Review Request 121757: Use target environment whenever available, and implement TU pinning (WIP)

[Olivier Jean de Gaalon]

> On Jan. 8, 2015, 6:07 p.m., Milian Wolff wrote:
> > clangparsejob.cpp, line 314
> > <https://git.reviewboard.kde.org/r/121757/diff/4/?file=338962#file338962line314>
> >
> >     this should use the tuUrl here, no? and we'd get ast sharing for free then, if I'm not mistaken.

Can't do that as long as ParseSession has a url/file pointing to the parsejob that created it.


> On Jan. 8, 2015, 6:07 p.m., Milian Wolff wrote:
> > duchain/clangindex.h, line 34
> > <https://git.reviewboard.kde.org/r/121757/diff/4/?file=338965#file338965line34>
> >
> >     not really a multi hash you are using. do you want to?

Ah, I did originally, but you said in your last review to change to a QSet?


- Olivier Jean de


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://git.reviewboard.kde.org/r/121757/#review73516
-----------------------------------------------------------


On Jan. 7, 2015, 5:16 p.m., Olivier Jean de Gaalon wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://git.reviewboard.kde.org/r/121757/
> -----------------------------------------------------------
> 
> (Updated Jan. 7, 2015, 5:16 p.m.)
> 
> 
> Review request for KDevelop, Kevin Funk and Milian Wolff.
> 
> 
> Repository: kdev-clang
> 
> 
> Description
> -------
> 
> Note: This requires IBuildSystemManager::hasIncludesOrDefines, which should probably be renamed to isTranslationUnit()
> 
> The idea here is to use the AST of the buildsystem's translation units wherever possible, since that's the only file that's guaranteed to have correct defines and includes (and furthermore can handle strange cases such as '#include inside namespace' and picky include/define ordering).
> This proposal essentially allows TU environments to override non-TU environments and then get "pinned" so that the non-TU will continue to use the TU AST to build its duchain.
> This could eventually be exposed in the UI, allowing the user to pick different "views" of headers (overriding the pinned TU).
> The pinning data needs to be fleshed out to handle expiry (header no longer included) and be serialized with an ItemRepository.
> 
> I'm looking for feedback on this idea... where does it break? Any better ideas?
> 
> This more or less works, we can flesh it out more if we want to take this route.
> 
> 
> Diffs
> -----
> 
>   clangparsejob.h 1cc8282 
>   clangparsejob.cpp 297b836 
>   duchain/clanghelpers.h 4216454 
>   duchain/clanghelpers.cpp 5f29a2a 
>   duchain/clangindex.h 991c7ca 
>   duchain/clangindex.cpp 02c93cf 
>   duchain/clangparsingenvironment.h 7bb8111 
>   duchain/clangparsingenvironment.cpp 1decc14 
>   duchain/clangparsingenvironmentfile.h 953ee94 
>   duchain/clangparsingenvironmentfile.cpp b3d0563 
>   duchain/clangpch.cpp 1ce3457 
>   duchain/parsesession.h b688fb1 
>   duchain/parsesession.cpp f41768a 
>   duchain/tuduchain.cpp 33a0570 
>   tests/test_duchain.cpp 7db9fea 
> 
> Diff: https://git.reviewboard.kde.org/r/121757/diff/
> 
> 
> Testing
> -------
> 
> Tests fail in the same manner as before
> 
> 
> File Attachments
> ----------------
> 
> kdevelop patch
>   https://git.reviewboard.kde.org/media/uploaded/files/2015/01/02/5c0314bf-f241-4ad0-93f6-9a0a48a79135__kdevelop-hasincordef.diff
> platform patch
>   https://git.reviewboard.kde.org/media/uploaded/files/2015/01/02/edc11881-06c6-4570-9afd-824a99687116__kdevplatform-hasincordef.diff
> 
> 
> Thanks,
> 
> Olivier Jean de Gaalon
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kdevelop-devel/attachments/20150108/527f07fc/attachment.html>