Re: r333874 - Reland "Move #include manipulation code to new lib/Tooling/Inclusions."

2018-06-04 Thread Nico Weber via cfe-commits
(Cost as in "shared library buildsof clang are ~35% slower than regular clang" last we checked.) On Mon, Jun 4, 2018 at 2:14 PM, Nico Weber wrote: > Sure, the bots could do that, the more general point was that many > (probably most) people who want to enable asserts don't want to use shared >

Re: r333874 - Reland "Move #include manipulation code to new lib/Tooling/Inclusions."

2018-06-04 Thread Nico Weber via cfe-commits
Sure, the bots could do that, the more general point was that many (probably most) people who want to enable asserts don't want to use shared libs. I believe that only a few people use that config, so turning things around why should everyone building with asserts have to pay the cost of

Re: r333874 - Reland "Move #include manipulation code to new lib/Tooling/Inclusions."

2018-06-04 Thread David Zarzycki via cfe-commits
Hi Roman – I agree that people shouldn’t need to manually fix this every time it happens. Hi Nico – What is stopping the Chromium buildbots from setting BUILD_SHARED_LIBS to false? That way the default BUILD_SHARED_LIBS policy of llvm and/or clang doesn’t matter, right? Dave > On Jun 4,

Re: r333874 - Reland "Move #include manipulation code to new lib/Tooling/Inclusions."

2018-06-04 Thread Roman Lebedev via cfe-commits
This is n-th time i do this kind of fix. Sure this does resolve the issue, but it would be much better to use proper means, and tell people to fix their code, in an automatic way :) On Mon, Jun 4, 2018 at 7:36 PM, Nico Weber wrote: > All chromium bots build with asserts on and we definitely

Re: r333874 - Reland "Move #include manipulation code to new lib/Tooling/Inclusions."

2018-06-04 Thread Nico Weber via cfe-commits
All chromium bots build with asserts on and we definitely don't want shared libs :-) Having the people who care about this build config make sure it keeps working seems like a good approach to me. On Mon, Jun 4, 2018 at 8:15 AM, David Zarzycki via cfe-commits < cfe-commits@lists.llvm.org> wrote:

Re: r333874 - Reland "Move #include manipulation code to new lib/Tooling/Inclusions."

2018-06-04 Thread David Zarzycki via cfe-commits
What’s stopping the project from automatically setting BUILD_SHARED_LIBS to TRUE if asserts are enabled? That’d automatically “force” active developers to test the scenario on their local machines, but not change the way release builds work. > On Jun 4, 2018, at 8:11 AM, Roman Lebedev via

Re: r333874 - Reland "Move #include manipulation code to new lib/Tooling/Inclusions."

2018-06-04 Thread Eric Liu via cfe-commits
Thanks for fixing this Roman! On Mon, Jun 4, 2018 at 2:11 PM Roman Lebedev wrote: > This broke building as shared libraries. > > I've committed the fix, but the bot coverage is clearly severely missing, > given that such breakages are very often. > > Roman. > > On Mon, Jun 4, 2018 at 12:04 PM,

Re: r333874 - Reland "Move #include manipulation code to new lib/Tooling/Inclusions."

2018-06-04 Thread Roman Lebedev via cfe-commits
This broke building as shared libraries. I've committed the fix, but the bot coverage is clearly severely missing, given that such breakages are very often. Roman. On Mon, Jun 4, 2018 at 12:04 PM, Eric Liu via cfe-commits wrote: > Author: ioeric > Date: Mon Jun 4 02:04:12 2018 > New Revision:

Re: r333874 - Reland "Move #include manipulation code to new lib/Tooling/Inclusions."

2018-06-04 Thread David Zarzycki via cfe-commits
FYI – This broke `cmake -DBUILD_SHARED_LIBS=TRUE` builds. > On Jun 4, 2018, at 5:04 AM, Eric Liu via cfe-commits > wrote: > > Author: ioeric > Date: Mon Jun 4 02:04:12 2018 > New Revision: 333874 > > URL: http://llvm.org/viewvc/llvm-project?rev=333874=rev > Log: > Reland "Move #include