.com/torvalds/linux.git and set it up as a remote project using Remote Tools. 2. Filter out all of the arch/* directories, except the one you are working on. Also filter out the drivers directory, because that will speed up the indexing considerably. 3. Add the remote include paths of <workspace>/include <workspace>/include/linux and <workspace>arch/<your_arch>/include. 4. Do an index operation. It should take 5-10 minutes (plus the added time for the $*(@# Photran indexer), after which you should see all of the unresolved inclusion errors.
I should add that I'm seeing these "Unresolved inclusion: ..." errors in the Error Log view.
Until recently the GNU scanner discovery stuff was pretty broken, and that will mean things would not get parsed right unless you painstakingly set up all of the include paths, and *macros* at the proper values. SR1 should help with that a lot. I would suggest trying out a newer build... a June build is pretty old at this point. If you still have issues then we can dig further.
Are you still seeing this with the properly matched CDT/RDT?
No, this problem went away once I switched to the nightly build of the CDT. Should we wait for SR1 to come out, then re-test?
I'm going to mark this as WORKSFORME then. If you see this in the official SR1, then feel free to reopen it.